Social Security as a Retirement Resource for Near-Retirees, by Race and Ethnicity, Nativity, Benefit Type, and Disability Status

by Benjamin Bridges and Sharmila Choudhury
ORES Working Paper No. 109 (released October 2007)

This paper analyzes Social Security benefits as a retirement resource for selected subgroups of current and recent cohorts of near-retirees. The paper examines the distribution of benefits among (1) several race-ethnic subgroups, (2) the native-born and the foreign-born, (3) worker, spouse, and survivor beneficiaries, and (4) the disabled and the nondisabled. We use improved data (actual earnings history data) to produce more accurate measures of benefits. We look at how the average values of several benefit measures such as Social Security wealth and earnings replacement rates differ among the selected subgroups and discuss reasons for these differences. We find that substantial differences in earnings levels and/or mortality levels among these subgroups interact with Social Security program provisions to produce sizable differences in the values of our benefit measures.


The authors are respectively with the Division of Economic Research and the Division of Policy Evaluation, Office of Research, Evaluation, and Statistics, Office of Policy Social Security Administration.

Acknowledgments: The authors thank Rob Gesumaria for programming assistance and Paul Davies, Linda Del Bene, Ed DeMarco, Howard Iams, Dean Leimer, Michael Leonesio, Joyce Manchester, Linda Maxfield, and Mark Sarney for constructive comments.

Working papers in this series are preliminary materials circulated for review and comment. The findings and conclusions expressed in them are the authors' and do not necessarily represent the views of the Social Security Administration.

Summary

This paper provides an in-depth examination of one component of retirement resources, Social Security benefits, for specific subgroups of current and recent near-retirees. It examines the distribution of benefits among (1) several race-ethnic subgroups that include non-Hispanic whites, non-Hispanic blacks, Asians, and Hispanics, (2) the native-born and the foreign-born, (3) worker, spouse, and survivor beneficiaries, and (4) the disabled and nondisabled. Our choices of subgroups are driven by the longstanding interest by policymakers in many of these subgroups as well by the need to address the conflicting or missing empirical evidence with regard to these subgroups.

The paper considers benefits for people who turn age 61 during the 1993–2007 period. We choose age 61 because it is the last age before the age of first eligibility for Social Security retired-worker and spouse benefits, which is age 62. We compute a variety of benefit measures (Social Security wealth, annualized benefit payouts, and earnings replacement rates), some of which have not been used in previous studies. We rely primarily on actual earnings history data in computing streams of benefits. The use of observed earnings histories allows us to capture the large variation in these histories, unlike methods that estimate earnings histories on the basis of a single earnings equation. The study uses the Modeling Income in the Near Term (MINT) data files, which include Social Security Administration (SSA) administrative earnings and benefit history records exact-matched to the 1990–1993 panels of the U.S. Census Bureau's Survey of Income and Program Participation. Measuring benefits in innovative ways and using improved data, the paper explores in detail the benefits of subgroups who command considerable interest.

What are the effects of various economic, demographic, and Social Security program factors on the differences in benefit measures of these subgroups? Some of our results have been reported in the literature. For example, we report that among race-ethnic subgroups, because of their higher indexed taxable earnings, whites receive the highest amounts of Social Security wealth and annualized payouts. Taxable earnings replacement rates are the lowest for whites and higher for minority race-ethnic subgroups, because of the progressivity of the Social Security benefit formula. Immigrants of all race-ethnic subgroups receive lower average Social Security wealth and annualized payouts than the native-born primarily because of their lower indexed taxable earnings. When we look at persons by longest-held benefit type, survivor beneficiaries receive the highest amounts of Social Security wealth, because of their much longer lives. In contrast, because of their markedly shorter lives, the disabled, as defined in our paper, receive considerably less in median amounts of Social Security wealth than do the nondisabled.

Some other interesting findings emerge from our study of these subgroups. For example, over time Hispanics have very slow growth in Social Security wealth compared with that of the other race-ethnic subgroups. A key underlying variable is the growth in earnings. Median indexed taxable earnings increases are considerably smaller for Hispanics than for the other three race-ethnic subgroups. For immigrants, the taxable earnings replacement rate is not a very good measure of how effective Social Security is in replacing average career earnings; this is especially so for Asians who have the highest average age of entry into the United States. Age of entry into the country is an important variable. Immigrants who enter before age 23 have benefits similar to those of the native-born. When we consider benefit types, we find that the small subgroup of male spouse beneficiaries receive substantially lower Social Security wealth than any other male or female benefit-type subgroup. Female survivor beneficiaries, on the other hand, receive significantly higher wealth amounts than any of the seven other benefit-type subgroups. In addition, compared with the other race-ethnic subgroups, a larger share of black beneficiaries receives disability and/or survivor benefits.

I. Introduction

Social Security benefits are the major retirement resource (wealth and income) for U.S. retirees. In 2004, 66 percent of aged beneficiary units (those aged 65 or older) received at least half of their income from these benefits, while for 34 percent, these benefits accounted for at least 90 percent of their income. These benefits were especially important for low earners and for certain population subgroups such as blacks, Hispanics, and widows. Moreover, benefits are now almost universal. The proportion of the aged units receiving Social Security benefits rose from 69 percent in 1962 to 89 percent in 2004.1

The purpose of this paper is to analyze Social Security benefits as a retirement resource for selected subgroups of interest among the population of near-retirees. The subgroups that are considered to be vulnerable when studying the economic well-being of the older population have, in many instances, been racial and ethnic minorities, immigrants, the disabled, and survivor beneficiaries. How they fare under Social Security is of interest to policy makers and researchers who seek to understand the well-being of the elderly. Also, the benefit outcomes for these subgroups acquire additional importance when the program is projected to become financially insolvent. Improved understanding of the benefit outlook under current Social Security law for these at-risk subgroups should be helpful to policymakers in evaluating proposals made to change current law in response to the long-term solvency outlook or other considerations.

This paper examines the distribution of benefits among (1) several race-ethnic subgroups that include non-Hispanic whites, non-Hispanic blacks, Asians, and Hispanics, (2) the native-born and the foreign-born, (3) worker, spouse, and survivor beneficiaries, and (4) the disabled and nondisabled. It examines benefits for today's near-retirees and for earlier cohorts of near-retirees. The near-retirees in this study are people who turn age 61 during the 1993–2007 period. We choose age 61 because it is the last age before the age of first eligibility for Social Security retired-worker and spouse benefits, which is 62. The paper shows how the average values of several benefit measures (Social Security wealth, annualized benefit payouts, and earnings replacement rates) differ among the selected subgroups. These measures include only benefits received by persons when they are aged 62 or older. We look at some reasons for these differences and discuss the effects of various economic, demographic, and Social Security program factors on these benefit measures.

The Social Security program provides monthly benefits to qualified retired and disabled workers and to their dependents and survivors. To qualify for benefits, a worker must have at least a specified amount of work in covered employment. For those who qualify for benefits, the benefit amount increases, but less than proportionally, with lifetime taxable earnings in covered employment. In other words, the benefit formula is progressive. (The worker pays payroll taxes on these earnings.) Benefit payments to near-retirees usually continue until they die. Although under Social Security law a person's benefits do not depend on the person's race, ethnicity, nativity, or sex, substantial differences in earnings levels and/or mortality levels by these characteristics can produce sizable differences in Social Security benefit levels among these subgroups.

Our choices of subgroups are driven by the longstanding interest by the policymaking community in these subgroups. They are also driven by our desire to address the conflicting claims made with regard to some subgroups, as with race-ethnic minorities, as well as by the lack of sufficient empirical evidence for other subgroups, as with immigrants and various Social Security beneficiary types. We briefly provide some information about our chosen subgroups.

With regard to race-ethnic subgroups, a common theme in distributional analyses is that Social Security benefits are important to most race-ethnic minorities. For example, according to a report based on a U.S. Census Bureau survey in 2004 about half of black and Hispanic aged beneficiary units received 90 percent or more of their income from Social Security.2 Studies have shown that these particular race-ethnic subgroups tend to have lower earnings, on average, and thus are helped by the progressivity of the Social Security benefit formula. Some minority groups, for example as with blacks, participate to a greater extent than other race-ethnic groups in Social Security's Disability Insurance program. Yet, it has been pointed out that blacks, on average, have shorter life spans, which mean fewer years of benefit receipt.

Another issue is how the foreign-born fare under Social Security when compared with the native-born. Not much research has been done on this issue. A worker's Social Security benefit depends on the worker's lifetime taxable earnings in employment covered by Social Security. In computing an immigrant's lifetime taxable earnings, the work-years spent outside the United States are treated under Social Security law, in the great majority of cases, as years in noncovered employment and hence as years of zero taxable earnings. Because many immigrants have considerable earnings outside the United States, this program feature lowers the benefits of the immigrant subgroup relative to those of the native-born subgroup. However, the progressivity of the benefit formula partially offsets the effect of this zero-earnings feature. The importance of this feature depends on the age at which immigrants enter the United States. This issue is particularly relevant for the large Hispanic minority and the smaller Asian minority, both subgroups of which have substantial shares of foreign-born members.

Social Security provides benefits to distinct beneficiary categories. Among adults, the program provides benefits to disabled workers, retired workers, spouses, and surviving widow(er)s. There have been a number of proposals to revise the treatment of auxiliary beneficiaries, mainly spouses and survivors.3 The paper examines benefits separately for these beneficiary designations to show how each of these subgroups fares. How the disabled fare in the retirement years compared with the nondisabled has been of continuing concern, even more so as policymakers advocate reforming the current Social Security program.

The focus here is the availability to various subgroups of Social Security benefits as a retirement resource and not on issues related to money's worth which concern the relation of benefits received to taxes paid. The paper builds on our previous work that focused on intercohort changes in Social Security benefits of near-retirees but which did not disaggregate results for the subgroups described above.4 The benefit measures used here are affected primarily by lifetime earnings, marital histories, mortality, and benefit rules. Because many of the differences in Social Security benefit outcomes for the selected subgroups are associated with these underlying factors, an attempt will be made to assess the role that these factors play in driving these differences. The sizeable overlaps among these various subgroups will be considered in the paper.

This paper attempts to provide clear and comprehensive answers regarding one component of retirement resources, that is, Social Security benefits. It provides an in-depth examination of Social Security benefits for specific subgroups of current and recent near-retirees. We compute a variety of benefit measures that have not been used in previous studies. We rely primarily on actual earnings history data in computing streams of benefits. The use of observed earnings histories allows us to capture the large variation in these histories, unlike methods that estimate earnings histories on the basis of a single earnings equation. The study uses the Modeling Income in the Near Term (MINT) data files, which include Social Security Administration (SSA) administrative earnings and benefit history records exact-matched to the 1990–1993 panels of the United States Census Bureau's Survey of Income and Program Participation (SIPP). Because of the extensive content of this data set, we are able to use fewer imputations and projections than have a number of other studies. Any imputations and projections that were required were done by MINT modelers using sophisticated analytical methods. Measuring benefits in innovative ways and using improved data, the paper explores in detail the benefits of subgroups who command considerable interest.

The following is an outline of the paper: section II discusses the data, section III explains the various benefit measures that are used; and sections IV–VII present empirical analyses for the selected subgroups. The last section has concluding observations.

II. Data

As stated earlier, for this study we use data from the Modeling Income in the Near Term (MINT) model project.5 The MINT project is a large-scale effort that has been underway since the late 1990s. Much of the developmental work was done for the Social Security Administration by analysts at the Urban Institute, RAND Corporation, and Brookings Institution. The starting sample is from the 1990, 1991, 1992, and 1993 panels of the Census Bureau's Survey of Income and Program Participation. In this survey of the noninstitutionalized population, interviews were conducted once every 4 months for 28–36 months. The initial SIPP interviews were conducted in 1990–1993 and almost all of the final SIPP interviews were conducted during the 1992–1995 period. The SIPP collected information on income and wealth components, mortality, marital histories, institutionalization, immigration, various demographic and socioeconomic variables (for example, race, ethnicity, nativity, and education), and many other variables.

As part of the MINT project, SSA administrative records were exact-matched to SIPP data for sample members born during the 1926–1965 period. These administrative records include earnings history, benefit history, and death information through 1999;6 these records also include sex and date of birth. Exact-matches were made for about 92 percent of the sample members, and administrative records were imputed by MINT modelers for the remaining 8 percent of the sample. Thus, we have SIPP data through the 1992–1995 period and administrative data through 1999. For years after this time range, the MINT model projects institutionalizations, marital histories, dates of death, earnings histories, and benefit histories, using information from both SSA administrative records and the SIPP. In addition, persons are projected to enter the sample by means of immigration. These economic and demographic projections were designed to be generally consistent with the intermediate assumptions of the 2002 Old-Age, Survivors, and Disability Insurance (OASDI) Trustees Report.7 Additional information about MINT imputations and projections is given in Appendix A of Bridges and Choudhury (2005). For a detailed description and evaluation of the MINT3 model and data, see Toder and others (2002). Also see Panis and Lillard (1999) for a detailed description and evaluation of the MINT projections of marital histories, mortality, and disability status.

For our study, we use a subset of the MINT sample members born during the 1932–1946 period. The data set used in this study has notable strengths. First, longitudinal administrative data are available through 1999. Thus, earnings history data are available through age 53 for the youngest birth cohort analyzed (those born in 1946) and through age 67 for the oldest birth cohort (born in 1932). Benefit record information is available for the great majority of members of the three oldest cohorts (born 1932–1934) and for many members of the next three cohorts (born 1935–1937). Second, the combined SIPP panels provide a large sample. Each of our single-year birth cohorts is represented by a sample of more than 1,000 persons. Studies of retirement resources of near-retirees typically use much smaller samples.

III. Definitions and Benefit Measures

This section discusses the empirical constructs of the study: the definitions of cohorts of near-retirees and of program participants and the benefit measures (Social Security wealth, annualized payout, and earnings replacement rates).

Cohorts of Near-Retirees

The unit of analysis is the person and not some larger unit such as a marital unit or family. In studies that use longitudinal data, the person is often the unit of analysis. The composition of the larger units changes over time. For example, the marital status of most persons changes one or more times during their adult lifetime.

The paper looks at 15 single-year cohorts, that is, those persons attaining age 61 during the period from calendar year 1993–2007. Each single-year cohort consists of all persons who reach age 61 during that year and are members of the noninstitutionalized population at the end of that year; that is, at the beginning of the year most of them can first receive Social Security retirement benefits. Each of the four SIPP panels (1990–1993) includes persons from each of our 15 single-year cohorts.

To facilitate the presentation of results and to avoid small sample sizes for certain subgroups, we combine these 15 single-year cohorts into three groups of five single-year cohorts. The first and oldest cohort of near-retirees, the 1993 cohort, combines five single-year cohorts of persons who reach age 61 in 1993–1997. The 1998 cohort combines persons who reach 61 in 1998–2002, and the last cohort, the 2003 cohort, consists of persons reaching age 61 in 2003–2007. From here onward, whenever we use the term cohort, we refer to these 5-year groups. When we refer to single-year cohorts we will use the term single-year cohort. Benefits of cohort members are evaluated as of January 1 of the year they reach 62. To increase comparability among subgroups within a cohort and among cohorts, benefits of all members of a particular cohort are evaluated as of the year they reach a given age (62) rather than as of a given year (for example, 1993). All measures are in 2002 constant dollars.

Benefit Measures

In our study all benefit amounts are those payable under actually enacted Social Security law. In our benefit calculations we assume that the program provisions in effect in future years are those scheduled under current law. The most recent significant change in Social Security law, a change in the earnings test, was enacted in 2000.

Our benefit concept is shared benefits. For each year a person is married, the person's shared benefit equals one-half the benefits received by the couple. It is our view that shared benefit is superior to individual benefit received as a measure of the income support the person receives from the OASDI program. The individual benefits of husband and wife often are quite different. However, most married couples share their incomes.8 For each year a person is not married, the person's shared benefit equals the benefits received by the person.9

Our benefit measures such as Social Security wealth include benefits received in the year the person attains age 62 and in all later years. They do not include any benefits received before the year the person attains age 62. We focus on the support provided by Social Security to persons over their post age-61 years. We do not attempt to measure the support provided over a person's lifetime. Our measures include the benefits paid from the Old-Age and Survivors (OASI) and Disability Insurance (DI) trust funds to a worker, spouse, divorced spouse, surviving spouse, or surviving divorced spouse.

Social Security Wealth. For each person with benefits, we compute Social Security wealth—the present value of shared benefits evaluated as of January 1 of the year the person reaches age 62. Real Social Security wealth (SSW) is expressed in January 1, 2002, dollars.10 Our annual discount rate series consists of the rates of return on OASI trust fund assets.11 Projected Consumer Price Index for Urban Wage Earners and Clerical Workers (CPI-W) values and trust fund interest rates are based on the intermediate assumptions of the 2002 Trustees Report.

Social Security wealth is a measure of the total support provided by Social Security to a person over the period from the year the person attains age 62 until the person's death. The value of a person's SSW depends importantly on the person's longevity and past and future (projected) marital history.

Annualized Social Security Wealth Payout. For each person with benefits, we compute an annualized Social Security wealth payout that is equal to the constant real annual payment over all the person's potential benefit years that has a present value equal to the person's Social Security wealth. In other words, the person's SSW is converted into an annuity, which provides constant real annual payments over the person's potential benefit years. As with SSW, annualized payout is expressed in January 1, 2002, dollars. All years from the year the person reaches age 62 through the last year before the year of death are potential benefit years.12,13 The person's number of potential benefit years is the maximum number of years (starting with the year the person reaches age 62) that the person could receive benefits. After 1999, the year of death is projected by the MINT model.

Annualized payout, which has not been used in previous studies, is a useful measure of the average annual support provided by Social Security after age 61.14 It is less affected by differences within cohorts or increases over cohorts in longevity than is the Social Security wealth measure.15 We use annualized payout as the numerator of our earnings replacement rates.

Earnings and Replacement Rates. There are a number of possible earnings replacement rate measures. For example, replacement rates have been defined as the percent of average earnings for the last few years before benefit receipt that are replaced by benefits. Instead, our replacement rates measure the extent to which average career earnings are replaced by benefits. In the next four paragraphs, we discuss how we arrive at our two career earnings measures, average wage-indexed taxable earnings, and average wage-indexed less-censored earnings, before we go on to describe our two earnings replacement rates.

The annual taxable earnings (wages and self-employment income) of a worker is that part of the worker's total earnings from employment covered by Social Security that is at or below the legislated taxable maximum (the maximum amount of annual earnings that is subject to Social Security payroll tax and is included in the calculation of benefits). For each year after 1981, the legislated taxable maximum has been indexed by SSA's U.S. average annual wage series. Therefore, since 1983 the ratio of the legislated taxable maximum to the average annual wage has been roughly constant at about 2.3 to 2.5. The ratio was 2.3 to 2.4 during the 1983–1989 period and 2.4 to 2.5 during the 1990s. Before 1983, this ratio was always below 2.3 and varied substantially. The ratio was 1.0 to 1.7 during the 1951–1978 period and 2.0 to 2.2 during the 1979–1982 period.16

We compute a measure of earnings that is less censored than taxable earnings and that unlike taxable earnings has censoring limits that are a constant percentage of average annual wage series amounts. The annual less-censored earnings of a worker is that part of the worker's total earnings from employment covered by Social Security that is estimated to be at or below a hypothetical taxable maximum that for each year was set at about 2.45 times the average annual wage. The SSA earnings records included in our MINT data file include annual amounts of taxable earnings, but not amounts of total covered earnings. For each year before 1990, MINT modelers estimate covered earnings in excess of the legislated taxable maximums using SSA administrative data on quarters of coverage and Current Population Survey wage data.17 The 1951–1989 hypothetical maximums are then applied to these estimated earnings to get less-censored earnings. For years after 1989, less-censored earnings are simply set equal to taxable earnings; for these years the legislated taxable maximums were 2.4 to 2.5 times the average annual wage. For each year of the 1951–1989 period the hypothetical maximum exceeds the legislated maximum; for each year of this period less-censored earnings are less censored than taxable earnings. We believe that less-censored earnings are superior to taxable earnings in approximating relative differences in total earnings both within cohorts among subgroups and across cohorts.

We compute average taxable wage-indexed earnings as follows. For each person, shared taxable earnings for each year of the computation period are indexed, using the average wage series, to wage levels as of the beginning of the year the person reaches age 62. The indexed earnings are then averaged over the person's computation period. Finally, this average is expressed in January 1, 2002, dollars to get our measure of average wage-indexed shared taxable earnings.18 For average wage-indexed shared taxable earnings, we often use the term indexed taxable earnings. The computation period for these indexed taxable earnings begins with 1951 or the year the person reaches age 22, whichever comes later, and it ends with the year the person reaches age 61. In the computation of indexed taxable earnings for immigrants who enter the United States after 1950 and after they reach age 22, all years before their year of immigration are treated as years of zero earnings. Projected average annual wages in the MINT data file are based on the intermediate assumptions of the 2002 Trustees Report.

Average wage-indexed shared less-censored earnings are computed in an analogous way.19 For average wage-indexed shared less-censored earnings we often will use the term indexed less-censored earnings. The computation period for indexed less-censored earnings begins with 1951, the year the person reaches age 22, or the year the person immigrates to the United States, whichever comes later; the period ends with the year the person reaches age 61. Thus, except for immigrants who enter the United States after 1950 and after the year they reach age 22, the computation periods for indexed less-censored earnings are the same as those for indexed taxable earnings. For such immigrants, the computation periods for indexed less-censored earnings are shorter than those for indexed taxable earnings.

For each person with some shared earnings, we calculate two earnings replacement rates—one for average wage-indexed shared taxable earnings and another for average wage-indexed shared less-censored earnings. Given that the numerator of our earnings replacement rates, annualized payout, is a shared benefit measure, we need shared earnings measures for the denominators of these replacement rates.20 One reason for selecting measures of average wage-indexed career earnings for the replacement rate measures is because one goal of the Social Security program is to provide benefits that replace a portion of a measure of average wage-indexed career earnings. In addition, for a given single-year cohort, average wage-indexed career earnings provides a useful indicator of a worker's average position over their career in the economy's earnings distribution. We present results for the taxable earnings replacement rate because this rate and the replacement rate measure implicit in Social Security law have some similar features. The less-censored earnings replacement rate is our proxy for a total earnings replacement rate; it is superior to the taxable earnings replacement rate as a measure of the adequacy of Social Security benefits because its denominator is a better proxy for the person's average preretirement standard of living.

A person's taxable earnings replacement rate is the person's annualized payout expressed as a percentage of the person's indexed taxable earnings. The following features are common to our taxable earnings replacement rate and the replacement rate measure implicit in Social Security or OASDI law. Under that law, a person's initial benefit is determined as a percentage of the person's average wage-indexed monthly earnings (average indexed monthly earnings or AIME), and over time the person's initial benefit is kept constant in real terms. The numerator of the taxable earnings replacement rate is the annualized payout, which is a constant real benefit and is related to the price-indexed OASDI initial benefit. The denominator of the taxable earnings replacement rate is average indexed taxable earnings from age 22 through age 61. Indexed taxable earnings and OASDI average indexed monthly earnings have some similar features, but differ in several ways. Both are indexed using the SSA average annual wage series, and their averaging periods are similar.21 The same AIME computation procedure applies to all of our cohorts of near-retirees.

The less-censored earnings replacement rate is the percentage of indexed less-censored earnings replaced by Social Security benefits. As stated earlier, the less-censored earnings replacement rate is our proxy for a total earnings replacement rate; it is superior to the taxable earnings replacement rate as a measure of the adequacy of Social Security benefits because its denominator is a better proxy for the person's average standard of living over their work career. For both immigrants and the native-born, indexed less-censored earnings is superior to indexed taxable earnings as a proxy for a person's average standard of living over their work career because the former earnings measure includes earnings up to a constant relative taxable maximum and is less censored. In addition, for immigrants the average less-censored measure has the advantage that its computation period does not include any years before the year of immigration that are treated as years of zero earnings. Bear in mind, however, that indexed less-censored earnings of immigrants who enter the United States at quite different ages cover quite different portions of these immigrants' work lives.

Both the taxable and less-censored earnings replacement rates are age-62 replacement rates, that is, they give the percentages of a person's earnings wage-indexed to January 1 of the year the person reaches age 62 that are replaced by the person's constant real annualized payout. As average real economy-wide earnings increase in the years after age 61, the person's annualized payout declines relative to average economy-wide earnings.

Program Participants and Beneficiaries

In this paper, Social Security program participants are near-retirees with some shared earnings (that is, with positive indexed taxable earnings), and those with no shared earnings are nonparticipants. Social Security beneficiaries are those with both shared indexed earnings and shared benefits (that is, with positive Social Security wealth and annualized payouts). For each of the three cohorts, 95.2 percent to 95.6 percent of program participants are beneficiaries. The very small group of nonparticipants (about 1 percent of our sample) is excluded entirely from this paper's analyses. In our results for race-ethnic groups and for immigrants and native-born, we include program participants regardless of whether they have shared benefits, that is, our tables include participants who have positive indexed taxable earnings but receive no benefits—nearly always because of employment histories that are not strong enough to qualify for benefits or because they die before claiming benefits. On the other hand, the tables for persons classified by benefit type and by disability status provide data for beneficiaries only; program participants with no shared benefits are excluded from these tables.

IV. Findings by Race-Ethnic Subgroups

We present results for selected race-ethnic subgroups. We are able to classify near-retirees into a larger number of race-ethnic subgroups than is typically available. Of particular note is our inclusion of a category for Asians. Hispanics, who may be of any race, are a separate category. Thus, our subgroups are: (1) whites (non-Hispanic whites), (2) blacks (non-Hispanic blacks), (3) Asians (non-Hispanic Asians and Pacific Islanders), (4) Hispanics, and (5) "others" (non-Hispanic American Indians, Eskimos, and Aleuts). We present some results for these subgroups disaggregated by sex.22

This section's tables present data for Social Security program participants. Note that this paper's analysis deals only with persons who live to at least age 61 and only with the shared benefits they receive after the year they reach age 61.

We briefly examine a few demographic characteristics of our sample. As seen in Table A-1, whites account for 79 percent to 81 percent (81 percent of the 1993 cohort, 81 percent of the 1998 cohort, and 79 percent of the 2003 cohort) of our near retirees. Blacks, Asians, Hispanics, and "others" account for 9 percent, 3 percent to 4 percent, 7 percent to 8 percent, and less than 1 percent, respectively. In our tables, the "other" subgroup is not shown separately, but is included in calculating numbers for the totals that combine all subgroups.

Table A-1. Percentage of near-retirees with selected characteristics, by cohort and race-ethnicity
Characteristic and cohort All White Black Asian Hispanic
Male
1993 48 49 42 51 49
1998 48 48 42 53 50
2003 48 49 44 48 48
Foreign-born
1993 10 6 5 77 41
1998 10 5 7 79 48
2003 12 6 7 77 48
Entered United States at age 23 or older
1993 7 3 4 68 30
1998 6 2 5 67 34
2003 8 3 4 66 34
Married at age 62
1993 74 76 58 83 70
1998 73 74 61 86 73
2003 71 73 59 76 71
Total number of near-retirees (thousands)
1993 10,033 8,123 898 268 674
1998 11,115 9,032 960 296 752
2003 13,911 11,030 1,250 521 1,045
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).
NOTE: The "others" subgroup is not shown separately because it contains less than 1 percent of near-retirees; it is included in the "All" column.

Looking into characteristics by race-ethnicity, we see that the percentage of males is lowest for blacks (42 percent to 44 percent) and a bit higher for whites, Asians, and Hispanics at 48 percent to 49 percent, 48 percent to 53 percent, and 48 percent to 50 percent, respectively. The percentages of those married at age 62 are higher for whites (73 percent to 76 percent) and Asians (76 percent to 86 percent) than for Hispanics (70 percent to 73 percent) and blacks (58 percent to 61 percent). As is to be expected, large percentages of Asians (77 percent to 79 percent) and Hispanics (41 percent to 48 percent) immigrated to the United States—most of them as adults; the comparable percentages for whites (5 percent to 6 percent) and blacks (5 percent to 7 percent) are much smaller.23,24 We will discuss the impact of these subgroup differences in immigration on our results.

Social Security Wealth

Social Security wealth is the present value at age 62 of Social Security benefits received from age 62 until death. Thus, SSW depends importantly on projected longevity. Among the variables used in projecting MINT mortality beyond 1999 are sex, earnings, education, marital status, disability benefit status, and race (white and black). The Hispanic and other race (mostly Asian) classifications are used only in projecting deaths before age 65. For the 1993, 1998, and 2003 cohorts, projected deaths account for 94 percent, nearly 100 percent, and 100 percent of all deaths, respectively. Thus, MINT-based estimates of longevity and of SSW may not be as accurate for Hispanics and Asians as for whites and blacks.25

Median Social Security wealth is highest for whites primarily because they have the highest median indexed taxable earnings.26 For example, the wealth levels of blacks are 72 percent to 74 percent of those of whites (Table A-2). In addition, whites live longer than blacks. High indexed taxable earnings produce high annual benefits. Longer lives result in more years of benefit receipt. The other three subgroups have median indexed taxable earnings equal to 51 percent to 71 percent of those of whites. Among the minority subgroups for the two youngest cohorts, Hispanics have the lowest indexed taxable earnings and blacks have the highest. Blacks have mean number of potential benefit years equal to 84 percent to 89 percent of those of whites. The preceding general patterns also hold for each sex; for example, among women and among men, SSW is highest for whites.

Table A-2. Social Security benefit measures and related measures for near-retirees, by cohort and race-ethnicity
Measure and cohort All White Black Asian Hispanic
Social Security wealth (median, 2002 dollars)
1993 122,258 129,451 93,772 92,589 90,689
1998 147,003 156,568 116,291 116,134 99,231
2003 164,961 178,168 129,261 126,076 99,980
Annualized payout (median, 2002 dollars)
1993 6,338 6,463 5,756 5,020 5,456
1998 7,487 7,676 6,712 5,504 5,778
2003 8,292 8,588 7,578 6,019 5,959
Taxable earnings replacement rate (median, percent)
1993 33.9 33.2 41.0 35.6 38.4
1998 32.2 31.4 37.0 32.4 38.6
2003 31.0 30.0 37.3 34.3 38.0
Less-censored earnings replacement rate (median, percent)
1993 30.6 29.7 38.9 24.3 35.2
1998 30.0 29.5 35.7 23.6 33.4
2003 29.5 28.8 36.3 25.6 34.9
Taxable earnings (median, 2002 dollars)
1993 18,454 19,676 13,032 13,519 13,919
1998 22,915 24,305 17,084 15,970 14,178
2003 26,198 28,534 18,913 17,433 14,578
Less-censored earnings (median, 2002 dollars)
1993 20,276 21,743 13,645 19,313 14,657
1998 24,437 25,997 17,555 20,482 15,799
2003 27,237 29,581 19,631 21,985 16,426
Years of benefit receipt (mean)
1993 20.22 20.76 17.89 17.85 17.60
1998 20.62 21.20 17.99 18.91 17.76
2003 21.02 21.72 18.23 19.76 17.56
Years of potential benefit receipt (mean)
1993 21.47 21.89 19.39 20.76 19.32
1998 21.97 22.36 19.69 22.68 20.01
2003 22.34 22.89 19.30 22.94 19.90
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).
NOTE: The "others" subgroup is not shown separately because it contains less than 1 percent of near-retirees; it is included in the "All" column.

Other things being equal, subgroups with higher proportions of immigrants will have lower median indexed taxable earnings for beneficiaries and higher proportions of program participants who are nonbeneficiaries. We have seen in Table A-1 that the Asian and Hispanic subgroups contain very high proportions of immigrants. A reason that immigrants have lower median indexed taxable earnings than the native-born is that for many immigrants their computation periods for indexed taxable earnings begin before they immigrate; in the computation of indexed taxable earnings all such years before the year of immigration are treated as years of zero earnings. The computation period for indexed taxable earnings begins with the later of 1951 or the year the person reaches age 22. For example, immigrants who entered the United States in 1980 at age 35, will have their earnings for ages 22–34 set to zero. These 13 years of zero earnings are included in computing their lifetime indexed taxable earnings. The majority of immigrants (62 percent to 66 percent) enter the United States after the year they reach age 22.

Another important feature regarding immigrants and their benefits is that relatively more of them have employment histories that are not strong enough to qualify for benefits. The percentage of program participants with no Social Security wealth is 4 percent for whites, 5 percent to 7 percent for blacks, and 8 percent to 9 percent of Hispanics. The percentage for Asians is generally similar to that for Hispanics.

We do not display results by sex here (available upon request). We find that for each race-ethnic subgroup, median Social Security wealth is greater for women than for men. The main causes for this difference are (1) that women have much higher average number of years of benefit receipt and (2) our use of a shared concept of wealth rather than an individual concept.27 The ratio of Social Security wealth of women to that of men differs among race-ethnic subgroups. This ratio is highest for Hispanics.

When we look at changes from the 1993 cohort to the 2003 cohort, the relative increase in median Social Security wealth is much smaller for Hispanics than the increases for the other three racial-ethnic subgroups. Again, a key underlying variable shows similarly large differences. The relative increase in median indexed taxable earnings for Hispanics is much smaller than the increases for the other subgroups.28

Annualized Social Security Wealth Payout

Our annualized payout is a measure of the average annual support in real dollars provided by Social Security over the years beyond age 61. It is computed by spreading Social Security wealth over all potential benefit years. The effects of errors in the mortality projections for Hispanics and Asians on estimates of annualized payout for these subgroups should be relatively small because errors in Social Security wealth should be largely offset by errors in number of potential benefit years.

Again, as with Social Security wealth, the median annualized payout is highest for whites, driven primarily by their higher indexed taxable earnings. For the remaining other subgroups, annualized payouts are 69 percent to 89 percent of those of whites. Blacks have the second highest annualized payouts (87 percent to 89 percent of those of whites), while Hispanics and Asians have the lowest.

Reporting on results by sex (not shown here), we find that these general patterns hold for both men and women; for example, among women and among men, annualized payouts are highest for whites. For race-ethnic subgroups, the annualized payouts of women and men are similar. The ratios of these payouts for women to those for men are in the .94 to 1.18 range. The corresponding ratios for Social Security Wealth are higher than those for annualized payouts because of the longevity differences between men and women.

From the 1993 cohort to the 2003 cohort, the relative increase in median annualized payout is much smaller for Hispanics than for whites and blacks, as seen in Table A-2. We do not show the results separately by sex, but the relative increases for Hispanics are also much smaller than for whites and blacks for both men and women.

Taxable Earnings Replacement Rate

Our taxable earnings replacement rate measures the extent to which annualized payout replaces average indexed taxable earnings. The rate is somewhat like the replacement rate measure implicit in OASDI law.29

As seen in Table A-2, median taxable earnings replacement rates are lowest for whites, while those of the other subgroups are 103 percent to 127 percent of those of whites.30 Asians have the second lowest taxable earnings replacement rates, and blacks and Hispanics have the highest. Note that median indexed taxable earnings of whites are much higher than those of the other subgroups. Differences in median indexed taxable earnings among the other subgroups are usually not large. Thus, the progressivity of the Social Security benefit formula is an important reason why taxable earnings replacement rates of whites are lower than those of the other subgroups.

From the 1993 cohort to the 2003 cohort, median taxable earnings replacement rates of whites and blacks fall considerably, by 10 percent and 9 percent. Rates are almost unchanged for Hispanics.31 Indexed taxable earnings of both whites and blacks rose about 45 percent over this period. On the other hand, indexed taxable earnings increased only 5 percent for Hispanics. This differential earnings growth interacted with Social Security's progressive benefit formula to produce much of the above difference in intercohort movement of earnings replacement rates.

Again, we briefly point to some results by race and sex that are not included in the tables. Median taxable earnings replacement rates of whites, blacks, and Hispanics are considerably higher for women than for men. A primary reason for this is that, for whites and blacks, median indexed taxable earnings are markedly lower for women than for men. White women have lower replacement rates than do blacks and Hispanics of either sex.

Less-Censored Earnings Replacement Rate

Our measure of less-censored earnings replacement rates tells us the extent to which annualized payout replaces average indexed less-censored earnings, our proxy for total earnings. As seen in Table A-2, median less-censored replacement rates are lowest for Asians, ranging from 24 percent to 26 percent. They are second lowest for whites, ranging from 29 percent to 30 percent. Thus, less-censored earnings replacement rates of Asians are 80 percent to 89 percent of those of whites; while those of blacks and Hispanics are higher at 121 percent to 131 percent and 113 percent to 121 percent of those of whites.

Why are less-censored earnings replacement rates for Asians low relative to those of the other race/ethnic subgroups? One can look to how their less-censored earnings replacement rates compare with their taxable earnings replacement rates. The ratio of less-censored earnings replacement rate to taxable earnings replacement rate is only 0.68 to 0.75 for Asians compared with 0.87 to 0.97 for the other three subgroups. That is, the two earnings replacement rates are quite different for Asians. This is driven by the relatively large difference between their indexed less-censored earnings and indexed taxable earnings. The ratio of median indexed less-censored earnings to median indexed taxable earnings is much higher for Asians (1.26 to 1.43) than for the other three subgroups (1.03 to 1.11). Immigrating after age 22 is a key reason why indexed less-censored earnings is greater than their indexed taxable earnings because the computation of indexed less-censored earnings does not include years before immigration. Some two-thirds of Asian near retirees are adult immigrants. Only 2 percent to 5 percent of whites and blacks are adult immigrants. Of Hispanic near-retirees about one-third are adult immigrants. Therefore, for Asians in particular, due to the wedge between their indexed less-censored and indexed taxable earnings, the taxable earnings replacement rate measure is not a very good measure of how effective Social Security is in replacing average career earnings.

The general patterns of race-ethnic differences in less-censored earnings replacement rates also hold for each sex in tabulations that are not shown here. For example, among women and among men, less-censored earnings replacement rates are lowest for Asians and second lowest for whites.

Section Summary

We find that because of their higher indexed taxable earnings, whites, as a group, receive more Social Security wealth and annualized payout than other race-ethnic groups. The lower indexed taxable earnings of Asians and Hispanics are due, in large part, to the fact that many of them immigrate to the United States as adults; program rules assign zero earnings to years before immigration. In addition, whites have more years of benefit receipt than blacks because they live longer. Certain aspects of the Social Security program, such as the progressive benefit formula, advantage those with lower lifetime earnings. Thus, blacks, Hispanics, and Asians have higher taxable replacement rates than whites because those groups have lower lifetime taxable earnings than whites. For Asians (a group with a very high proportion of immigrants), this taxable earnings replacement-rate measure is not a very good measure of how effective Social Security is in replacing average career total earnings. The indexed taxable earnings of Asians are particularly low relative to their indexed less-censored earnings, which is our proxy for indexed total earnings, because of a large number of years of zero taxable earnings for earnings that were received before entering the United States. Other race-ethnic groups do not exhibit as large differences between these two earnings replacement rates as do Asians. With regard to sex differences (results not displayed here), women in every race-ethnic group have considerably higher Social Security wealth than men, because they live longer. In terms of Social Security wealth, Hispanic women fare particularly well relative to Hispanic men.

V. Findings by Immigrant Status

In this section, we consider the following: How do immigrants fare under Social Security compared with the native-born? How do Social Security outcomes for immigrants differ among race-ethnic subgroups? How does age at time of immigration affect Social Security outcomes for immigrants?32

The starting MINT sample is from the 1990, 1991, 1992, and 1993 panels of the SIPP. Members of this starting sample were asked their year of immigration and source country. In addition, persons are projected to enter the MINT sample by means of immigration in the years after the end of the SIPP interview. Imputed immigrants account for roughly 3 percent of immigrants in the 1993 cohort of near-retirees, 9 percent in the 1998 cohort, and 15 percent in the 2003 cohort.33,34

This section's tables show results for program participants. Nonparticipants (near retirees with no shared earnings) account for less than 0.5 percent of the native-born, but for 6 percent to 10 percent of immigrants. Immigrants account for 10 percent to 12 percent of all Social Security program participants (Table B-1).

Table B-1. Percentage of near-retirees with selected characteristics, by immigrant status and cohort
Characteristic Immigrant Native-born
1993 1998 2003 1993 1998 2003
Male 48 50 49 48 47 48
Married at age 62 77 79 76 74 73 71
Race-ethnicity
White 47 42 39 85 86 85
Black 5 6 5 9 9 9
Asian 21 20 25 1 1 1
Hispanic 27 31 31 4 4 4
Education
Dropout 36 32 29 24 17 13
High school graduate 40 42 44 58 62 59
College graduate 24 26 28 19 21 27
Age entered United States
Up to 23 34 38 35 100 100 100
23–32 27 25 26 0 0 0
33–42 21 17 19 0 0 0
43–52 12 12 13 0 0 0
52–61 5 9 7 0 0 0
Total number of near-retirees (thousands) 996 1,151 1,610 9,037 9,964 12,301
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).

Among immigrants, about 50 percent are Asian or Hispanic, whereas these subgroups comprise only about 5 percent of our native-born population. Correspondingly, among immigrants, about 39 percent to 47 percent are white and 5 percent to 6 percent are black compared with about 85 percent and 9 percent among the native-born. The sex compositions of the immigrant and native-born subgroups are quite similar. For immigrants, the proportions that are married are slightly higher and the proportions that are divorced are lower. Relative to the native-born, a larger share of immigrants are high school dropouts or college graduates. This means that a smaller share of immigrants are in the middle category of being only high school graduates. In other words, immigrants have several characteristics that are distinct from those in the general native-born population.

About a third of immigrants enter the United States before they reach age 22. Less than 10 percent enter the United States after age 53. Table B-1 shows that the majority of immigrants in our cohorts enter the United States during their prime working years.

Social Security Wealth

Immigrants have much lower median indexed taxable earnings than the native-born. This results in median Social Security wealth of immigrants falling short of that of the native-born.35 The relative shortfall has increased over time (Table B-2).36 For the 1993, 1998, and 2003 cohorts, median indexed taxable earnings of immigrants are 20 percent, 33 percent, and 44 percent lower than those of the native-born. One reason immigrants have lower indexed taxable earnings is that their computation periods for these earnings begin before many of them immigrate.37 Relatively more immigrants have employment histories that are insufficiently strong to qualify for benefits.

Table B-2. Social Security benefit measures and related measures for near-retirees, by immigrant status and cohort
Measure Immigrant Native-born
1993 1998 2003 1993 1998 2003
Social Security wealth (median, 2002 dollars) 99,838 109,737 108,101 125,681 151,789 172,338
Annualized payout (median, 2002 dollars) 5,456 6,018 5,849 6,403 7,601 8,478
Taxable earnings replacement rate (median, percent) 34.8 33.9 33.9 33.8 32.1 30.7
Less-censored earnings replacement rate (median, percent) 27.0 26.0 27.2 31.2 30.4 29.7
Taxable earnings (median, 2002 dollars) 14,981 15,757 15,274 18,802 23,596 27,723
Less-censored earnings (median, 2002 dollars) 19,064 19,937 19,420 20,394 24,859 28,294
Years of benefit receipt (mean) 18.6 18.4 18.7 20.4 20.9 21.3
Years of potential benefit receipt (mean) 21.2 21.4 21.6 21.5 22.0 22.4
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).

Among immigrants, whites have greater median Social Security wealth than other groups (Table B-3). It is highest for whites because they have the highest median indexed taxable earnings and because they live longer than most other race-ethnic groups. The other subgroups have median indexed taxable earnings equal to 48 percent to 80 percent of those of whites. Median Social Security wealth of white immigrants falls a bit short of that of the native-born (all race-ethnic subgroups combined).

Table B-3. Social Security benefit measures and related measures for near-retiree immigrants, by cohort and race-ethnicity
Measure and cohort White Black Asian Hispanic
Social Security wealth (median, 2002 dollars)
1993 118,566 85,235 84,424 71,664
1998 140,795 72,433 104,593 70,876
2003 143,061 70,801 113,717 76,649
Annualized payout (median, 2002 dollars)
1993 6,178 4,578 4,437 4,476
1998 7,202 5,311 5,105 4,850
2003 7,430 5,702 5,294 4,805
Taxable earnings replacement rate (median, percent)
1993 34.1 31.6 34.1 36.9
1998 31.1 37.8 34.2 42.5
2003 30.5 32.0 36.2 40.6
Less-censored earnings replacement rate (median, percent)
1993 27.5 24.5 23.0 29.9
1998 25.8 24.7 23.1 29.5
2003 26.1 26.5 24.8 31.8
Taxable earnings (median, 2002 dollars)
1993 18,294 14,576 11,423 11,495
1998 21,824 12,207 12,672 10,965
2003 22,297 13,581 14,579 10,768
Less-censored earnings (median, 2002 dollars)
1993 22,536 18,395 18,879 13,778
1998 26,003 19,674 19,483 13,849
2003 26,066 19,639 20,558 14,407
Years of benefit receipt (mean)
1993 20.2 17.1 17.5 16.9
1998 20.3 17.5 18.6 15.9
2003 21.0 15.7 19.2 15.8
Years of potential benefit receipt (mean)
1993 22.3 20.3 21.1 19.6
1998 22.7 20.4 22.8 19.0
2003 23.4 17.8 22.9 18.9
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).

From the 1993 cohort to the 1998 cohort, median Social Security wealth of immigrants increases substantially for whites and Asians, but is virtually unchanged for Hispanics. For the 1993–2003 period, the percentage increases in median SSW are larger for whites and Asians than for Hispanics.38

Among immigrants, median Social Security wealth declines markedly as age at entry into the United States increases (Table B-4).39 For example, median Social Security wealth is zero for the subgroup with age at entry of 53–61, indicating that at least 50 percent of this subgroup have no SSW. Median indexed taxable earnings decreases as age at entry increases.40 As age at entry increases there is a corresponding increase in the number of years in the computation period for indexed taxable earnings that are treated as years of zero earnings. The share of program participants with some shared benefits falls from 95 percent to 98 percent for those who enter before age 33 to 39 percent to 44 percent for those who enter at ages 53–61. Note that median Social Security wealth of immigrants who immigrate before age 23 is similar to that of the native-born.

Table B-4. Social Security benefit measures and related measures for near-retiree immigrants, by cohort and age at entry to United States
Measure and cohort Age at entry
Under 23 23–32 33–42 43–52 53–61
Social Security wealth (median, 2002 dollars)
1993 129,171 108,507 101,214 39,473 0
1998 158,459 120,244 116,599 40,502 0
2003 159,154 134,555 88,070 38,236 0
Annualized payout (median, 2002 dollars)
1993 6,608 5,769 5,368 3,078 0
1998 7,411 6,709 5,208 2,881 0
2003 7,747 7,241 4,709 2,326 0
Taxable earnings replacement rate (median, percent)
1993 36.0 33.1 35.9 36.3 0
1998 32.9 35.0 38.4 45.8 0
2003 32.7 33.6 39.1 40.8 0
Less-censored earnings replacement rate (median, percent)
1993 32.3 27.5 23.6 22.6 0
1998 31.1 30.4 24.1 19.2 0
2003 31.9 29.2 26.7 19.8 0
Taxable earnings (median, 2002 dollars)
1993 19,250 17,376 13,409 7,199 1,052
1998 23,553 19,103 12,981 5,401 999
2003 23,165 20,518 11,477 4,658 1,020
Less-censored earnings (median, 2002 dollars)
1993 20,678 20,691 19,313 11,476 4,876
1998 24,636 21,330 19,937 13,618 5,782
2003 23,757 24,154 16,755 12,210 5,857
Years of benefit receipt (mean)
1993 21.1 18.9 19.2 14.7 7.8
1998 20.8 20.1 20.1 13.6 6.2
2003 21.6 20.8 17.6 14.5 6.7
Years of potential benefit receipt (mean)
1993 22.1 20.9 21.0 19.7 22.1
1998 22.1 21.3 22.1 19.9 19.7
2003 22.5 22.1 20.3 20.9 20.2
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).

Annualized Social Security Wealth Payout

Just as with Social Security wealth, the lower median indexed taxable earnings of immigrants causes the median annualized payout of immigrants to fall short of that of the native-born. This relative gap has also increased over time. For the 1993, 1998, and 2003 cohorts, median annualized payouts of immigrants are 15 percent, 21 percent, and 31 percent lower than those of the native-born (Table B-2). For these cohorts, as stated earlier, median indexed taxable earnings of immigrants are 20 percent, 33 percent, and 44 percent lower than those of the native-born.

Among immigrants, whites have the highest median indexed taxable earnings and correspondingly receive the largest median annualized payouts. Payouts of the other race-ethnic subgroups are 65 percent to 77 percent of those of whites (Table B-3). When comparing white immigrants with the native-born, we find that median annualized payouts of immigrants are less than those of the native-born population (all race-ethnic subgroups combined) by 3 percent to 12 percent. Over time, from the 1993 cohort to the 2003 cohort, the percentage increases in median annualized payouts are larger for whites and Asians than for Hispanics.

The importance of the age of entry into the United States is highlighted in Table B-4. Among immigrants, median annualized payouts fall markedly as age of entry increases. For those who immigrate before age 23, annualized payouts are similar to those of the native-born.

Taxable Earnings Replacement Rate

Median taxable earnings replacement rates of immigrants slightly exceed those of the native-born, and the relative difference has increased a bit over time. For the 1993, 1998, and 2003 cohorts, median replacement rates for immigrants are 3 percent, 6 percent, and 12 percent higher (Table B-2). We have seen that median indexed taxable earnings of immigrants is less than that of the native-born, and that this relative difference has increased over time. These differences in indexed taxable earnings operate through the progressive benefit formula to produce higher taxable earnings replacement rates for immigrants.

We stated earlier that relatively more immigrants than native-born have U.S. employment histories that are insufficient to qualify for benefits. Generally, a person needs at least 10 years of U.S. earnings to establish eligibility for retirement benefits for one's self or for one's spouse. The ratios of beneficiaries to program participants (those with some shared indexed taxable earnings) are 96 percent for the native-born and 89 percent to 91 percent for immigrants. For the 1993, 1998, and 2003 cohorts, median replacement rates for immigrant beneficiaries only are 5 percent, 14 percent, and 20 percent higher than those for the native-born.

Gustman and Steinmeier (1998) argue that beneficiaries who immigrate as adults are treated too favorably by the Social Security program. They argue that these immigrants have rates of return on contributions and replacement rates for taxable earnings that are too high. They note that two features of the Social Security benefit calculation are responsible for the relatively favorable treatment of such immigrant beneficiaries: the benefit formula is progressive with respect to indexed taxable earnings and the determination of computation periods causes the indexed taxable earnings of such immigrant beneficiaries to be artificially low. Naturally, these arguments to do not apply to immigrants who pay some Social Security taxes, but do not qualify for benefits.

Table B-3 shows that when we focus on immigrants alone, the 1998 and 2003 cohorts' median taxable earnings replacement rates are lowest for whites (31 percent) and highest for Hispanics (41 percent to 43 percent). The primary reason for this pattern is the progressivity of the Social Security benefit formula. For these two cohorts, median indexed taxable earnings of Hispanics is 48 percent to 50 percent of that of whites.

Among immigrants, median taxable earnings replacement rates generally increase as age at entry increases from younger than age 23 to 43–52 (Table B-4). A primary reason for this pattern is the progressivity of the benefit formula. Median indexed taxable earnings decreases as age at entry increases over this age-at-entry range.

Less-Censored Earnings Replacement Rate

Median less-censored earnings replacement rates of immigrants fall short of those of the native born by 8 percent to 14 percent (Table B-2).41 How do our two earnings replacement rates compare between the native-born and immigrants? We find that for the native-born, their less-censored earnings replacement rates are 3 percent to 8 percent lower than taxable earnings replacement rates because their indexed less-censored earnings are larger than their indexed taxable earnings. The hypothetical less-censored taxable maximums exceed the legislated taxable maximums. Thus, some earnings that are above the legislated maximums are below the hypothetical maximums. For immigrants, their less-censored earnings replacement rates are considerably lower (20 percent to 23 percent) than their taxable earnings replacement rates primarily because their indexed less-censored earnings are far greater than their indexed taxable earnings, more so than for the native-born. This is because their computation periods for indexed less-censored earnings are often shorter than those for indexed taxable earnings.

Table B-3 shows that among immigrants, median less-censored earnings replacement rates are lowest for Asians (23 percent to 25 percent) and highest for Hispanics (30 percent to 32 percent). This pattern differs from that for taxable earnings replacement rates where whites had the lowest replacement rates. This is because our subgroups vary in how their indexed taxable earnings compare with their indexed less-censored earnings. Note that in the calculation of the earnings replacement rates, the denominators of the taxable earnings replacement rate and the less-censored earnings replacement rate are indexed taxable earnings and indexed less-censored earnings, respectively; but both replacement rates have the same numerator, namely, annualized payout. The differences in the two earnings replacement rates arise because of differences in the denominator. Asians have a relatively low ratio of indexed taxable earnings to indexed less-censored earnings, in part because they have the highest average age at entry; late entry tends to reduce median indexed taxable earnings, relative to median indexed less-censored earnings.

Among immigrants, median less-censored earnings replacement rates decrease as age at entry increases from younger than age 23 to 43–52 (Table B-4). Taxable earnings replacement rates generally increase over this age-at-entry range. This difference results because as age at entry increases over this range, median indexed taxable earnings fall markedly relative to median indexed less-censored earnings.

Section Summary

Primarily due to their lower indexed taxable earnings, immigrants of every race-ethnic subgroup, on average, receive lower Social Security wealth and annualized payouts than the native-born (all race-ethnic subgroups combined). Despite having some earnings, a larger share of immigrants, compared with the native-born, have earnings histories that are insufficient to qualify for any benefits. Age at entry plays a very important role in determining benefit levels, with our results showing a strong negative correspondence between immigrants' benefit levels and age at entry into the country. The importance of age at entry is strengthened by our finding that immigrants who enter before age 23 have benefits that are similar to those of the native-born.

However, immigrants as a whole have somewhat higher taxable earnings replacement rates than the native-born. Note the relatively high taxable earnings replacement rates for Hispanic and Asian immigrants, especially for Hispanic immigrants. On the other hand, for certain immigrants, particularly Asians, the taxable earnings replacement rate measure is not a very good measure of Social Security benefits as a percentage of an immigrant's average standard of living over their work career. Because only earnings after immigrating to the United States are used in the computation of indexed less-censored earnings, for this purpose the less-censored earnings replacement-rate measure is better for immigrants in particular. We find that less-censored earnings replacement rates for immigrants as a whole are somewhat lower than those of the native-born.

Another finding relates to how Hispanic immigrants have fared over time. Their Social Security wealth and annualized payout levels increase only modestly for recent cohorts. This relatively stagnant pattern for benefits is not observed among white and Asian immigrants. Because immigrants comprised almost half of all Hispanic near-retirees, immigration accounts in part for the slow growth in benefits of Hispanic near-retirees.

VI. Findings by Social Security Benefit Type

Proposals to modify Social Security often would, by design, affect some types of beneficiaries and not others, for example, reduce spouse benefits, increase survivor benefits, and so forth. In considering reform proposals it is useful to know how various types of beneficiaries fare under current law in terms of annualized benefits, replacement rates, and so on. In this section, we present results by benefit-type subgroups. Benefit-type differences can be associated with gender and race-ethnicity. For example, we have seen that women receive higher amounts of Social Security wealth because of greater longevity. But also important is their beneficiary designation. Many women receive high Social Security wealth amounts as a result of being long-lived survivor beneficiaries. Below, we present results for beneficiary subgroups disaggregated by sex and in some instances by race-ethnicity.42 This section's tables present results for beneficiaries.

Benefit measures in this paper include worker, spouse, divorced spouse, surviving spouse, and surviving divorced-spouse benefits paid from the OASI and DI trust funds. We classify these benefits into four broad benefit types: retired worker only, disabled worker only, spouse (spouse and divorced spouse), and survivor (surviving spouse and surviving divorced spouse). For years after 1999, benefit types are projected by the MINT model.

A person's benefit type is the type of their own benefit; the person's spouse may receive a different type of benefit.43 A dually entitled beneficiary is one who is entitled to a worker benefit and to a larger spouse or survivor benefit. Here we treat the dually entitled as spouse or survivor beneficiaries.44 For a person who is a disabled-worker beneficiary (worker only or dually entitled) in the year just before the year they reach the full retirement age, we treat any worker-only benefit that the person receives in a later year as a disabled-worker benefit.

For each year of benefit receipt after the year the person reaches age 61, we determine their yearly benefit type. About one-fourth of beneficiaries change benefit type during their retirement years, and these changes are concentrated among women. Only 4 percent to 5 percent of men change benefit type compared with 43 percent to 46 percent for women (Table C-1). Thus, 90 percent to 93 percent of those who change benefit type are women.

Table C-1. Number of near-retiree beneficiaries, by cohort, sex, and number of benefit types during retirement (in thousands)
Sex Total Number of benefit types
One Two or more
1993
All 9,582 7,161 2,421
Women 4,983 2,747 2,236
Men 4,599 4,414 185
1998
All 10,584 7,792 2,792
Women 5,576 2,987 2,589
Men 5,008 4,805 203
2003
All 13,293 10,007 3,286
Women 6,874 3,905 2,969
Men 6,419 6,102 317
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).

Of all the persons with multiple benefit types, 92 percent to 93 percent have only two types. Among women with two benefit types, 36 percent to 49 percent move between the spouse and survivor types; 38 percent to 50 percent move between the retired-worker and survivor types (Table C-2). Among men with two benefit types, 54 percent to 59 percent move between the retired-worker and survivor types.

Table C-2. Number of near-retiree beneficiaries with two benefit types, by cohort, sex, and benefit-type combinations during retirement (in thousands)
Sex Total Benefit-type combinations
Retired worker
and spouse
Retired worker
and survivor
Spouse and
survivor
Spouse/survivor
and disabled a
1993
All 2,249 204 884 1,028 133
Women 2,070 164 779 1,006 121
Men 179 40 105 22 12
1998
All 2,556 192 1,141 1,061 162
Women 2,371 151 1,038 1,047 135
Men 185 41 103 14 27
2003
All 3,022 281 1,522 1,014 205
Women 2,721 215 1,361 987 158
Men 301 66 161 27 47
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).
a. Includes (1) spouse and disabled combinations and (2) survivor and disabled combinations.

Because many beneficiaries change benefit type during their retirement years, we decided it would be useful to determine a longest-held benefit type for each beneficiary and then use this variable to classify beneficiaries. A person's longest-held benefit type is their most common yearly benefit type for the period that starts with the year he or she reaches age 62. This is the benefit type variable that is used in the remainder of this section.

In terms of longest-held benefit type, 55 percent to 61 percent of beneficiaries are retired workers, 12 percent to 15 percent are spouses, 17 percent to 20 percent are survivors, and 9 percent to 10 percent are disabled workers (Table C-3). Men are predominantly retired-worker beneficiaries. Women are mostly spread among three benefit types—retired-worker (33 percent to 44 percent), spouse (20 percent to 27 percent), and survivor (31 percent to 36 percent). Retired workers and disabled workers are predominantly male, and spouse and survivors are overwhelmingly female. Men account for 63 percent to 69 percent of retired workers and 70 percent to 79 percent of disabled workers. Women, on the other hand, account for 86 percent to 92 percent of spouses and 93 percent to 95 percent of survivors.

Table C-3. Number of near-retiree beneficiaries, by cohort, sex, and longest-held benefit type (in thousands)
Sex Total Retired
worker
Spouse Survivor Disabled
1993
All 9,582 5,305 1,453 1,875 949
Women 4,983 1,660 1,336 1,785 202
Men 4,599 3,645 117 90 748
1998
All 10,584 6,185 1,422 1,939 1,038
Women 5,576 2,174 1,279 1,848 275
Men 5,008 4,011 144 92 762
2003
All 13,293 8,160 1,636 2,271 1,226
Women 6,874 2,997 1,404 2,113 361
Men 6,419 5,163 233 159 864
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).

We briefly mention the shares of benefit types by race-ethnicity (numbers not shown in our tables). Asians have the highest share of beneficiaries who are retired workers (61 percent to 68 percent); blacks have the highest shares of beneficiaries who are survivors (20 percent to 24 percent) and disabled workers (16 percent to 18 percent) and the lowest shares who are retired workers (49 percent to 54 percent) and spouses (7 percent to 10 percent). Among whites, the shares of beneficiary types are as follows: survivors (17 percent to 19 percent), disabled workers (8 percent to 9 percent), retired workers (56 percent to 60 percent), and spouses (13 percent to 16 percent). These simple descriptive statistics tell us a bit about how our cohorts of different race-ethnic subgroups use various aspects of the Social Security program.45

Social Security Wealth

Using longest-held benefit type as the classifier, we find that across all beneficiaries, median shared Social Security wealth is highest for survivors and lowest for disabled workers (Table C-4).46

Table C-4. Social Security benefit measures and related measures for the 2003 cohort of near-retiree beneficiaries, by sex and longest-held benefit type
Measure and sex Total Retired
worker
Spouse Survivor Disabled
Social Security wealth (median, 2002 dollars)
All 172,043 168,196 163,466 239,306 94,201
Women 202,355 191,476 179,196 245,894 125,787
Men 143,082 155,297 65,197 139,122 87,138
Annualized payout (median, 2002 dollars)
All 8,437 8,371 8,176 8,743 8,747
Women 8,471 8,171 8,545 8,777 8,743
Men 8,418 8,475 3,708 8,249 8,753
Taxable earnings replacement rate (median, percent)
All a 28.5 35.3 40.5 a
Women a 29.4 34.3 40.0 a
Men a 28.0 55.3 51.5 a
Less-censored earnings replacement rate (median, percent)
All a 27.3 33.1 39.3 a
Women a 28.4 32.0 38.8 a
Men a 26.7 44.6 50.4 a
Taxable earnings (median, 2002 dollars)
All a 30,116 22,392 21,194 a
Women a 27,886 25,067 21,636 a
Men a 31,227 5,921 13,701 a
Less-censored earnings (median, 2002 dollars)
All a 31,145 24,304 22,273 a
Women a 28,638 26,457 22,780 a
Men a 32,379 8,052 14,275 a
Years of benefit receipt (mean)
All 22.0 21.4 22.0 28.7 13.5
Women 25.5 25.2 22.8 29.4 16.4
Men 18.2 19.3 16.6 19.3 12.3
Years of potential benefit receipt (mean)
All 22.9 22.4 22.9 29.5 13.7
Women 26.3 26.0 23.4 30.2 16.6
Men 19.2 20.3 19.6 20.3 12.5
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).
a. Replacement rates for the disabled are not calculated because many disabled beneficiaries begin receiving benefits before age 62. Therefore, there are no totals available for those measures.

Among women, survivor beneficiaries have the highest Social Security wealth and disabled-worker beneficiaries have the lowest. This SSW pattern results because survivor beneficiaries have the highest mean number of benefits receipt years and disabled-worker beneficiaries have the lowest.

Among men, spouse beneficiaries have by far the lowest Social Security wealth and disabled workers have the second lowest. Spouse beneficiaries have the second lowest mean number of benefit receipt years and disabled-worker beneficiaries have the lowest. In addition, spouse beneficiaries have very low shared indexed taxable earnings.

When we compare women and men, for each longest-held benefit type subgroup, shared Social Security wealth is greater for women than for men. The ratio of female to male median Social Security wealth amounts is especially high for spouse and survivor beneficiaries (2.21 to 2.75 and 1.61 to 1.83). Women have more years of benefit receipt than men because they live longer. For example, for survivor beneficiaries, the mean number of benefit receipt years for women is especially high when compared with that of men; the ratio of female to male mean number of years of benefit receipt ranges from 1.42 to 1.52.

In tabulations by race-ethnicity (not shown here), Social Security wealth is highest for survivors for all of the race-ethnic subgroups and lowest for disabled workers for three of these subgroups. Average numbers of years of benefit receipt are highest for survivors and lowest for disabled workers for each of these subgroups. For each benefit type, Social Security wealth of whites is the highest, with average number of years of benefit receipt of whites higher than that of most other race-ethnic subgroups.

Annualized Social Security Wealth Payout

Using longest-held benefit type as the classifier, we find that, across all beneficiaries, shared annualized payout amounts are highest for survivors and disabled workers and lowest for spouses (Table C-4).

For women, median shared annualized payout amounts differ somewhat by benefit type (Table C-4). These amounts are slightly higher for survivors than for retired workers and spouses. Individual spouse benefits are typically 50 percent of the individual benefit of the higher-earning spouse. Individual survivor benefits are typically 100 percent of the individual benefit of the deceased spouse. This percentage difference (50 percent versus 100 percent) tends to increase shared annualized payouts of survivor beneficiaries relative to those of retired-worker and spouse beneficiaries.

For men, shared annualized payouts of spouse beneficiaries are drastically lower than those of the other seven male or female benefit-type subgroups. Spouse beneficiaries have very low shared indexed taxable earnings.

When comparing men with women, we find the following:

The ratios of female to male median amounts are considerably higher for Social Security wealth than for annualized payout amounts because women have more years of benefit receipt and they live longer.

The pattern of higher annualized payout amounts for survivors and disabled workers than for spouses and retired workers, holds across race-ethnic subgroups (results not shown here). For each benefit type, indexed taxable earnings are highest for whites resulting in their higher annualized payouts compared with the other race-ethnic subgroups.

Earnings Replacement Rates

Here we examine earnings replacement rates for retired workers, spouses, and survivors.47 Again we use longest-held benefit type as the classifier. Median taxable earnings replacement rates are highest for survivors and lowest for retired workers.

Among women, taxable earnings replacement rates are highest for survivors and lowest for retired workers (Table C-4). Note that the taxable earnings replacement rate has annualized payout as its numerator and indexed taxable earnings as its denominator. Survivors have relatively high annualized payout amounts and relatively low indexed taxable earnings. Conversely, retired workers have relatively low annualized payout amounts and relatively high indexed taxable earnings.

Among men, taxable earnings replacement rates are lowest for retired workers. Retired-worker beneficiaries have quite high indexed taxable earnings.

When we compare men with women, taxable earnings replacement rates of retired workers are slightly higher for women than they are for men. On the other hand, replacement rates for survivors are a lot higher for men primarily because male survivor beneficiaries have relatively low indexed taxable earnings. Taxable earnings replacement rates for spouse beneficiaries are markedly higher for men, again because the small subgroup of male spouse beneficiaries have very low indexed taxable earnings.

We find that for each of the race-ethnic subgroups (not shown in our tables), taxable earnings replacement rates are lowest for retired workers. This is primarily because for each race-ethnic subgroup, indexed taxable earnings are highest for retired workers. For each benefit-type subgroup, taxable earnings replacement rates are lowest or second lowest for whites, and indexed taxable earnings are highest for whites. Thus, the progressivity of the benefit formula is an important reason why taxable earnings replacement rates of whites are usually lower than those of the other subgroups.

The patterns for less-censored earnings replacement rates are rather similar to those for taxable earnings replacement rates.

Section Summary

Most beneficiaries do not change benefit type during their retirement years. About one-fourth of beneficiaries, overwhelmingly female, do change types, usually moving from spouse to widow or from retired worker to widow.

Analyzing benefits by longest-held benefit type, we find that survivor and disabled-worker beneficiaries receive the highest levels of annualized payout amounts, and spouse beneficiaries generally the lowest. Survivors also receive the highest Social Security wealth amounts, but now disabled workers receive the lowest, driven by the long and short lives of the two subgroups. Earnings replacement rates are highest for survivors and lowest for retired workers, driven in part by lower and higher indexed taxable earnings for the two subgroups. When we compare men with women by longest-held benefit type, we find that female spouse and survivor beneficiaries receive higher Social Security wealth and annualized payouts than do their male counterparts, whereas female worker beneficiaries receive annualized payouts modestly lower than those of male worker beneficiaries.

The preceding patterns generally hold across race-ethnic subgroups as well. For each benefit-type subgroup, nonwhites receive higher replacement rates than whites because of the progressivity of the benefit formula. A larger share of blacks are survivors or disabled workers than of any other race-ethnic subgroup.

In addition, two subgroups of beneficiaries are very different in terms of values of benefit measures. Male spouse beneficiaries (only 1 percent to 2 percent of all beneficiaries) receive amounts of Social Security wealth that are lower than those of any other male or female benefit-type subgroup. On the other hand, female survivor beneficiaries (16 percent to 19 percent of all beneficiaries) have the most years of benefit receipt and get considerably higher amounts of Social Security wealth than any other subgroup.

VII. Findings by Disability Status

How do the disabled fare under Social Security compared with the nondisabled? How are these disabled/nondisabled differences associated with sex and race-ethnicity?48 In this section, we present results for disability-status subgroups, and discuss some reasons for these differences.

We classify beneficiaries, that is, program participants with shared benefits, into two disability benefit-status subgroups (disabled and nondisabled) using longest-held benefit type codes. Because this paper focuses on shared benefits, we use both the person's type code and the spouse's type code in determining a person's disability status. The disabled are persons whose longest-held benefit type is disabled-worker only and/or whose spouse's longest-held benefit type is disabled-worker only.49,50 This shared-record disabled subgroup is considerably broader than the person-record disabled-worker only subgroup examined in the previous section; those whose person-record longest-held benefit type is disabled worker only account for 60 percent to 61 percent of these shared-record disabled (Table D-1). The remaining 39 percent to 40 percent of our disabled are persons who do not receive disabled-worker only benefits themselves but have a spouse who receives such benefits. All other beneficiaries are classified as nondisabled.51 These shared-record disabled account for 15 percent to 16 percent of all beneficiaries, 17 percent to 20 percent of male beneficiaries, and 13 percent to 14 percent of female beneficiaries. The 4 percent to 5 percent of program participants with no shared benefits are not dealt with in this section.

Table D-1. Percentage of near-retiree beneficiaries with selected characteristics, by disability status and cohort
Characteristic Disabled Nondisabled
1993 1998 2003 1993 1998 2003
Reason for disability status
Both person and spouse disabled 7 5 5 0 0 0
Only person disabled 54 55 55 0 0 0
Only spouse disabled 40 40 40 0 0 0
Neither disabled 0 0 0 100 100 100
Male 58 56 55 46 46 47
Foreign-born 9 9 10 10 10 11
Married at age 62 82 81 79 74 72 70
Race-ethnicity
White 71 76 73 83 83 81
Black 12 15 15 8 7 8
Asian 2 1 3 3 2 4
Hispanic 9 7 8 6 6 7
Total number of beneficiaries (thousands) 1,576 1,734 2,050 8,012 8,851 11,244
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).
NOTE: Disability status is based on the individual's and the spouse's disability benefit types. By contrast, in Tables C-1 through C-4, disability status is based only on the individual's benefit type.

In considering our results in this section it is important to keep in mind the following facts about the subgroup we call the disabled. First, our disabled include not only disabled workers, but also persons with spouses who are disabled workers. Second, our disabled all live to at least age 61. It is important to note that many disability beneficiaries die before reaching age 61. Third, on average, our disabled first receive disability benefits when in their mid- to late fifties. For all disability beneficiaries, the average age of first receipt of benefits is well below the mid-fifties.52 Fourth, in determining a person's longest-held benefit type, we do not convert disabled-worker beneficiaries to retired-worker beneficiaries when they reach the full retirement age. Thus, our disabled group is defined somewhat differently than the typical disabled population.

Looking at the demographics of our disabled, we find that the disabled are more male than the nondisabled. Some 55 percent to 58 percent of the disabled are men compared with 46 percent to 47 percent of the nondisabled (Table D-1). Most disabled men (77 percent to 82 percent) are persons whose person-record longest-held benefit type is disabled-worker only. In contrast, most disabled women (61 percent to 69 percent) are persons who do not have a person-record longest-held benefit type of disabled-worker only, but have a spouse with a longest-held benefit type of disabled-worker only.

The shares of blacks, Hispanics, and Asians in our disabled subgroup are larger than their shares in our nondisabled population. Some 24 percent to 27 percent of the disabled are minorities compared with 17 percent to 19 percent for the nondisabled. The disabled comprise a larger share of black beneficiaries (23 percent to 28 percent) than of any other race-ethnic subgroup.

Immigrants account for 9 percent to 10 percent of the disabled and 10 percent to 11 percent of the nondisabled. The percentages married at age 62 are higher for the disabled (79 percent to 82 percent) than they are for the nondisabled (70 percent to 74 percent).

Below we discuss empirical estimates of Social Security wealth and of annualized payouts for the disabled and nondisabled. Because many of the disabled near-retirees begin to receive disability benefits a number of years before they reach age 62, we decided not to present any replacement rate estimates in this section.

Social Security Wealth

Our measure of Social Security wealth does not include benefits received before the year the person reaches age 62. Yet, for the near-retiree disabled, the great majority begin to receive disability benefits before reaching age 62.

Our disabled have fewer years of benefit receipt because they die sooner. Therefore, it is not surprising that median Social Security wealth of the disabled is considerably less than that of the nondisabled. This is true for both men and women; this is also true for whites, blacks, and Hispanics. Median Social Security wealth of the disabled is 27 percent to 28 percent lower and mean number of years of benefit receipt of the disabled are 23 percent to 24 percent lower than those of the nondisabled (Table D-2).

Table D-2. Social Security benefit measures for near-retiree beneficiaries, by cohort, disability status, and sex
Measure and cohort Disabled Nondisabled
All Women Men All Women Men
Social Security wealth (median, 2002 dollars)
1993 97,208 140,186 69,340 133,274 152,458 111,954
1998 115,950 142,669 92,374 161,969 187,183 133,564
2003 130,706 174,196 101,272 178,702 208,082 151,851
Annualized payout (median, 2002 dollars)
1993 6,896 6,851 7,024 6,349 6,481 6,209
1998 7,955 7,710 8,194 7,554 7,611 7,492
2003 8,689 8,601 8,771 8,392 8,437 8,347
Years of benefit receipt (mean)
1993 16.7 22.2 12.8 22.1 24.8 18.8
1998 16.8 20.4 14.1 22.6 25.8 18.8
2003 17.4 21.8 13.8 22.8 26.1 19.2
Years of potential benefit receipt (mean)
1993 16.8 22.3 12.9 23.0 25.7 19.9
1998 17.0 20.5 14.3 23.6 26.6 20.0
2003 17.7 22.0 14.2 23.8 26.9 20.3
SOURCE: Authors' calculations using data from Modeling Income in the Near Term (MINT3).
NOTE: Disability status is based on the individual's and the spouse's disability benefit types. By contrast, in Tables C-1 through C-4, disability status is based only on the individual's benefit type.

When men and women are looked at separately, we find that median Social Security wealth of disabled men is 31 percent to 38 percent lower and mean number of benefit-receipt years is 24 percent to 32 percent lower than for nondisabled males; the corresponding figures for women are 8 percent to 24 percent and 11 percent to 21 percent.

As with the nondisabled, median Social Security wealth is considerably larger for disabled women than for disabled men. The two main causes of this difference are (1) that women have much higher average number of years of benefit receipt and (2) our use of a shared concept of wealth rather than an individual concept.

Again, as with the nondisabled, because whites have higher average number of years of benefit receipt, disabled median Social Security wealth is larger for whites than for the group of minorities (results not shown here).

Annualized Social Security Wealth Payout

Annualized payouts of the disabled exceed those of the nondisabled by 4 percent to 9 percent. For men and women, payout amounts for the disabled are higher by 5 percent to 13 percent and 1 percent to 6 percent. The corresponding amounts for disabled whites, blacks, and Hispanics are higher by 4 percent to 11 percent, 4 percent to 12 percent, and 15 percent to 22 percent.

A factor that markedly increases annualized payouts of the disabled relative to those of the nondisabled is the following: Non-Disability Insurance benefits are reduced for early benefit receipt, that is, early retirement. For a full retirement age of 66, these reductions can be as large as 25 percent for retired-worker benefits, 30 percent for spouse benefits, and 19 percent for surviving spouse benefits. There are no comparable reductions for DI benefits.

Two factors that decrease annualized payouts of the disabled relative to that of the nondisabled are the following:

  1. The indexing of retired-worker benefits differs from that for disabled-worker benefits. Retired-worker benefits are based on earnings wage-indexed to wage levels as of the year the beneficiary reaches age 60. Cost-of-living adjustments (that is, price-indexing) to these retirement benefits begin at the end of the year the person reaches age 62. By contrast, disabled-worker benefits are based on earnings wage-indexed to wage levels as of the year that is 2 years before the year of first receipt of disability benefits. The cost-of-living adjustments to disability benefits start at the end of the year of first disability benefit receipt. For near-retiree disabled-worker only beneficiaries, the median age of first receipt of disability benefits is 57 or 58.

    Because the average wage measure usually increases at a faster percentage rate than the price index, these differences in indexing usually cause the annualized payouts of the disabled to decrease relative to those of the nondisabled. For the 1998 (1993) cohort, this indexing difference decreases (increases) annualized payouts of disabled-worker only beneficiaries by about 10 percent (1 percent) relative to those of the retired-worker only beneficiaries.

  2. An individual's relative annual taxable earnings are annual taxable earnings relative to the Social Security Administration's average annual wages. The average relative taxable earnings of disabled-worker beneficiaries over their computation periods are less than those of retired-worker beneficiaries over their usually longer computation periods.53,54 The smaller earnings amount for disabled-worker beneficiaries would reduce annualized payouts of the disabled relative to those of the nondisabled even if both types of benefits were wage-indexed to the same beneficiary age and were price-indexed starting from the same beneficiary age. We find that our estimates of medians of average relative taxable earnings of disabled-worker only beneficiaries are 9 percent to 14 percent less than those of retired-worker only beneficiaries.55

Differences in median annualized payouts by sex, are quite small. The ratios of median annualized payouts for women to that for men are 0.94 to 0.98 for the disabled and 1.01 to 1.04 for the nondisabled.

Among the disabled, race-ethnic differences (not shown in our tables) in median annualized payouts are a bit larger than are differences by sex. For the disabled, the median annualized payouts of blacks and Hispanics are 5 percent to 14 percent and 10 percent to 19 percent less than those of whites; the corresponding figures for nondisabled blacks and Hispanics are 9 percent to 12 percent and 14 percent to 30 percent.

Section Summary

Our definition of the disabled is somewhat different than that used by the Social Security program. It is an expanded definition in one sense because in determining who is disabled, we take into account the disability status of one's spouse. On the other hand, because our focus is on near retirees, all of our disabled live to at least age 61, and we measure their shared benefits from age 62 forward. Thus, our disabled are persons whose shared benefits from age 62 forward include a sizable portion of disability benefits; about 60 percent receive disabled-worker benefits themselves and the remaining 40 percent have spouses who receive such benefits. On average, they do not begin receiving disability benefits until their mid-to-late fifties. In determining a person's longest-held benefit type, we do not convert disabled-worker benefits to retired-worker benefits at the full retirement age.

Our near-retiree disabled subgroup is, as expected, different from the near-retiree nondisabled. Males account for a larger proportion of the disabled. In addition, minority race-ethnic groups, especially blacks, make up a larger share of our specific definition of the disabled, just as they did when disability status was defined on the basis of individual records only.

By one measure, namely Social Security wealth, we find that because our disabled subgroup die sooner, it receives considerably less in median amounts than does the nondisabled subgroup. These differences in Social Security wealth for the disabled exist for both men and women and for all of the race-ethnic groups. Among the disabled, women receive more than men, and whites receive more than blacks and Hispanics. However, it is very important to note that had we taken into account all benefits that our disabled received before the year they reached age 62, the nature of these differences may have been quite different. But because the focus of this paper is near-retirees, including the disabled, we examine Social Security benefits only from the year they reach age 62.

Using another measure of benefits, namely annualized payouts, we find that median amounts for the disabled are slightly higher than those of the nondisabled. This small excess is the result of a number of offsetting factors. These offsetting effects include the following: (1) Old-age benefits are reduced for early retirement (there are no comparable reductions for disability benefits); (2) wage-indexing for disability benefits usually stops before reaching age 60, which serves to reduce disability benefits; and (3) the average relative earnings of disabled-worker beneficiaries over their computation periods appear to be less than those of retired-worker beneficiaries over their longer computation periods. For both benefit types, earnings are measured relative to SSA's average annual wages. The lower amount for the disabled subgroup would reduce annualized payouts of the disabled relative to those of the nondisabled even if both types of benefits were indexed in the same way.

Annualized payouts of the disabled subgroup are larger than those of the nondisabled for both women and men and for whites, blacks, and Hispanics. Among the disabled, there are no appreciable differences by sex in payout amounts, but blacks and Hispanics receive somewhat smaller payouts than do whites.

VIII. Concluding Remarks

The various measures of Social Security benefits as a retirement resource for subgroups of near-retirees that are presented here should be useful in different ways. As noted at the onset, it is important to examine how particular subgroups fare, especially if they are considered economically vulnerable and/or may be subject to program changes. We focused on a few select subgroups, namely race-ethnic subgroups, immigrants and the native-born, Social Security beneficiary types, and the disabled and the nondisabled. Our results provide substantial empirical evidence on how the Social Security program affects these populations of concern. A major strength of the results is that for the most part they are based on actual earnings histories—an advantage shared by very few studies on the subject.

Some of our results for near-retirees may be unsurprising. For example, we report that among race-ethnic subgroups, whites receive the highest amounts of Social Security wealth and annualized payouts because of their higher indexed taxable earnings. Taxable earnings replacement rates, on the other hand, are the lowest for whites and higher for minority race-ethnic subgroups, because of the progressivity of the Social Security benefit formula. Immigrants of all race-ethnic subgroups, on average, receive lower Social Security wealth and annualized payouts than the native-born as a whole, primarily because of their lower indexed taxable earnings. When we look at persons by longest-held benefit type, survivor beneficiaries receive the highest amounts of Social Security wealth, because of their much longer lives. In contrast, because of their markedly shorter lives and because we consider Social Security benefits only if received after age 61, the near-retiree disabled, as defined in our paper, receive considerably less in median amounts of Social Security wealth than do the nondisabled.

We are also able to point to other interesting findings from our study of these subgroups. For example, over time, Hispanics have very slow growth in Social Security wealth compared with that of the other race-ethnic subgroups. A key underlying variable is the growth in earnings. Median indexed taxable earnings increases are considerably smaller for Hispanics than they are for the other three race-ethnic subgroups. For immigrants, the taxable earnings replacement rate is not a very good measure of how effective Social Security is in replacing average career total earnings; this is especially so for Asians whose indexed taxable earnings are particularly low relative to their indexed less-censored earnings (our proxy for indexed total earnings) in considerable part because of their having the highest average age of entry into the United States. Age of entry into the country is an important variable. Immigrants who enter before age 23 have benefits similar to those of the native-born.

When we consider benefit types, we find that the small subgroup of male spouse beneficiaries receive substantially lower Social Security wealth than any other male or female benefit-type subgroup. Female survivor beneficiaries, on the other hand, receive significantly higher wealth amounts than any of the seven other male or female benefit-type subgroups. In addition, compared with the other race-ethnic subgroups, a larger share of black beneficiaries receives disability and/or survivor benefits.

Under Social Security law, a person's benefits do not depend on the person's race, ethnicity, nativity, or sex. That notwithstanding, this paper has highlighted that substantial differences in earnings levels and/or mortality levels by these characteristics produce sizable differences in Social Security benefit levels among these subgroups of near-retirees.

Notes

1. Social Security Administration (2006b).

2. Social Security Administration (2006b).

3. Iams and Sandell (1998).

4. Bridges and Choudhury (2005, 2007).

5. We use MINT3 data files created in April 2003.

6. The administrative records contain amounts of annual taxable earnings beginning with 1951.

7. Board of Trustees (2002). Two key economic assumptions of Trustees Reports are those with regard to inflation and the growth of average earnings. The 2002 report uses actual historical data on average wages through calendar year 2000 and on consumer price levels through early 2002.

8. To some extent the incomes of the members of a couple are a product of joint decision making.

9. Given the content of the MINT data file, the sharing of benefit income within a larger unit, such as the family, could not be considered.

10. Through the price index of January 1, 2002, the price index for January 1 of a given year is the average of the published price index for January of that year and the published price index for December of the previous year. For years after 2002, the price index value for January 1 of a given year is the average of the projected price index for that year and the projected price index for the previous year.

11. From the perspective of Social Security beneficiaries, the trust fund interest rate can be viewed as a proxy for a U.S. government bond rate series, since the trust fund interest rate is based on marketable U.S. Treasury obligations. From the perspective of the Social Security program the trust fund interest rate is the rate at which the trust fund is able to transform funds over time.

One can argue for using an interest rate lower or higher than the trust fund rate. Using a lower or higher interest rate would of course change the levels of the estimated SSW for subgroups, but within a cohort would be expected to usually leave unchanged the rankings of the subgroups in terms of size of SSW, for example, SSW of whites greater than that of blacks.

12.The number of potential benefit years equals 0 for persons who die in the year they reach age 62, equals 1 for persons who die in the year they reach 63, and so on.

13. For the year of a person's death, the MINT benefit calculator does not credit the person with any individual or shared benefits. For example, in the case of a beneficiary who dies in July 2000, the MINT calculator does not credit the person with any benefits for calendar year 2000. For the year the person begins to receive benefits, the benefit calculator credits the person with 12 months of benefits unless that is the year that the person dies.

14. A somewhat similar measure is used in Smith, Toder, and Iams (2003/2004). See their "Overall Approach" section.

15. The cohort or cohort subgroup with greater average longevity than another such group can be said to have additional potential benefit years, most of which will also be years in which the beneficiaries receive real annual benefits that are at least as large as those received in their earlier years. These additional benefits result in additional Social Security wealth. To compute annualized payout of this longer-lived group, its larger SSW is spread over a larger number of potential benefit years. Thus, greater longevity usually causes a smaller percentage increase in annualized payout than in Social Security wealth.

16. The proportion of all workers (of any age) in covered employment with covered earnings at or above the legislated taxable maximums was 6 percent during the 1983–1989 period and 5 percent to 6 percent during the 1990s. The percentages during the 1951–1978 and the 1979–1982 periods were 15 to 36 and 7 to 10.

17. For each year of the 1951–1977 period, the MINT model uses information from SSA administrative records on the quarter in which a person's earnings reached the legislated taxable maximum to assign a person to a covered earnings interval. Means for each interval were derived from the earnings data collected by the Census Bureau in its Current Population Surveys (CPS). Each person is assigned the mean earnings for their interval.

For the 1978–1989 period, the administrative records do not contain information on the quarter in which an individual's earnings reached the legislated taxable maximum. For this later period, covered earnings above the legislated taxable maximum were set at the CPS average of earnings above the legislated taxable maximum for each year.

See Butrica and others (2001) for additional information on the MINT projection method for less-censored earnings. MINT modelers coined the phrase "less-censored earnings."

18. Because the numerator of the replacement rate, annualized payout, is expressed in January 1, 2002, dollars, we need to express the denominator of the replacement rate, indexed taxable earnings (TX-EARN), in January 1, 2002, dollars.

P2002 is the CPI as of January 1, 2002, and PT is the CPI as of January 1 of year T (the year the person reaches age 62). AET is average wage-indexed shared taxable earnings indexed to the average wage level prevailing as of January 1 of year T.

TX-EARN = (P2002 / PT) AET.

19. SSW is evaluated as of January 1 of the year the person reaches age 62. Annualized payout, the numerator of our replacement rates, is derived from SSW. Thus, we want to wage-index less-censored earnings (the denominator of the less-censored earnings replacement rate) to the wage level as of the beginning of the year the person reaches age 62. Having the timing of its numerator and denominator consistent makes the less-censored earnings replacement rate a better measure of the adequacy of Social Security benefits. We chose to wage-index taxable earnings to the same date as that used for wage-indexing less-censored earnings.

20. For each year a person is married, the person's shared earnings equal one-half of the earnings of the couple. For each year a person is not married, the person's shared earnings equal the person's own earnings.

21. For purposes of determining retired-worker benefits, the worker's AIME is determined as follows. Annual taxable earnings through age 60 are indexed, using the average wage series, to wage levels as of the year the worker reaches age 60; annual earnings after age 60 are not wage-indexed. The sum of the 35 highest earnings is divided by 420 (35 × 12) to get AIME. For disabled workers the calculation of AIME usually employs a shorter computation period (less than 35 years). Given that we use a shared benefit measure, annualized payout, we needed a shared earnings measure. For various conceptual and data reasons, we could not compute a shared AIME measure.

22. Table A-2 does not present data on benefit measures (SSW, and so on) by sex.

23. Persons are projected to enter the MINT sample by means of immigration in the years after the end of the SIPP interviews. A hot-deck imputation procedure is used for this purpose.

24. Some 66 percent to 68 percent of Asians and 30 percent to 34 percent of Hispanics enter the United States after the year they reach age 22; the comparable figures for whites and blacks are 2 percent to 3 percent and 4 percent to 5 percent.

25. There is considerable evidence that, other things being the same, mortality rates for Hispanics are lower than those for non-Hispanic whites; see Franzini, Ribble, and Keddie (2001) and Liao and others (1998). Thus the MINT-based estimates of Hispanic longevity and SSW are likely to be too low. There is some evidence that mortality rates for Asians, other things being equal, may be lower than those for non-Hispanic whites; see Rogers and others (1996).

26. Haveman and others (2006), Wolff (2002), and Liu and Rettenmaier (2003) are three recent studies that present some estimates of Social Security wealth by race-ethnic subgroups. Their data sets differ from each other and from our data set. The focus of each of these studies is rather different from the focus of our study. Each study uses only two race-ethnic subgroups.

Haveman and others (2006) use samples from the New Beneficiary Study and from the Health and Retirement Study to examine the overall retirement income adequacy of persons who retired in the early 1980s and in the mid-1990s. One of their findings is that the average Social Security wealth of whites exceeds that of nonwhites.

Wolff (2002) uses samples from three Surveys of Consumer Finances to estimate the overall retirement income adequacy of persons aged 59–64 in 1983, 1989, and 1998. One of his findings is that the average Social Security wealth of non-Hispanic whites exceeds that of the combined group of blacks and Hispanics.

Liu and Rettenmaier (2003) use a set of hypothetical workers in their study of the money's worth of Social Security for workers born from 1935 through 1980. One of their findings is that the average Social Security wealth of whites exceeds that of blacks.

We see that the findings of these three studies are generally consistent with our findings.

27. Most married women receive smaller annual benefits (auxiliary or worker) than their husbands. Thus, shared benefit is greater than individual benefit for most married women and less than individual benefit for most married men.

28. Because average indexed monthly earnings are wage-indexed, the rates of growth of average real Social Security wealth and of average real annualized payouts from the 1998 cohort to the 2003 cohort are sensitive to the projected growth of the average annual wage series through 2006. Similarly, because indexed taxable earnings and indexed less-censored earnings are wage-indexed, their rates of growth are sensitive to the projected growth of the average annual wage series through 2008. In this paper we use projections of the average annual wage series from the 2002 Trustees Report, which overstated the growth of the average annual wage over the 2000–2004 period. Using the 2002 Report projections, the real average annual wage facing the 2003 cohort at age 62 exceeded that faced by the 1998 cohort by 9 percent; using the 2006 Report assumptions, the comparable figure is 4 percent. Thus, our tables probably substantially overstate the 1998–2003 growth rates for real Social Security wealth, real annualized payouts, real indexed taxable earnings, and real indexed less-censored earnings.

29. Under OASDI law, for disabled-worker beneficiaries, the year they become disabled and later years are usually disregarded in determining average indexed monthly earnings. In determining indexed taxable earnings (and indexed less-censored earnings), we include such years if they are before the year the worker attains age 62. Many of the near retirees who receive disability benefits start receiving them before reaching age 62. For our near retirees, the median age of first receipt of disability benefits is 57 or 58. Some 15 percent of Social Security program participants receive shared disability benefits. Including such post-disability years in the computation of indexed taxable earnings and indexed less-censored earnings for disability beneficiaries causes modest increases in taxable earnings replacement rates and less-censored earnings replacement rates for the race-ethnic subgroups and for immigrant-status subgroups.

30. The impact of the lack of precision of mortality projections for Hispanics and Asians on estimates of taxable earnings replacement rates and less-censored earnings replacement rates for these subgroups should be relatively small.

31. The overstatement of the 2000–2004 growth of the average annual wage (referred to in note 28) should have only small effects on our estimates of median taxable earnings replacement rates and less-censored earnings replacement rates. This overstatement of wage growth causes offsetting overstatements of the numerator and denominators of our replacement rates.

32. A paper that deals with immigrants and Social Security in a somewhat different way is Cohen and Iams (2007).

33. A hot-deck imputation procedure is used in selecting postinterview immigrants from a donor pool of immigrants from the SIPP sample. The imputation is done so as to approximate estimated control totals of immigrants by time period, sex, age at immigration, and source region. The records of the selected donors are then updated to the year of projected immigration. All imputed immigrants enter the United States as adults. These immigrant projections could contain considerable error. For details, see Berk and Smith (2003).

34. The SIPP panels contain, to an unknown degree, undocumented or illegal immigrants. The SIPP coverage rate for the undocumented is probably quite low relative to those of legal immigrants and of the native-born. The estimated control totals for immigrant imputations do not include the undocumented. For our analysis, MINT's treatment of the undocumented causes very little problem; most of the undocumented enter the United States before age 35, and most of them stay in the country less than 10 years. See the discussion in Duleep and Dowhan (2006).

35. Gustman and Steinmeier (1998) use the Health and Retirement Study sample to examine Social Security's treatment of immigrants and natives born from 1931 through 1941. One of their findings is that the average Social Security wealth of the native-born exceeds that of immigrants.

36. There is some evidence that, other things including race-ethnicity being the same, mortality rates may be lower for immigrants than for the native-born; for example, see Rogers and others (1996).

37. For the small minority of immigrants whose benefits are based on totalization agreements, their benefits are not computed under the usual OASDI rules. In 2004, about 100,000 immigrants, emigrants, and others received some U.S. OASDI benefits under totalization agreements.

38. Blacks account for only 5 percent to 6 percent of immigrants.

39. Gustman and Steinmeier (1998) find that the average Social Security wealth of immigrants generally is lower the later the year of immigration.

40. For beneficiaries only (those with positive SSW), median SSW and median indexed taxable earnings also generally decrease as age at entry increases.

41. For beneficiaries only, median less-censored earnings replacement rates for immigrants are 2 percent to 12 percent lower.

42. Our tables for this section do not present data for benefit measures (Social Security wealth, and so on) by race-ethnicity.

43. Although our benefit measures are for shared benefits, we decided to determine benefit type using only the person's type code. We thought that using both the person's code and the spouse's code to determine the person's benefit category (for example, person is retired worker and spouse is retired worker, person is spouse beneficiary and their spouse is a retired-worker beneficiary, and so on) would produce so many benefit categories that our analysis by benefit category would become very cumbersome.

44. Weaver (1997) presents estimates of average benefit amounts by type of benefit. In defining benefit types he treats dual beneficiaries as auxiliary beneficiaries as we do. However, his estimates are for individual benefits and are thus not comparable to our estimates of shared benefits.

45. Whites account for 71 percent to 76 percent of disabled workers, 79 percent to 81 percent of survivors, and 81 percent to 85 percent of the other benefit types. Immigrants account for 8 percent to 13 percent of the various benefit-type subgroups.

46. Table C-4 gives data only for the 2003 cohort. Patterns found in the data for the various benefit measures for the other two cohorts are similar to those found for the 2003 cohort.

47. Indexed taxable earnings, the denominator of the taxable earnings replacement rate, is an average of indexed earnings through age 61. Because many disabled-worker beneficiaries first receive disability benefits a number of years before reaching age 62, we chose not to present taxable earnings replacement rates for these beneficiaries. For near-retiree disabled-worker only beneficiaries, the median age of first receipt of disability benefits is 57 or 58.

48. Our tables for this section do not present data for benefit measures (Social Security wealth, and so forth) by race-ethnicity.

49. The person's and spouse's longest-held benefit types are for the same time period, namely, the period that starts with the year the person reaches age 62 and ends with the person's death.

50. Survivor beneficiaries whose benefits are based on the earnings of a disabled worker will not be classified as disabled in cases in which the disabled worker died before the survivor beneficiary reached age 62.

51. This definition of nondisabled differs from the definition of the non-Disability Insurance Benefit population in Bridges and Choudhury (2005).

52. For each year of the 1983–2004 period, the average age of first receipt of disabled-worker benefits (worker only plus dually entitled) is 48–50 years. See Table 6.C2 of Social Security Administration (2006a).

53. Under Social Security law the determination of AIME computation periods for disabled-worker benefits differs from that for retired-worker benefits. In our calculations of average relative individual taxable earnings of disabled-worker beneficiaries and retired-worker beneficiaries, we approximate computation periods as follows. For both types of benefits, our computation period starts with 1951 or the year the person reaches age 22, whichever comes later. For retired-worker benefits, the period ends with the year the person reaches age 61. For disabled-worker benefits, the computation period ends with the year before the year of first receipt of disability benefits.

54. This measure (average relative individual taxable earnings) is not used anywhere else in this paper.

55. Disabled workers who survive to age 61 have higher average earnings than those who die before age 61.

References

Berk, Jillian, and Karen Smith. 2003. Task 1. Add immigrants to MINT. Letter report to Leslie Muller of the Social Security Administration. Washington, DC: Urban Institute (March 18).

Board of Trustees of the Federal Old-Age and Survivors Insurance and Disability Insurance Trust Funds. 2002. 2002 annual report. Washington, DC: U.S. Government Printing Office.

Bridges, Benjamin, and Sharmila Choudhury. 2005. Social Security as a source of retirement resources for near-retirees. ORES Working Paper No. 106, Social Security Administration, Office of Policy, Office of Research, Evaluation, and Statistics, Washington, DC (May).

———. 2007. Social Security benefits as a retirement resource for U.S. near-retirees. Review of Income and Wealth 53(3): 538–567.

Butrica, Barbara, Howard M. Iams, James Moore, and Mikki D. Waid. 2001. Methods in Modeling Income in the Near Term (MINT I). ORES Working Paper No. 91, Social Security Administration, Office of Policy, Office of Research, Evaluation, and Statistics, Washington, DC (June).

Cohen, Lee, and Howard Iams. 2007. Income adequacy and Social Security differences between the foreign-born and U.S.-born. International Migration Review 41(3): 555–578.

Duleep, Harriet Orcutt, and Daniel J. Dowhan. 2006. Incorporating immigrant flows into microsimulation models. Unpublished paper, Social Security Administration, Office of Policy, Office of Research, Evaluation, and Statistics.

Franzini, Luisa, John C. Ribble, and Arlene M. Keddie. 2001. Understanding the Hispanic paradox. Ethnicity and Disease 11(Autumn): 496–518.

Gustman, Alan L., and Thomas L. Steinmeier. 1998. Social Security benefits of immigrants and U.S. born. NBER Working Paper No. 6478, National Bureau of Economic Research, Cambridge, MA.

Haveman, Robert, Karen Holden, Barbara Wolfe, and Andrei Romanov. 2006. The sufficiency of retirement savings: Comparing cohorts at the time of retirement. Pension Research Council Working Paper No. 14, University of Pennsylvania, Philadelphia, PA (September).

Iams, Howard M., and Steven H. Sandell. 1998. Cost-neutral policies to increase Social Security benefits for widows: A simulation for 1992. Social Security Bulletin 61(1): 34–43.

Liao, Youlian, Richard S. Cooper, Guichan Cao, Ramon Durazo-Arvizu, Jay S. Kaufman, Amy Luke, and Daniel L. McGee. 1998. Mortality patterns among adult Hispanics: Findings from the NHIS, 1986 to 1990. American Journal of Public Health 88(2): 227–232.

Liu, Liqun, and Andrew J. Rettenmaier. 2003. Social Security outcomes by racial and education groups. Southern Economic Journal 69(4): 842–864.

Panis, Constantijn, and Lee Lillard. 1999. Near-term model development. Draft final report, SSA contract No. 600-96-27335, RAND Corporation, Santa Monica, CA.

Rogers, Richard G., Robert A. Hummer, Charles B. Nam, and Kimberley Peters. 1996. Demographic, socioeconomic, and behavioral factors affecting ethnic mortality by cause. Social Forces 74(4): 1419–1438.

Smith, Karen, Eric Toder, and Howard Iams. 2003/2004. Lifetime distributional effects of Social Security retirement benefits. Social Security Bulletin 65(1): 33–61.

[SSA] Social Security Administration. 2006a. Annual Statistical Supplement to the Social Security Bulletin, 2005. Washington, DC: SSA, Office of Policy, Office of Research, Evaluation, and Statistics.

———. 2006b. Income of the population 55 or older, 2004. Washington, DC: SSA, Office of Policy, Office of Research, Evaluation, and Statistics.

Toder, Eric, Lawrence Thompson, Melissa Favreault, Richard Johnson, Kevin Perese, Caroline Ratcliffe, Cori Uccello, Timothy Waidmann, Jillian Berk, Romina Woldermariam, Gary Burtless, Claudia Sahm, and Douglas Wolf. 2002. Modeling Income in the Near Term: Revised projections of retirement income through 2020 for the 1931–1960 birth cohorts. Report prepared under SSA contract No. 600-96-27332. Washington, DC: Urban Institute and Brookings Institution (June).

Weaver, David A. 1997. The economic well-being of Social Security beneficiaries, with an emphasis on divorced beneficiaries. Social Security Bulletin 60(4): 3–17.

Wolff, Edward N. 2002. Retirement insecurity: The income shortfalls awaiting the soon-to-retire. Washington, DC: Economic Policy Institute.