CONGRESS OF THE UNITED STATES CONGRESSIONAL BUDGET OFFICE Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Social Security replacement rates—benefits expressed as a percentage of preretirement earnings—vary substantially depending on how they are measured. Percent 60 Individual Shared 40 55 55 49 49 44 45 20 36 39 36 36 0 Last 5 Years of Last 20 Years, Including Highest 35 Years of Highest 35 Years of All Earnings From Substantial Earnings No or Very Low Earnings Wage-Adjusted Earnings Price-Adjusted Earnings Age 22 Through Age 61 Late-Career Earnings Lifetime Earnings Measures Measures Median Replacement Rates for Long-Career Workers Born in the 1960s APRIL 2019 Notes Unless otherwise indicated, all years referred to in this report are calendar years. Similarly, unless otherwise noted, all benefits referred to in this report are scheduled benefits, which are those calculated under current law regardless of the amounts in the Social Security trust funds. Numbers in the text, tables, figures, and exhibits may not add up to totals because of rounding. Definitions of various terms appear at the end of this report. www.cbo.gov/publication/55038 Contents Summary 1 Background 2 How Did CBO Develop Its Projections? 2 What Population Did CBO Analyze? 3 What Measures of Social Security Benefits Did CBO Examine in Its Analysis? 3 How Do Social Security Benefits Compare With Poverty Thresholds? 5 How Do Benefits Compare With Preretirement Earnings? 6 How Does Taking Taxes Into Account Affect the Results? 7 How Do Benefit Measures Change Over the Course of Retirement? 7 How Does Uncertainty Affect the Projections? 8 How Does This Report Relate to Other Analyses by CBO? 8 Initial Benefits Compared With the Official Federal Poverty Threshold 10 Exhibits 1 through 3 TABLE OF CONTENTS SOCIAL SECURITY REPLACEMENT RATES AND OTHER BENEFIT MEASURES: AN IN-DEPTH ANALYSIS II Initial Individual and Shared Replacement Rates 14 Exhibits 4 through 11 Numerators and Denominators of Replacement Rates 15 Exhibit 4 Individual Replacement Rates 17 Exhibits 5 through 7 Shared Replacement Rates 21 Exhibits 8 through 11 Multiyear Shared Replacement Rates 26 Exhibits 12 through 14 Appendix A: CBO’s Analytical Approach 30 Appendix B: Sensitivity of Findings to Certain Analytical Choices 37 Definitions 43 About This Document 45 List of Exhibits Initial Benefits Compared With the Official Federal Poverty Threshold 10 1. Percentage of Long-Career Workers With Initial Individual Benefits Below the Poverty Threshold 11 2. Percentage of Long-Career Workers With Benefits Below the Poverty Threshold, Using Different Measures of Benefits 12 3. Comparison Between Long-Career and Short-Career Workers With Initial Benefits Below the Poverty Threshold 13 Initial Individual and Shared Replacement Rates 14 Numerators and Denominators of Replacement Rates 15 4. Median Numerators and Denominators of Replacement Rates for Long-Career Workers Born in the 1960s, Using Individual and Shared Measures 16 Individual Replacement Rates 17 5. Median Replacement Rates for Long-Career Workers, Using Initial Individual Benefits and Two Different Measures of Earnings 18 6. Variations in Median Replacement Rates for Long-Career Workers Born in the 1960s, Using Different Measures of Earnings 19 7. Variations in Median Replacement Rates for Long-Career Workers Born in the 1960s, Using Different Hypothetical Claiming Ages 20 LIST OF EXHIBITS SOCIAL SECURITY REPLACEMENT RATES AND OTHER BENEFIT MEASURES: AN IN-DEPTH ANALYSIS IV Shared Replacement Rates 21 8. Comparison Between Median Replacement Rates for Long-Career Workers Born in the 1960s, Using Individual and Shared Measures and Different Measures of Earnings 22 9. Median Replacement Rates for Long-Career Workers Born in the 1960s, With Initial Shared Benefits and the Last 20 Years of Shared Earnings, by Household Type 23 10. The Relationship Between Initial Household Benefits, Poverty Thresholds, and Shared Replacement Rates for Long-Career Workers Born in the 1960s 24 11. Median Replacement Rates for Long-Career Workers Born in the 1960s, Using Initial Shared Benefits and the Last 20 Years of Shared Earnings, Adjusted for Taxes 25 Multiyear Shared Replacement Rates 26 12. Median Multiyear Replacement Rates for Long-Career Workers, Using the Average of All Shared Benefits and the Last 20 Years of Shared Earnings 27 13. Median Replacement Rates Before and After Widowhood for Long-Career Workers Born in the 1960s, Using Shared Benefits and the Last 20 Years of Shared Earnings 28 14. Uncertainty Surrounding Shared Multiyear Replacement Rates for Long-Career Workers Born in the 1960s 29 List of Tables and Figures Tables A-1. Illustration of Restrictions Used to Produce the Sample for the 1960s Birth Cohort 32 B-1. Median Initial Replacement Rates for Long-Career Workers, Using Different Numerators and the Last 20 Years of Earnings, Including Years With No or Very Low Earnings 39 Figures 1. Median Initial Benefit for Long-Career Workers Born in the 1960s, if Claimed at Age 65 5 A-1. Sensitivity of Shared After-Tax Replacement Rates for Long-Career Workers Born in the 1960s to Alternative Specifications of Federal Income Taxes 34 A-2. Comparison of Earnings Levels for Long-Career Workers With OCACT’s Hypothetical Scaled Workers in the 1960s Cohort 35 A-3. Replacement Rates for OCACT’s Hypothetical Workers and CBO’s Long-Career Workers Born in the 1960s, Using Wage-Adjusted Earnings 36 B-1. Percentage of Long-Career Workers With Initial Individual Benefits Below 200 Percent of the Poverty Threshold 38 B-2. Difference in Median Initial Shared Replacement Rates for Long-Career Workers Born in the 1960s Resulting From Adjustments for Economies of Scale Within the Household 42 Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Summary the 1960s and 1980s, even more are projected lowest earnings quintile are about two to three In this report, the Congressional Budget Office to have retired-worker benefits above that times higher, on average and across cohorts, examines whether Social Security benefits enable threshold. than replacement rates for workers in the retired workers to meet their basic needs and the highest quintile. extent to which benefits replace preretirement ■ Replacement rates that compare benefits earnings. Focusing on workers with long careers, with earnings just before retirement show ■ If future benefits are limited to the annual who generally have higher average earnings than that, across cohorts, benefits replace about revenues credited to Social Security once the all workers, CBO finds that those benefits enable two-fifths of substantial late-career earnings, program’s combined trust funds are exhausted, most of those workers to cover their essential falling short of providing income continuity which is projected to occur in 2031—that is, living expenses as measured by the official federal as workers transition out of the labor force. payable benefits—the fraction of workers with poverty threshold. However, the extent to which (Substantial earnings are annual earnings that initial benefits below the poverty threshold benefits replace preretirement earnings varies are at least half of the worker’s average indexed is projected to increase slightly between the substantially, depending on the way benefits and earnings.) 1940s cohort and the 1960s cohort, and then earnings are measured. to increase substantially more for the 1980s ■ Replacement rates that are designed to cohort. Replacement rates based on payable To show benefits from different perspectives, capture overall changes in the standard of benefits would be significantly lower than the CBO presents alternative specifications of basic living between working years and retirement replacement rates based on scheduled, or full, needs measures and of the Social Security replace- show that Social Security benefits replace a benefits. ment rate—that is, the amount of Social Security significantly higher percentage of average benefits received in retirement, expressed as a earnings over a lifetime, adjusted for changes CBO also compares individual measures of bene- percentage of preretirement earnings. Among in prices over time. fits with household-based measures and before-tax CBO’s findings: measures with after-tax measures. In addition, ■ Because the Social Security benefit formula the agency compares replacement rates that are ■ Retired-worker benefits for most long-career is progressive, meaning that benefits replace based on a single year of benefits with rates that workers born in the 1940s exceed the official a larger share of earnings for lower-income are based on the receipt of benefits over multiple federal poverty threshold. For workers born in workers, replacement rates for workers in the years. According to CBO’s projections: SUMMARY Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis ■ Household benefits, counting those available for Social Security benefits totaled $977 billion, cannot answer the question of whether a retired to a retired worker and his or her spouse, fall accounting for almost one-quarter of all fed- worker has sufficient income in retirement.2 To below the official poverty threshold for only a eral spending. Because most of the program’s determine the adequacy of retirement income, an very small percentage of long-career workers. benefits—69 percent (or $674 billion) in fiscal analysis would need to incorporate all sources of year 2018—are paid to retired workers, Social income as well as a definition of what is considered ■ Replacement rates that are based on benefits Security is often characterized as a retirement “adequate.” and earnings shared within a household program, although it also pays benefits to disabled are similar to replacement rates based on workers and to their dependents. This report presents results from a wide range of individual benefits and earnings. specifications side by side and generally for the Social Security benefits are an important source same group of workers. (Throughout the report, ■ Accounting for payroll taxes and federal of income for the nation’s elderly. In 2018, the the unit of analysis is the individual worker.) By income taxes results in after-tax replacement average benefit paid to a retired-worker beneficiary doing so, the report enables a straightforward rates that are several percentage points higher was about $1,400 per month. However, the total comparison of different measures and sheds light than before-tax replacement rates. amount paid to the household of a retired worker on how the choice of a particular specification is typically greater. More than half of people age affects the results. ■ Replacement rates that are based on average 65 or older are married, and their spouses and benefits received throughout the course of other dependents may also receive Social Security How Did CBO Develop Its Projections? retirement are similar to replacement rates that benefits on the basis of their own work history CBO’s analysis is based on the agency’s long-term are based on a single (initial) year of benefits. or because they are eligible for spousal or other projections of trends in a host of demographic However, based on benefits and earnings dependent benefits. and economic variables, which are the same as shared within a household, married workers those underlying The 2018 Long-Term Budget experience a sharp reduction in replacement Evaluating the adequacy of those benefits is not a Outlook.3 For workers born in the 1940s—the first rates after the death of a spouse. clear-cut task. Some benefit measures, or gauges cohort CBO examined—the projections reflect of benefit generosity, are designed to examine historical earnings data because those workers The analysis in this report focuses on Social whether Social Security enables retired workers to have nearly completed their careers. For workers Security benefits and does not examine other afford essential living expenses. Such basic needs in later cohorts—those born in the 1960s and the sources of income that retired workers may have. measures include comparisons of benefits with the 1980s—more years of their earnings are projected. Therefore, it does not provide an assessment of the official federal poverty threshold. Other measures, Those earnings and other characteristics, includ- adequacy of retirement income overall. known as Social Security replacement rates, are used to determine the extent to which benefits enable Background retirees to maintain their preretirement standard of 2. For an introduction to the issues surrounding the measurement of retirement income, see Congressional Social Security is the largest single program in living. Budget Office, Measuring the Adequacy of Retirement the federal budget.1 In fiscal year 2018, outlays Income: A Primer (October 2017), www.cbo.gov/ Social Security benefits are just one of several publication/53191. 1. For an overview of Social Security, see Congressional different sources of income available to retirees, 3. See Congressional Budget Office, The 2018 Long- Budget Office, Social Security Policy Options, 2015 and workers may have income other than earn- Term Budget Outlook (June 2018), www.cbo.gov/ (December 2015), www.cbo.gov/publication/51011. ings. Consequently, looking at those benefits alone publication/53919. 2 BACKGROUND Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis ing the age at which benefits are claimed, were 2017, about $5,000, is comparable to working full funds’ balances. As long as a trust fund’s balance projected using CBO’s long-term model.4 That time for about a third of the year while earning the is sufficient to cover required payments, benefits model draws on data from the Social Security federal minimum wage.) Including workers with can be paid without the need for any legislative Administration’s Continuous Work History shorter careers in the computation of individual action. However, according to CBO’s most recent Sample and from the Census Bureau’s Survey of retired-worker replacement rates would make the projections, under current law, the balances in the Income and Program Participation and Current analysis less meaningful because those workers combined OASDI trust funds would be exhausted Population Survey. (See Appendix A for more often receive spousal or survivor’s benefits that are in 2031.6 So, in addition to projecting scheduled details about CBO’s analytical method.) significantly higher than benefits that are based on benefits, which are calculated under the Social their own work record. Security Act without regard to the Social Security What Population Did CBO Analyze? trust funds’ balances, CBO projected payable CBO’s analysis focused on workers who, on the What Measures of Social Security Benefits benefits—that is, benefits as calculated under basis of their own earnings, are eligible to receive Did CBO Examine in Its Analysis? current law and reduced as necessary to conform Social Security retired-worker benefits by the time This report focuses on the benefits available to the limits imposed by the trust funds’ balances. they reach the earliest eligibility age (EEA) of 62. to retired workers through the Old-Age and Payable benefits would be less than the scheduled (Those workers are not disabled and are eligible for Survivors Insurance (OASI) program, the largest amounts once the trust funds were exhausted Old-Age Insurance, or OAI, benefits. They may component of Social Security. (Social Security because annual outlays would be limited to annual claim benefits at any age starting with the EEA.) consists of two parts: OASI, which pays benefits revenues credited to Social Security. Because workers who are not eligible for benefits to retired workers as well as to eligible dependents generally have lower lifetime earnings than work- and survivors; and Disability Insurance, or DI, Most important, CBO developed its findings ers who are eligible, workers examined for this which makes payments to disabled workers and to using both individual and household-based mea- report have higher lifetime earnings than workers their dependents until those workers reach the age sures of benefits. Although individual measures are in the overall population. at which they are eligible to receive full retired- simpler and more common in studies of retire- worker benefits under OASI.) Because CBO’s ment income, household-based measures provide a In addition, CBO focused primarily on workers analysis focused on nondisabled workers who more comprehensive perspective on benefits. with significant attachment to the labor force. are eligible for retirement benefits on the basis Referred to in this report as long-career work- of their own work history, benefits paid through In CBO’s analysis, individual benefit measures ers, they are defined as having at least 20 years the DI program were excluded from all measures were generally constructed using only the retired- of significant earnings—that is, earnings above described in this report. worker benefit—which is based on the beneficia- 10 percent of the average wage index (AWI) in ry’s own earnings—even though the beneficiary each year.5 (For example, 10 percent of the AWI in Benefits for OASI and DI alike are financed from might also be eligible for a benefit on the basis of a trust funds, which are credited with tax revenues, spouse’s earnings. If the spouse is alive, that benefit 4. For a brief summary of CBO’s long-term model, see mainly from payroll taxes, and interest on the is the spousal benefit; if the spouse is deceased, Congressional Budget Office, An Overview of CBOLT: The Congressional Budget Office Long-Term Model Board’s 2015 Technical Panel on Assumptions and (April 2018), www.cbo.gov/publication/53667. Methods. See 2015 Technical Panel on Assumptions 6. See Congressional Budget Office, The 2018 Long- 5. In focusing on long-career workers, CBO follows and Methods, Report to the Social Security Advisory Board Term Budget Outlook (June 2018), www.cbo.gov/ recommendations of the Social Security Advisory (September 2015), https://go.usa.gov/xUFa6. publication/53919. 3 BACKGROUND Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis that benefit is the survivor’s benefit.7 However, distribution for a given group.) For simplicity, eligible to receive.9 This measure is used in generally, beneficiaries can receive spousal benefits the individual benefit measures described in this the comparison of benefits with poverty only if their own earnings are substantially lower report generally include only individual retired- thresholds. Because official federal poverty than their spouse’s earnings, and they can receive worker benefits. thresholds vary on the basis of household survivor’s benefits only if their earnings are lower size, CBO compares total household benefits than those of their deceased spouse. As a result, a For a more comprehensive analysis, CBO included with the corresponding poverty threshold to typical long-career worker is either ineligible for not only the benefit that the worker is eligible to determine whether benefits fall short of that that benefit or eligible only for a small additional receive on the basis of his or her own earnings, but threshold. (For example, in 2018, the poverty benefit. For example, for long-career workers born also any spousal or survivor’s benefit and, most threshold for a single person age 65 or older in the 1960s, median initial individual benefits, important, any benefit that the worker’s spouse is was $12,043, and the threshold for two people if claimed at age 65, are projected to be about eligible to receive. Because spouses tend to share was $15,178. Therefore, a single person with $20,100 per year, whereas the median initial resources and most retired workers are married, a benefit of $10,000 would have a household benefits that also include any spousal or survivor’s such a household-based measure better accounts benefit below the poverty threshold, but a benefits are projected to be about $20,300 (see for all the Social Security benefits available to married couple with a household benefit Figure 1). (Initial benefits are defined as benefits a retired worker. (In this report, a household is of $20,000 would have a benefit above the received at age 65, if first claimed at that age. defined as consisting of a single person or a mar- poverty threshold.) Median benefits are those in the middle of the ried couple.)8 ■ The second measure—referred to as a shared 7. At the full retirement age, the eligible spouse of a retired The household-based measures used for CBO’s measure—reflects the amount of benefits worker is entitled to benefits that amount to 50 percent analysis fall into two categories: attributable to each person in the household. of the retired worker’s primary insurance amount (PIA) if This type of measure is used in the analysis of he or she is not eligible for benefits on the basis of his or ■ The first measure—referred to as a household Social Security replacement rates. The shared her own earnings. If the spouse also has earned benefits measure—reflects the sum of all retired- measure is defined as the sum of all retired- but has a PIA that is less than 50 percent of the primary beneficiary’s PIA, the spouse’s payments are increased worker, spousal, and survivor’s benefits worker, spousal, and survivor’s benefits that to meet the 50 percent threshold. A spouse whose that the worker and his or her spouse are the worker and his or her spouse are eligible PIA (based on his or her own earnings) is greater than to receive, adjusted for household size and for 50 percent of the primary beneficiary’s PIA receives no economies of scale to be comparable with the additional amount. (If the spouse claims benefits before individual-based measures. Because household his or her full retirement age, the spousal benefits are lower than 50 percent of the primary beneficiary’s PIA.) 8. CBO’s long-term model does not project living arrangements. Although defining a household as a 9. If the worker is married, the household benefit measure At the full retirement age, the eligible widow(er) of single person or a married couple represents a departure is the sum of all retired-worker and spousal benefits a retired worker is generally entitled to benefits that from the traditional definition, which characterizes a that both the worker and his or her spouse are eligible amount to 100 percent of the retired worker’s PIA, if he household as all people who live under the same roof, to receive on the basis of their earnings histories. If the or she is not eligible for benefits on the basis of his or her the definition that CBO uses is consistent with the living worker is not married, the household benefit measure own earnings. If the widow(er) also has earned benefits arrangements of 72 percent of people age 65 or older. reflects the retired-worker, spousal, and survivor’s benefits but has a PIA that is less than 100 percent of the primary See Loraine A. Wes and others, 65+ in the United States: that the worker is eligible to receive on the basis of his or beneficiary’s PIA, the widow(er)’s payments are generally 2010, Current Population Reports (Census Bureau, her earnings or on the basis of the earnings of a divorced increased to meet the 100 percent threshold. June 2014), https://go.usa.gov/xUFCc (PDF, 12 MB). or deceased spouse. 4 BACKGROUND Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis members tend to share expenses for housing, 1980s—the agency uses the middle cohort, Figure 1 . utilities and other items, two spouses generally the 1960s, in illustrative examples throughout Median Initial Benefit for Long-Career need less than twice the income that they this report.11 The specific outcomes differ Workers Born in the 1960s, if Claimed at would need if living separately. Therefore, across cohorts; however, the patterns in CBO applied a common economies-of-scale findings based on scheduled benefits are Age 65 adjustment known as a square-root scale, roughly similar. Findings based on payable Annual Amounts in 2018 Dollars which implies that a married person would benefits worsen for later cohorts following the 30,000 need about 30 percent less income than a exhaustion of the combined Social Security single person living alone to maintain the trust funds.) same general standard of living.10 Although shared benefits are generally larger 20,000 Because of that adjustment, the median shared than individual benefits, CBO finds that shared benefit for long-career workers is substantially replacement rates are comparable with individual higher than the median individual benefit. replacement rates. Accounting for shared resources 26,500 For example, for long-career workers born within a household increases both the benefits 10,000 20,100 20,300 in the 1960s, the median shared benefit is (numerators) and the earnings (denominators) about one-third higher than the median used in the calculation of replacement rates for individual benefit (see Figure 1). (Although workers who are married, so shared replacement CBO presents measures of Social Security rates are similar to individual replacement rates for 0 benefits for three cohorts—1940s, 1960s, and men and women combined.12 That relationship Retired-Worker Retired-Worker Shared differs for men and women separately, however, Benefit Benefit Plus Benefit because women tend to have lower earnings Spousal or 10. The square-root scale is used in the construction of than men and because the Social Security bene- Survivor’s the equivalence scales for the Supplemental Poverty Benefit Measure. See Liana Fox, The Supplemental Poverty fit formula is progressive, meaning that benefits Measure: 2017, Current Population Reports (Census replace a larger share of earnings for lower-income Source: Congressional Budget Office. Bureau, September 2018), https://go.usa.gov/xEU5C workers. As a result, individual replacement rates (PDF, 692 KB). Applying a square-root scale to compute for female workers tend to be several percentage shared benefits differs from the traditional shared-benefit computation, which divides the combined benefits of the points higher than their shared replacement rates, worker and his or her spouse by two. In this report, CBO 11. The main findings for all three cohorts are available whereas the opposite is true for male workers. presents the results using the square-root adjustment to through an interactive tool that accompanies this report better capture the well-being of people in two-person at www.cbo.gov/publication/55038. How Do Social Security Benefits Compare With households relative to one-person households. However, most findings using the traditional shared-earnings 12. Shared replacement rates are identical to individual Poverty Thresholds? approach that does not incorporate economies of scale replacement rates for workers who never marry. For Comparing individual retired-worker benefits with are similar to the findings based on the square-root workers who marry, the two rates are rarely the same. For poverty thresholds demonstrates that the majority scale adjustment (see Appendix B for a comparison of example, they would be the same if the two spouses had of long-career workers have initial retired-worker replacement rates with and without economies of scale identical earnings in each year that were counted toward benefits that exceed those thresholds. When the applied). the measures of earnings and benefits. 5 BACKGROUND Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis analysis is expanded to include household benefits, do not include all ineligible people (for instance, Social Security benefits, it does not project income an even larger fraction of workers is projected to single workers with insufficient work histories), from other sources, which are encompassed by the have initial benefits above the poverty thresholds. who are counted in the Census measures. Census measures. Those findings suggest that Social Security benefits generally enable the vast majority of long-career Second, CBO’s analysis incorporated the assump- How Do Benefits Compare With Preretirement workers to meet their basic needs, as measured in tion that workers and their spouses both claim Earnings? relation to poverty thresholds, in retirement. benefits at age 65, which generally results in Social Security replacement rates vary substantially slightly higher estimates of annual benefits than depending on how they are measured. On the one For many workers with shorter careers, however, the actual benefits received because more than half hand, replacement rates designed to compare ben- initial household benefits fall short of meeting of eligible people claim before or at age 65. Social efits with earnings just before retirement—that is, basic needs. For example, among OAI-eligible Security benefits are reduced if claimed before the late-career replacement rates—show that benefits workers born in the 1960s who have fewer than full retirement age (FRA) and increased if claimed alone are generally insufficient to maintain work- 20 years of significant earnings, about 40 percent afterward. By CBO’s estimate, individual first-year ers’ preretirement income as they leave the labor are projected to have household benefits below benefits at the projected age of claiming are about force. For example, for workers born in the 1960s, the poverty thresholds. Nevertheless, because 5 percent lower, on average and across cohorts, median late-career replacement rates that are based short-career workers make up a small percentage than hypothetical benefits claimed at age 65. (The on substantial earnings (adjusted for changes in of all OAI-eligible workers in that cohort (14 per- projected age for claiming benefits is the age at prices over time) in the last five years before the cent), the percentage of all OAI-eligible workers which CBO’s long-term model projects a future workers reach age 62 amount to less than 40 per- projected to have household benefits above the beneficiary would first claim benefits on the basis cent. (Substantial earnings are annual earnings that poverty thresholds is high (90 percent). of his or her work history and other individual are at least half of the worker’s average indexed characteristics.) earnings—that is, earnings over a person’s lifetime, It is important to note that comparing Social adjusted for changes in average wages over time.) Security benefits with the poverty thresholds dif- Third, many people receive income from sources fers from poverty-rate analyses conducted by the other than Social Security, such as pensions or On the other hand, replacement rates that focus Census Bureau for several reasons.13 First, the sam- other retirement income, and earnings.14 Although on the overall changes in the standard of living ple that CBO analyzed centers on workers who CBO’s long-term model projects earnings and between all working years and retirement show are eligible for retired-worker benefits on the basis that Social Security benefits replace a signifi- of their own earnings, rather than on all people. cantly higher percentage of average earnings over 14. Studies that considered multiple income sources available The household measures include workers’ spouses a lifetime, adjusted for changes in prices over in retirement found that Social Security plays an who may not be eligible for retired-worker benefits important role in enabling beneficiaries to avoid poverty. time. For workers born in the 1960s, the median on the basis of their own earnings; however, they For more information, see Irena Dushi, Howard M. Iams, replacement rate based on all earnings from age and Brad Trenkamp, “The Importance of Social Security 22 through age 61, including years with no or Benefits to the Income of the Aged Population,” Social very low earnings, is 55 percent. (Throughout this 13. For recent poverty estimates by the Census Bureau, see Security Bulletin, vol. 77, no. 2 (2017), https://go.usa. report, earnings refer to all labor earnings that Kayla Fontenot, Jessica Semega, and Melissa Kollar, gov/xEXMG; and C. Adam Bee and Joshua W. Mitchell, Income and Poverty in the United States: 2017, Current Do Older Americans Have More Income Than We Think? the workers have, not only the earnings under the Population Reports (Census Bureau, September 2018), Working Paper 2017-39 (Census Bureau, July 2017), https://go.usa.gov/xEXMV (PDF, 2.13 MB). https://go.usa.gov/xRG87. 6 BACKGROUND Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis taxable maximum.15 That measure of earnings changes in average claiming patterns over time. survivor’s benefit.17 However, if the two-person fully captures the income from workers’ wages and In contrast, multiyear benefits capture average household becomes a single-person household, the salaries.) benefits from the age at which CBO’s long-term surviving spouse is no longer able to share fixed model projects a future beneficiary would first expenses such as housing and utilities. As a result, How Does Taking Taxes Into Account claim benefits until death. Comparing findings his or her living standard tends to decline because Affect the Results? using those two types of measures can be infor- those fixed expenses consume a larger share of the Accounting for taxation increases replacement mative because the benefits that retired workers budget and leave less room for other spending. In rates. In its analysis, CBO accounted for Social receive may change over time. this example, the shared benefits would be lower Security and Medicare payroll taxes by subtracting after the death of a spouse than benefits adjusted the employee’s share of such taxes from each year There are two main reasons why real (inflation-­ for economies of scale during the marriage.18 of preretirement earnings. To account for federal adjusted) shared Social Security benefits may income taxes, CBO applied the federal tax law in change after the initial year in which they are Overall, multiyear replacement rates that are place in each year of earnings or benefits, subtract- received. First, a worker or his or her spouse could based on shared earnings and shared benefits ing projected taxes from the before-tax amounts. change his or her claiming status over time. For are similar to initial measures. For example, for In general, workers face higher average income example, a worker might claim benefits as soon as long-career workers born in the 1960s, the multi- tax rates than Social Security beneficiaries do. For he or she is eligible at age 62, whereas the spouse year shared replacement rate is about 1 percentage long-career workers born in the 1960s, account- might continue working and claim benefits a point higher than the initial shared replacement ing for both payroll taxes and federal income few years later. As a result, the shared benefits in rate—46 percent rather than 45 percent. However, taxes increases the median after-tax replacement the first few years would consist solely of a single married workers who become widowed in retire- rate (based on the last 20 years of earnings) by retired-worker benefit, whereas the shared benefits ment experience a significant decline in their over 5 percentage points, from 45 percent to in later years would consist of two retired-worker replacement rates. For married workers born in the 51 percent. benefits (divided by the square root of two, about 1960s, the median replacement rate declines by 1.4, to account for economies of scale). Moreover, 10 percentage points after the death of a spouse, How Do Benefit Measures Change if claimed at a later date, the spouse’s benefit could from 53 percent to 43 percent. Because a newly Over the Course of Retirement? be substantially higher because of the additional widowed worker can no longer share common In this report, CBO compares two types of benefit years of work, leading to a recomputation of the household expenses with a spouse, that worker’s measures: single-year and multiyear measures. The retired-worker benefit. In this example, the shared cost of living increases, reducing the amount of agency defines single-year—or initial—benefits benefits may increase over time. goods and services that can be afforded using the as hypothetical benefits received at age 65, if first claimed at that age. Holding the age at which Second, shared benefits could change over time benefits are claimed fixed at 65 enables a more because of marriage, widowhood, or divorce.16 17. The survivor’s benefit paid to a deceased worker’s spouse generally equals 100 percent of the worker’s PIA, if the straightforward comparison among different For example, after the death of one spouse, the survivor claimed at the full retirement age or later. cohorts of retirees by taking out the effect of surviving spouse is eligible to receive the higher of his or her own retired-worker benefit or the 18. If the traditional shared-earnings approach was used in this example instead of the economies-of-scale 15. The taxable maximum is the maximum amount of adjustment, shared benefits would be generally higher annual earnings subject to the Social Security payroll tax 16. In this report, widowhood refers to the state of being a during widowhood. See Appendix B for results that are ($128,400 in 2018). widow or widower. based on the traditional shared-earnings approach. 7 BACKGROUND Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis recalculated benefit. The lower median replace- age for claiming Social Security benefits. Finally, How Does This Report Relate to Other ment rate thus reflects the loss of economies of the projections in this report are uncertain because Analyses by CBO? scale within the household. of the inherent methodological challenges associ- This report builds on information presented in ated with projecting outcomes far into the future. Measuring the Adequacy of Retirement Income: A How Does Uncertainty Affect the Projections? (See Appendix A for a description of CBO’s Primer, which was published in October 2017.19 CBO’s findings depend critically on its projections analytical method.) The agency previously published several differ- of key economic, demographic, and behavioral ent measures of Social Security benefits in CBO’s factors, and all such long-term projections are As an illustration of the uncertainty, CBO found 2016 Long-Term Projections for Social Security: inherently uncertain. The main demographic that projections of labor force participation that Additional Information.20 In that report, CBO factors that affect the results are projections of are about 3 percentage points higher or lower showed mean Social Security replacement rates mortality and fertility. (Projections of mortality than in CBO’s 2018 extended baseline would that were based on benefit amounts net of income affect the number of years that beneficiaries receive result in replacement rates for workers born in taxes paid on benefits, reflecting the amounts benefits. Projections of fertility affect both gross the 1960s that are less than 1 percentage point received by beneficiaries. This report presents both earnings and federal income taxes.) The main higher or lower than the replacement rates shown. before-tax and after-tax measures of benefits and economic factors are labor force participation and (The replacement rates in this report are computed earnings, as well as a number of new specifica- the growth rate of productivity, which influence using projections of labor force participation tions that help address additional questions about earnings trajectories as well as projected benefits. and other variables that are the same as those benefits. This report also shows median replace- (The labor force participation rate is the percent- underlying CBO’s 2018 extended baseline, which ment rates, rather than the mean replacement rates age of people in the civilian noninstitutionalized generally reflects current law and spans the period presented in the previous report, because median population who are age 16 or older and either from 2018 through 2048.) Alternative projections rates are more widespread in the literature on working or actively seeking work. The growth rate of the growth rate of productivity that are 0.5 per- the adequacy of retirement income. (All median of productivity is calculated as the growth of total centage points higher or lower than those under­ replacement rates shown in this report reflect the factor productivity, which is the growth of real lying the extended baseline would cause the rates replacement rate for the person in the middle of output that is not explained by growth in labor or to be about 1 to 2 percentage points higher or the distribution for a given group. That person capital.) The behavioral factors that can affect the lower than shown. is not necessarily the person with earnings in the results include changes in household formation middle of a distribution of shared lifetime earnings and dissolution, as well as changes in the average for that group.) 19. See Congressional Budget Office, Measuring the Adequacy of Retirement Income: A Primer (October 2017), www.cbo. gov/publication/53191. 20. See Congressional Budget Office, CBO’s 2016 Long-Term Projections for Social Security: Additional Information (December 2016), www.cbo.gov/publication/52298. The data presented in that report were updated in September 2018 and can be found at www.cbo.gov/ publication/54428. 8 Initial Benefits Compared With the Official Federal Poverty Threshold Exhibits 1 through 3 evaluate the extent to which projected to be received in the first year of claim- are workers with 20 or more years of significant initial (single-year) Social Security benefits, if ing. (As shown in Table B-1 in Appendix B, that earnings—that is, earnings above 10 percent of the claimed at age 65, enable beneficiaries to meet variation would have little effect on the results.) average wage index in each year). The household their basic living needs in retirement. For those benefit measures also include any OAI benefits exhibits, CBO used 100 percent of the official Exhibit 1 addresses individual benefits. An indi- received by the spouses of the long-career workers, federal poverty threshold as the minimum dollar vidual benefit is the retired-worker benefit, which although the spouses may not be eligible workers amount needed to meet basic living needs. In is based on individual earnings through age 61, themselves. 2018, for example, that amount was $12,043 for that the worker could receive if claimed at age 65. a single person age 65 or older; for two people, Exhibit 3 illustrates how benefit measures for it was $15,178. (See Figure B-1 in Appendix B Exhibits 2 and 3 present information about both that group of workers compare with measures for for findings based on 200 percent of the official individual and household benefit measures. A short-career workers, who are eligible for OAI federal poverty threshold.) household benefit is the sum of benefits that the benefits on the basis of their individual earnings worker and his or her spouse could receive on through age 61 but who have fewer than 20 years Holding the age at which benefits are claimed the basis of their earnings through age 61 if each of significant earnings. Together, the long-career fixed at 65 enables an easier comparison among spouse claimed benefits at age 65. and short-career OAI-eligible workers who are not different cohorts of retirees by taking out the effect disabled make up about three-quarters of people of changes in average claiming patterns over time. The population analyzed in Exhibits 1 and 2 con- who survive through age 62. Of those OAI-eligible Although that hypothetical benefit measure does sists of retired workers who had long careers, who workers, more than 80 percent are long-career not account for variation in the observed or pro- are eligible to receive Old-Age Insurance (OAI) workers. jected timing of retirement, it provides a simple benefits on the basis of their individual earnings, approximation of retired-worker benefits that are and who are not disabled. (Long-career workers INITIAL BENEFITS COMPARED WITH THE OFFICIAL FEDERAL POVERTY THRESHOLD Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 1 . About 15 percent of long-career workers born in the Percentage of Long-Career Workers With Initial Individual Benefits Below the Poverty Threshold 1940s receive initial retired-worker benefits that are Percent insufficient to meet basic living needs as measured by the official federal poverty threshold. (Initial 40 Both Sexes benefits are defined as benefits received at age 65, if first claimed at that age.) For workers whose initial Scheduled Payable benefits are below the poverty threshold, the average 20 Benefits Benefits shortfall is about one-sixth of the threshold amount. 29 15 15 13 18 6 Because scheduled benefits for subsequent cohorts 0 grow with average wages, but the poverty thresholds Men generally grow more slowly—with prices—the frac- 40 tion of retired workers whose initial scheduled ben- efits are projected to be below the poverty threshold 20 is smaller for later cohorts, falling to 6 percent for 26 the 1980s cohort. However, with payable benefits, 8 11 0 6 6 6 that fraction increases to 29 percent for the 1980s cohort, indicating that the reduction in benefits fol- 40 Women lowing the projected depletion of the Social Security trust funds would prevent an additional 23 percent of workers in that cohort from meeting their basic 20 needs using benefits alone. (The combined Social 33 26 26 24 Security trust funds are projected to be depleted in 19 0 7 2031.) 1940s 1960s 1980s Considering workers by sex reveals a large discrep- Birth Cohort ancy in initial benefits for male and female workers. Source: Congressional Budget Office. More than a quarter of female workers born in the 1940s have initial benefits below the poverty thresh- Initial individual benefits are based on the assumption that workers first claim benefits at age 65. Benefits are computed for all old, whereas the same is true for only 6 percent of people who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. All benefit amounts are before taxes. male workers in that cohort. That discrepancy is pro- jected to decline over time as women’s earnings have Scheduled benefits are benefits as calculated under the Social Security Act, regardless of the balances in the trust funds. Payable grown, and are projected to continue growing, faster benefits are benefits as calculated under the act, reduced as necessary to ensure that outlays do not exceed the Social Security system’s revenues once the balances in the combined trust funds are exhausted, which is projected to occur in 2031. than men’s, narrowing the gap in benefits. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. The federal poverty threshold used here is the threshold for one person age 65 or older, adjusted for growth in prices over time. 11 INITIAL BENEFITS COMPARED WITH THE OFFICIAL FEDERAL POVERTY THRESHOLD Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 2 . Expanding the measure of benefits for an individual Percentage of Long-Career Workers With Benefits Below the Poverty Threshold, worker to include any spousal or survivor’s benefit Using Different Measures of Benefits that the worker is eligible to receive reduces the percentage of workers with initial benefits below Percent the poverty threshold. Moving from that measure Both Sexes to a household-level measure, which also includes 40 Retired-Worker all benefits available to the worker’s spouse, pro- Retired-Worker Benefits Plus the vides a more comprehensive perspective on benefits Benefits Worker’s Spousal or 20 Survivor’s Benefits Household and results in a greater reduction in the percentage Benefits of workers with initial benefits below the poverty 15 12 5 13 5 5 11 6 3 threshold. 0 40 Men On average, including spousal and survivor’s benefits has a relatively small effect on the percentage of long-career workers with individual benefits below 20 the poverty threshold, reducing that percentage from 3 3 5 15 percent to 12 percent for workers born in the 6 6 8 7 6 3 0 1940s. The reduction is considerably larger for female workers than for male workers. Because women tend 40 Women to have lower earnings than their spouses, they are more likely to receive a spousal or survivor’s benefit. 20 26 Moving to a household-level measure, which also 19 6 5 18 14 3 accounts for the benefits available to the worker’s 0 7 7 1960s spouse, yields a significantly smaller fraction of 1940s 1980s workers who cannot meet their basic needs with Birth Cohort benefits alone. On average, only about 5 percent Source: Congressional Budget Office. of retired workers born in the 1940s have house- hold benefits that fall below the household poverty Initial retired-worker benefits are based on the assumption that workers first claim benefits at age 65. Retired-worker benefits plus spousal or survivor’s benefits include benefits that workers can receive on the basis of their spouse’s earnings. If the worker is married, threshold. For female workers, that number is 7 per- the initial household benefit is the sum of the worker’s and the spouse’s benefits, reflecting the assumption that both spouses first cent, which is substantially lower than the 18 per- claim benefits at age 65. If the worker is not married, the initial household benefit captures the benefit that the worker is eligible to cent of women who have retired-worker, spousal, receive on the basis of his or her own earnings and on the basis of the earnings of the worker’s divorced or deceased spouse, if the and survivor’s benefits below the poverty threshold. worker was married. Benefits are computed for all people who are eligible to claim retirement benefits at age 62 and who are not On average, men have larger benefits than women receiving any benefit at age 61. All benefit amounts are before taxes. because they generally have higher lifetime earnings. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit Therefore, when the benefits of female workers and the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In their (male) spouses are combined, those benefits addition, workers who receive Disability Insurance benefits are excluded. are more likely to exceed the household poverty The federal poverty threshold used here is the threshold for one person (or two people) age 65 or older, adjusted for growth in prices threshold than women’s own benefits are to exceed a over time. The household benefit is compared with the poverty threshold in the year that the worker turns age 65. single-person poverty threshold.  12 INITIAL BENEFITS COMPARED WITH THE OFFICIAL FEDERAL POVERTY THRESHOLD Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 3 . In contrast with the outcomes for long-career work- Comparison Between Long-Career and Short-Career Workers With Initial Benefits ers, which are the focus of this report, the outcomes Below the Poverty Threshold for short-career workers—that is, workers with fewer than 20 years of significant earnings—look sub- Percent stantially worse. (Those workers make up less than Percentage of Workers With Initial Retired-Worker 20 percent of all OAI-eligible workers who survive Benefits Below the Poverty Threshold, by Length of Career through age 62.) 100 Short-Career Workers When only individual retired-worker benefits All OAI-Eligible are included in the computation, 94 percent of 50 Long-Career 94 Workers Workers 84 short-career workers born in the 1940s have initial 67 28 6 benefits that fall below the poverty threshold. That 15 13 23 14 0 share is six times larger than the share of long-career workers born in the 1940s with initial benefits below Percentage of Workers With Initial Household Benefits Below the Poverty Threshold, by Length of Career the threshold. Because short-career workers have 100 more years with no or very low earnings, median benefits for that group amount to less than half of the median benefits for long-career workers. 50 11 10 Expanding the measure of benefits to household 5 43 5 40 3 33 7 benefits reduces the percentage of short-career 0 1940s 1960s 1980s workers who cannot meet their needs using benefits Birth Cohort alone; however, that percentage remains high. For instance, 43 percent of short-career workers born Source: Congressional Budget Office. in the 1940s have household benefits below the household poverty threshold. For those workers, the Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. Short- differences in the percentages using retired-worker career workers are workers with fewer than 20 years of earnings above 10 percent of the average wage index. Workers who receive Disability Insurance benefits are excluded. benefits and household benefits are attributable to benefits that those workers receive on the basis of Initial retired-worker benefits are based on the assumption that workers first claim benefits at age 65. If the worker is married, the their spouse’s earnings, as well as to benefits that initial household benefit is the sum of the worker’s and the spouse’s benefits, reflecting the assumption that both spouses first claim benefits at age 65. If the worker is not married, the initial household benefit captures the benefit that the worker is eligible to receive their spouses receive. Of short-career workers born on the basis of his or her own earnings and on the basis of the earnings of the worker’s divorced or deceased spouse, if the worker in the 1980s, 33 percent are projected to receive was married. Benefits are computed for all people who are eligible to claim retirement benefits at age 62 and who are not receiving household benefits below the poverty threshold, any benefit at age 61. All benefit amounts are before taxes. compared with only 3 percent of long-career The federal poverty threshold used here is the threshold for one person (or two people) age 65 or older, adjusted for growth in prices workers.  over time. The household benefit is compared with the poverty threshold in the year that the worker turns age 65. OAI = Old-Age Insurance. 13 Initial Individual and Shared Replacement Rates Exhibits 4 through 11 present CBO’s projections Shared replacement rates are constructed to better economies of scale. For an individual who was of initial Social Security replacement rates under capture household resources. The numerator for never married, no adjustment for economies of different specifications of benefits (numerators) the shared replacement rate is the sum of Old-Age scale is applied.) and earnings (denominators). Exhibits 5 through and Survivors Insurance benefits that the worker 7 show individual replacement rates; Exhibits 9 and his or her spouse are eligible to receive on In all of the exhibits, CBO analyzed the same through 11 show shared replacement rates; and the basis of their earnings, including spousal and group of long-career workers who are eligible to Exhibits 4 and 8 compare individual and shared survivor’s benefits, if any, divided by the square receive OAI benefits on the basis of their individ- measures side by side. root of two to account for economies of scale. The ual earnings and who do not receive Disability denominator for the shared replacement rate is Insurance benefits at age 61 or at any time after- Individual replacement rates are constructed using constructed using the sum of the worker’s and his ward. (The shared benefit measures also include only the retired-worker benefit in the numerator or her spouse’s earnings in each year of marriage, any OAI benefits received by the spouses of the and the worker’s own earnings in the denominator. divided by the square root of two. (In years when long-career workers, although the spouses may not the worker is not married, his or her earnings be OAI-eligible workers themselves.) enter the calculation without the adjustment for Numerators and Denominators of Replacement Rates Exhibit 4 illustrates the extent of variation in both The denominators of the replacement rates— purchasing power of retirees’ benefits with the the numerators and denominators of individual preretirement earnings—vary on the basis of the purchasing power of their own earnings when they and shared replacement rates for workers born in time span for earnings, what earnings are counted, were still working; wage adjustment is used to the 1960s. The numerators of the replacement and whether the earnings are adjusted for overall compare the purchasing power of retirees’ benefits rates—Social Security benefits—vary on the basis changes in prices or wages. Price-adjusted earnings with the purchasing power of earnings of work- of the age at which benefits are claimed, whether account for inflation and reflect the purchasing ers who are currently in the labor force. Because they are individual or shared, and whether they are power of earnings over time. Wage-adjusted earn- wages generally grow faster than prices, using price measured in a single year or throughout the course ings account for changes in the average wage index indexing results in lower indexed earnings and, of retirement. and reflect the increase in the average wage earned therefore, a higher replacement rate than does by workers in the national economy over time. wage indexing. Price adjustment is generally used to compare the NUMERATORS AND DENOMINATORS OF REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 4 . Benefits and earnings that can be used in the calcula- Median Numerators and Denominators of Replacement Rates for Long-Career Workers tion of replacement rates vary substantially depend- Born in the 1960s, Using Individual and Shared Measures ing on how they are measured. Shared benefits at hypothetical claiming ages are larger than individual Annual Amounts in Thousands of 2018 Dollars retired-worker benefits because of the addition of Numerators (Benefits) spousal and survivor’s benefits and of any benefits Hypothetical Benefit, if Claimed at Age 62 21 available to the worker’s spouse, along with the 16 26 adjustment for economies of scale (see the top four Hypothetical Benefit, if Claimed at Age 65 20 sets of bars in the upper panel). Shared benefits at Hypothetical Benefit, if Claimed at FRA 30 the projected claiming ages are similar to individual 23 Shared 38 benefits because spouses typically claim benefits Hypothetical Benefit, if Claimed at Age 70 in different years (see the fifth set of bars). (The 29 Benefit at Projected Claiming Age 19 Individual projected shared first-year benefit generally reflects 19 Projected Average Benefit 26 only one spouse’s benefit because most spouses do From Age 62 Until Death not claim benefits in the same year. In contrast, the shared benefit at the hypothetical age of claiming is Denominators (Earnings) constructed by summing benefits of both spouses 65 Last 5 Years of Substantial Earnings (P) 55 using the benefit amounts for that hypothetical age. Last 5 Years of Earnings Above $1,000 (P) 56 See Appendix B.) 51 Last 5 Years, Including Years 54 Projected average annual shared benefits received With No or Very Low Earnings (P) 48 from age 62 until death—multiyear benefits—are Last 20 Years, Including Years 56 With No or Very Low Earnings (P) 45 very similar to hypothetical shared benefits claimed 72 at age 65 (both are about $26,000) because of Highest 35 Years of Earnings (W) 56 several offsetting factors, including claiming patterns Highest 35 Years of Earnings (P) 53 within households, changes in the composition 41 All Earnings From Age 22 48 of households over time, and growth in benefits Through Age 61, Including Years With 37 resulting from cost-of-living adjustments. (Multiyear No or Very Low Earnings (P) 0 10 20 30 40 50 60 70 80 shared benefits account for all years of benefits received by the household. In some years, only one Source: Congressional Budget Office. person in the household receives a benefit—for Shared measures of benefits and earnings reflect benefits and earnings of both spouses, adjusted for economies of scale. Individual example, when only one person has claimed benefits benefits and earnings include only retired-workers’ benefits and those workers’ own earnings. The average benefit from age 62 until or after a spouse has died—and in other years, both death is projected only on a shared basis. For earnings measures (denominators), “P” indicates that the earnings have been adjusted household members receive benefits.) for changes in prices, and “W” denotes that earnings have been adjusted for changes in wages. The last 5 (or last 20) years of earnings, including years with no or very low earnings, are computed starting with the last year with positive earnings before age 62, In calculating the denominators, average earnings stepping back through 4 (or 19) adjacent years, and averaging those values, including years with no or very low earnings. (Starting at measured over longer time spans tend to be lower the last year with positive earnings better captures the final years of work by excluding years when some workers may have already than earnings received in the last five years of work. retired.) Substantial earnings are annual earnings that are at least half of the worker’s (individual or shared) average indexed earnings. For earnings measured over longer time spans, All values are expressed on an individual-worker basis, are before taxes, and reflect annual amounts. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with adjusting those earnings for changes in wages over significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive time, rather than for changes in prices, results in a Disability Insurance benefits are excluded. FRA = full retirement age. substantially higher measure of earnings.  16 Individual Replacement Rates Exhibits 5 through 7 focus on individual replace- Information, December 2016). Replacement rates value of price-adjusted shared earnings over a life- ment rates. using a number of other configurations are shown time. (When the sample is divided into five groups in Exhibits 6 and 7. that are ranked according to those earnings, a Exhibit 5 shows replacement rates using all earn- quintile is one of those five groups. A present value ings from age 22 through age 61 and the last five The replacement rates are generally shown for is a single number that expresses a flow of current years of substantial earnings—a measure that CBO long-career workers born in the 1960s, by sex and future income or payments in terms of an has published in the past (see CBO’s 2016 Long- and for the lowest and highest quintiles of shared equivalent lump sum received or paid at a specific Term Projections for Social Security: Additional lifetime earnings, which are defined as the present time.) INDIVIDUAL REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 5 . Median replacement rates based on the last five years Median Replacement Rates for Long-Career Workers, Using Initial Individual Benefits and of substantial price-adjusted earnings are significantly Two Different Measures of Earnings lower than median replacement rates that are based on the last 40 years of any price-adjusted earnings, Percent including years with no or very low earnings. (The Both Sexes, All Quintiles former replacement rates are generally used to evalu- ate the extent to which benefits can provide income 100 All Earnings From Age 22 Through Age 61, continuity from the last several years of work, Scheduled Last 5 Years Including Years With Benefits whereas the latter replacement rates can be used to Scheduled Benefits Payable of Substantial No or Very Low Earnings Payable analyze the extent to which benefits enable workers Earnings Benefits to maintain the average standard of living experi- Benefits 50 enced during working years.) 59 59 55 52 60 43 Initial replacement rates based on the last 5 years 38 38 36 34 39 28 of substantial earnings before age 62 are less than 0 40 percent for long-career workers of both sexes and 1940s 1960s 1980s across cohorts. Consequently, individual retired- worker benefits alone are not projected to provide income continuity from the late-career earnings. 100 Using a longer measure of earnings (all earnings from age 22 through age 61) but the same mea- sure of benefits (scheduled retired-worker benefits, 50 if initially claimed at age 65), replacement rates 80 75 50 47 are substantially higher—between 55 percent and 22 21 34 32 60 percent of earnings for long-career workers across 0 cohorts. However, those rates suggest that benefits Lowest Quintile Highest Quintile are still likely to fall short of maintaining the average Source: Congressional Budget Office. standard of living experienced during working years. (For the 1960s and 1980s cohorts, replacement rates The initial individual benefit (the numerator) is the retired worker’s benefit, if first claimed at age 65. Scheduled benefits are benefits based on payable benefits are lower than those based as calculated under the Social Security Act, regardless of the balances in the combined trust funds. Payable benefits are benefits as calculated under the act, reduced as necessary to ensure that outlays do not exceed the Social Security system’s revenues once the on scheduled benefits.) balances in the combined trust funds are exhausted, which is projected to occur in 2031. Because of the progressive nature of Social Security’s Two different measures of price-adjusted earnings (denominators) are used in this exhibit. To compute the last 5 years of substantial benefit formula, replacement rates are much higher earnings (annual earnings that are at least half of the worker’s average indexed earnings), CBO started with the last year of substantial for workers with lower earnings (see the bottom earnings before age 62, identified the 4 preceding years with such earnings, and averaged those values. The measure of all earnings panel in the exhibit). For example, the median from age 22 through age 61, including years with no or very low earnings, captures the worker’s full earnings history up to age 62. All replacement rate based on all earnings from age 22 amounts are before taxes. Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent through age 61 is 80 percent for workers born in of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with the 1960s whose lifetime earnings fall in the lowest fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. earnings quintile, more than double the 34 per- cent for workers whose earnings fall in the highest quintile.  18 INDIVIDUAL REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 6 . In general, median replacement rates increase con- Variations in Median Replacement Rates for Long-Career Workers Born in the 1960s, siderably as a result of two analytical choices: using a Using Different Measures of Earnings longer period for measuring earnings and adjusting those earnings for changes in prices rather than for Percent changes in wages. (In this exhibit, the numerator Both Sexes for all replacement rates is the initial retired-worker 100 benefit, if first claimed at age 65.) 80 Highest 35 Years of Earnings (W) 66 67 58 63 49 55 49 50 Last 5 Years of Substantial This exhibit focuses on the 1960s cohort as an 50 36 36 38 40 44 31 34 Earnings (P) example of the pattern observed across all cohorts. 24 22 23 23 26 Last 5 Years Above $1,000 (P) For that group, median replacement rates across all 0 Last 5 Years, Including Years quintiles increase substantially as the period for mea- 100 Men With No or Very Low Earnings (P) suring earnings expands from 5 years to 40 years. 75 Last 20 Years, Including Years Because workers tend to have at least some years 60 65 66 64 With No or Very Low Earnings (P) with no or very low earnings during their lifetimes, 50 46 51 46 51 34 36 38 40 41 28 expanding the period for measuring earnings cap- 19 18 19 19 21 25 Highest 35 Years of Earnings (P) tures more of those years, leading to lower average All Earnings From Age 22 earnings (and, therefore, a smaller denominator in 0 Through Age 61, Including Years the calculation of the replacement rates). 100 Women With No or Very Low Earnings (P) 84 71 When earnings are adjusted for changes in prices 52 59 60 66 52 49 56 rather than for changes in wages, the median 50 38 37 39 40 46 31 27 28 28 34 41 45 replacement rate for all workers based on the highest 35 years of earnings increases from 36 per- 0 cent to 49 percent. Because wages typically grow All Quintiles Lowest Quintile Highest Quintile faster than prices, adjusting earnings for changes in prices results in smaller measures of earnings (and, Source: Congressional Budget Office. therefore, a smaller denominator in the calculation The initial individual benefit (the numerator) is the retired worker’s benefit, if first claimed at age 65. of the replacement rates). For earnings measures (denominators), “P” indicates that the earnings have been adjusted for changes in prices, and “W” denotes that earnings have been adjusted for changes in wages. The last 5 (or last 20) years of earnings, including years with no or very low Of the options analyzed, the denominator based earnings, are computed starting with the last year with positive earnings before age 62, stepping back through 4 (or 19) adjacent on the last 20 years of price-adjusted earnings, years, and averaging those values, including years with no or very low earnings. (Starting at the last year with positive earnings better including years with no or very low earnings, results captures the final years of work by excluding years when some workers may have already retired.) Substantial earnings are annual in replacement rate values for both sexes and all earnings that are at least half of the worker’s individual average indexed earnings. All values are before taxes. quintiles of lifetime earnings that are roughly in the Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving middle of the range.  any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. 19 INDIVIDUAL REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 7 . Median replacement rates are substantially higher Variations in Median Replacement Rates for Long-Career Workers Born in the 1960s, for people who claim benefits later. For example, Using Different Hypothetical Claiming Ages calculated using a single measure of earnings—the last 20 years of any price-adjusted earnings, includ- Percent ing years with no or very low earnings—median 100 Both Sexes replacement rates for the 1960s cohort increase Benefits Claimed at Age 62 Benefits Claimed at Age 65 by nearly 30 percentage points as the claiming age 95 Benefits Claimed at FRA increases from age 62 to age 70. That increase is 50 76 Benefits Claimed at Age 70 observed for both male and female workers and 62 66 across all quintiles of lifetime earnings. 50 53 44 26 30 37 35 21 The findings shown in this exhibit are the result of 0 the actuarial reductions in monthly benefits when 100 Men they are claimed before the full retirement age (FRA) and increases in monthly benefits due to delayed 94 retirement credits when benefits are claimed after 50 76 the FRA. (For workers in the 1960s cohort, the FRA 59 66 is age 67. For those workers, benefits claimed at age 48 41 53 62, for example, are 30 percent lower than benefits 33 30 17 21 24 claimed at the FRA.) 0 100 Women Although delaying the claiming of benefits signifi- cantly increases replacement rates, that delay has 94 comparatively little effect on the value of average 50 76 lifetime benefits. When people claim benefits earlier, 66 66 53 53 they collect smaller annual benefits for more years. 46 48 When people claim later, they collect larger annual 37 34 39 27 0 benefits for fewer years.  All Quintiles Lowest Quintile Highest Quintile Source: Congressional Budget Office. For this exhibit, CBO computes replacement rates using four measures of benefits (numerators): initial individual benefits, if first claimed at age 62, 65, the FRA, or 70. The earnings measure (the denominator) is based on the last 20 years of any individual price-adjusted earnings, including years with no or very low earnings. All values are before taxes. Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. FRA = full retirement age. 20 Shared Replacement Rates To provide a more comprehensive view of the (the individual worker) the same as that used in separately. To account for the generally lower Social Security benefits available to retired work- previous exhibits, thus ensuring that the individ- per-person living expenses in households with two ers, Exhibits 8 through 11 show replacement rates ual and shared replacement rates are comparable. or more members, CBO adjusts for those econo- that take into account the benefits and earnings of (As a result, the earnings and benefits of a spouse mies of scale. Shared benefits and shared earnings the worker’s spouse. In this report, shared bene- who is not a long-career worker are included when are computed as the sum of the worker’s and his fits at a hypothetical claiming age of 65 reflect all analyzing the shared earnings and benefits of the or her spouse’s benefits and earnings in each year retired-worker, spousal, and survivor’s benefits that long-career worker, but the spouse who has not during the marriage divided by the square root of both the worker and his or her spouse are eligible worked or worked for fewer than 20 years is not two, implying that a married person would need to receive on the basis of either of their earnings included in the analysis on his or her own.) about 30 percent less income than a single person histories. Similarly, shared earnings incorporate living alone to maintain the same general standard earnings of both spouses during the marriage. Because household members tend to share of living. expenses for housing, utilities, and other items, Shared replacement rates are computed for each two spouses typically need less than twice the long-career worker, keeping the unit of analysis income that either person would need if living SHARED REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 8 . Overall, shared replacement rates are similar to Comparison Between Median Replacement Rates for Long-Career Workers Born in the 1960s, individual replacement rates, and the variation in Using Individual and Shared Measures and Different Measures of Earnings replacement rates using different denominators is similar for the individual and shared measures. Percent (Shared replacement rates capture not only the 80 Both Sexes retired worker’s own benefits and earnings, but also Shared any spousal and survivor’s benefits and benefits and Individual earnings of the worker’s spouse.) However, the rela- 40 55 55 tionship between individual and shared replacement 39 44 45 49 49 36 36 36 rates is different for men and women. 0 For female workers born in the 1960s, shared 80 Men replacement rates tend to be lower than individual replacement rates, whereas for male workers, they 40 tend to be higher than individual rates. Because 45 46 49 51 55 women generally have smaller benefits and lower 36 39 41 34 36 earnings than men and because Social Security’s ben- 0 efit formula is progressive, incorporating the larger 80 Women benefits and earnings of their (male) spouses yields lower overall replacement rates for female workers. For male workers, incorporating the smaller benefits 40 and earnings of their (female) spouses leads to higher 45 52 49 59 54 overall replacement rates. 37 38 46 38 36 0 Last 5 Years of Last 20 Years, Including Highest 35 Years of Highest 35 Years of All Earnings From Shared replacement rates for male and female Substantial Earnings (P) Years With No or Earnings (W) Earnings (P) Age 22 Through Age 61 (P) Very Low Earnings (P) workers are much closer together than their individ- ual replacement rates. For two spouses, the shared Late-Career Earnings Lifetime Earnings benefit measures are equal, and the shared earnings Measures Measures measures are generally similar. (However, the shared Source: Congressional Budget Office. earnings measures for two spouses typically are not For this exhibit, CBO computes replacement rates using two measures of benefits (numerators): initial individual benefits and initial equal because of the individual years of earnings that shared benefits. The initial individual benefit is the retired worker’s benefit, if first claimed at age 65. The initial shared benefit reflects each spouse has before or after marriage.)  the sum of the worker’s and spouse’s initial benefits, if claimed age 65, adjusted for economies of scale. The earnings measures (denominators) are similarly computed either on an individual or a shared basis. CBO computes earnings measures in five different ways, ranging from the last 5 years of substantial earnings (that is, annual earnings that are at least half of the worker’s average indexed earnings) to all earnings from age 22 through age 61, including years with no or very low earnings. “P” indicates that the earnings have been adjusted for changes in prices, and “W” denotes that the earnings have been adjusted for changes in wages. All values are before taxes. Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. 22 SHARED REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 9 . Married workers have higher shared replacement Median Replacement Rates for Long-Career Workers Born in the 1960s, With Initial Shared rates than do workers who are divorced or widowed Benefits and the Last 20 Years of Shared Earnings, by Household Type or who are single. (For comparison with some other exhibits, the denominators of the replacement rates Percent are based on the last 20 years of shared earnings, Both Sexes including years with no or very low earnings.) For 60 example, for married workers born in the 1960s, the median shared replacement rate is 47 percent, whereas the median replacement rate for divorced workers is 41 percent. In CBO’s analysis, the shared 30 benefits of married workers are higher than their 45 45 47 45 individual retired-worker benefits because eligible 41 spouses are generally entitled to total individual benefits that are 50 percent of the retired worker’s 0 amount and because of the economies-of-scale All Single Married Divorced Widowed adjustment. In households consisting of a married couple, 60 replacement rates vary on the basis of whether one or both spouses worked. CBO defined one-earner households as those in which only one spouse is 30 57 eligible for OAI benefits on the basis of his or her 46 own earnings by age 62. Two-earner households are defined as those in which both spouses are 0 eligible for OAI worker benefits. (Among mar- One-Earner Two-Earner ried long-career workers born in the 1960s, fewer Household Household than 10 percent belong to one-earner households.) Because of the availability of the spousal benefit, Source: Congressional Budget Office. generally the median shared replacement rate in a For this exhibit, the marital status of a long-career worker at age 61 was used to determine his or her household type. One-earner one-earner household is significantly higher than the households are defined as households in which only one spouse is eligible for Old-Age Insurance benefits using earnings through age median replacement rate in a two-earner household. 61, whereas two-earner households are those in which two spouses are eligible. (The patterns in the replacement rates by marital The initial shared benefit (the numerator) reflects the sum of the worker’s and spouse’s initial benefits, if claimed age 65, adjusted for status and by the number of earners in the house- economies of scale. The earnings measure (the denominator) is based on the last 20 years of shared price-adjusted earnings, including hold are similar by sex.)  years with no or very low earnings. All values are before taxes. Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. 23 SHARED REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 10 . Median shared replacement rates for long-career The Relationship Between Initial Household Benefits, Poverty Thresholds, and Shared workers vary considerably depending on whether Replacement Rates for Long-Career Workers Born in the 1960s the worker’s shared benefit falls above or below the federal poverty threshold. For the vast majority of long-career workers, the sum of the worker’s initial Percentage of Long-Career Workers benefit and the spouse’s benefit—the household Above Poverty Threshold Below Poverty Threshold benefit—exceeds the poverty threshold. For those Household Benefits Above Versus 95 5 workers, the median shared replacement rates Below the Poverty Threshold range from 36 percent, using the highest 35 years 0 50 100 of wage-adjusted earnings, to 54 percent, using all earnings from age 22 through age 61, adjusted for Median Replacement Rates (Percent) changes in prices over time. Highest 35 Years of Earnings (W) For the 5 percent of long-career workers whose 36 56 initial household benefits are below the poverty threshold, median shared replacement rates are Last 5 Years of Substantial Earnings (P) 38 52 substantially higher than for workers whose benefits are above the threshold. Those replacement rates Last 20 Years, Including Years 45 73 are higher because workers with benefits below the With No or Very Low Earnings (P) threshold have lower earnings and because the bene- Highest 35 Years of Earnings (P) 49 77 fit formula is progressive. All Earnings From 54 91 The divergence between the two groups of workers Age 22 Through Age 61 (P) is particularly large when replacement rates are based 0 50 100 on longer periods of price-adjusted earnings. For example, whereas replacement rates based on the last Source: Congressional Budget Office. five years of substantial earnings are about 14 per- For comparison with the official federal poverty threshold, CBO computes household benefits as the sum of the worker’s and spouse’s centage points higher for workers whose household initial benefits, if claimed at age 65. If the worker is married, initial household benefits are compared with the poverty threshold for benefits fall below the poverty thresholds, their two people age 65 or older, adjusted for growth in prices. If the worker is not married, the initial household benefit (the retired-worker replacement rates based on all earnings from age benefit plus any spousal or survivor’s benefit that the worker is eligible to receive) is compared with the poverty threshold for one 22 through age 61 are about 37 percentage points person. higher. As the span lengthens, more years with no For replacement rates, CBO computes initial shared benefits as the sum of the worker’s and spouse’s initial benefits, if claimed at age or very low shared earnings tend to occur for such 65, adjusted for economies of scale. For the earnings measures, “P” indicates that the earnings have been adjusted for changes in workers. (The patterns in the replacement rates for prices, and “W” denotes that earnings have been adjusted for changes in wages. Substantial earnings are annual earnings that are at the two groups are similar for men and women.)  least half of the worker’s shared average indexed earnings. All values are before taxes. Benefit measures are computed for all people who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. 24 SHARED REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 11 . Because taxes are generally lower in retirement than Median Replacement Rates for Long-Career Workers Born in the 1960s, Using Initial Shared during working years, after-tax replacement rates Benefits and the Last 20 Years of Shared Earnings, Adjusted for Taxes are generally higher than before-tax replacement rates. To adjust for payroll taxes, CBO deducted the Percent Replacement Rates employee’s share of payroll taxes from each year of 100 Adjusted for Adjusted for Both preretirement earnings. To adjust for federal income Payroll Taxes Payroll Taxes and taxes, CBO deducted income taxes on earnings from Federal Income Taxes Before Taxes each year of preretirement earnings. Because Social 50 Security benefits are subject to taxation, the agency 73 79 77 45 49 51 also deducted income taxes on benefits from initial 25 27 30 benefits. (For comparison with previous exhibits, the 0 denominators of the replacement rates are based on Thousands of 2018 Dollars Numerators (Shared Benefits) the last 20 years of shared earnings, including years 100 with no or very low earnings.) For workers born in the 1960s, accounting for 50 payroll taxes increases the median shared replace- ment rate by 4 percentage points, from 45 percent 26 26 24 36 36 32 0 15 15 14 to 49 percent. Accounting for federal income taxes in addition to payroll taxes increases the replacement Thousands of 2018 Dollars Denominators (Shared Earnings) rate by another one and a half percentage points, resulting in an after-tax rate of 51 percent. 150 100 Because federal income taxes are progressive, mean- 142 ing that average tax rates generally rise as income 133 50 109 increases, workers in the highest earnings quintile 56 52 47 pay a larger share of their income in taxes than do 0 21 19 19 All Quintiles Lowest Quintile Highest Quintile workers in the lowest quintile. As a result, adjusting shared benefits for federal income taxes lowers the Source: Congressional Budget Office. median benefit for workers in the lowest quintile by less than 3 percent, from $14,900 to $14,500, while The before-tax replacement rate in each group is computed using before-tax amounts for the initial shared benefit (the numerator) the median benefit for workers in the highest quin- and the last 20 years of shared earnings (the denominator). Adjusting for payroll taxes deducts the employee’s share of taxes under the Federal Insurance Contributions Act from preretirement earnings, including payroll taxes for Old-Age and Survivors Insurance, tile is reduced by about 10 percent, from $36,000 to Disability Insurance, and Medicare. Adjusting for federal income taxes involves deducting income taxes on benefits as well as income $32,500. Similarly, for workers in the lowest quin- taxes on preretirement earnings. (See Appendix A for more information about how estimates of federal income taxes are obtained.) tile, adjusting for both payroll and federal income taxes lowers the shared earnings by about 8 percent, The earnings measure (the denominator) is based on the last 20 years of shared price-adjusted earnings, including years with no or very low earnings. from about $20,500 to $19,000. In contrast, for workers in the highest quintile, adjusting for taxes Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving lowers earnings by about 23 percent, from $141,500 any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of to $109,500 (see Appendix A).  earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. 25 Multiyear Shared Replacement Rates Exhibits 12 through 14 present results from benefits. For years when both the worker and the analysis, resulting in a sample that is 1 percent to CBO’s multiyear analysis. Unlike the previous spouse receive benefits, CBO applies an adjust- 2 percent smaller, on average and across cohorts, exhibits, which generally show hypothetical ment for household size and economies of scale, than the sample used in previous exhibits. (Within Social Security benefits at the age of 65, the fol- dividing the sum of their benefits by the square specific groups, sample sizes are between 0 and lowing exhibits summarize the projected benefits, root of two. In years when only one member of the 5 percent smaller than the ones used in the previ- adjusted for changes in prices over time, that are two-person household receives a benefit, there is ous exhibits.) received from the initial year of claiming until no adjustment for household size or economies of death. The multiyear benefit measure is a simple scale. (Doing otherwise could significantly under- The advantage of multiyear analysis is that it average of those benefits—that is, all years of state the resources available to the married couple allows an examination of the way benefit measures benefits are weighted equally. in such a year because the other spouse’s ongoing change over the course of retirement. For retired earnings would be omitted while his or her benefit workers who are married, a life event that often Multiyear shared benefits reflect all Old-Age would be zero because it would not have been results in significant changes to their financial and Survivors Insurance benefits received by the claimed yet. See Appendix B for the sensitivity of circumstances is the death of a spouse, which trig- worker and his or her spouse from the time the findings to alternative specifications.) gers an adjustment of the Social Security benefits. worker is age 62 until the end of life, including Exhibit 13 illustrates the magnitude of a change in spousal benefits and survivor’s benefits, if any For workers who are projected to die before they shared replacement rates for married workers who (see Exhibit 12). The first year of benefits for the or their spouses claim any benefits, multiyear become widowed in retirement. multiyear analysis is defined as the first year after replacement rates equal zero, complicating inter- the worker reaches age 62 in which either the pretation of the results. To address that concern, worker or the spouse is projected to receive any such workers are excluded from the following MULTIYEAR SHARED REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 12 . Multiyear replacement rates—computed on the Median Multiyear Replacement Rates for Long-Career Workers, Using the Average of basis of all shared scheduled benefits received from All Shared Benefits and the Last 20 Years of Shared Earnings age 62 until death—are roughly comparable with the initial replacement rates calculated under the Percent assumption that benefits are claimed at age 65. (That 80 similarity holds across and within quintiles of shared Multiyear lifetime earnings.) Initial Initial Scheduled Payable 60 Scheduled Benefits Multiyear Benefits Benefits Using the last 20 years of earnings as the denomina- Payable Benefits tor, the multiyear shared replacement rate based on scheduled benefits is 1 percentage point higher than 40 the initial shared replacement rate with scheduled 51 benefits for workers born in the 1960s. However, the 44 48 48 46 45 42 48 20 42 38 multiyear shared replacement rate based on payable 36 35 benefits is substantially lower than the initial replace- ment rate because benefit reductions following the 0 1940s 1960s 1980s projected depletion of the Social Security trust funds would affect multiyear benefits for nearly all workers 80 born in the 1960s, whereas they would affect initial 60 benefits for only a fraction of those workers. (The combined trust funds are projected to be depleted in 40 73 73 68 2031.) 58 20 27 21 26 24 Scheduled multiyear replacement rates increase from 0 44 percent for workers born in the 1940s to 51 per- Lowest Quintile Highest Quintile cent for workers born in the 1980s. That finding is, Source: Congressional Budget Office. in part, a result of an increase in earnings inequality over time. In contrast, multiyear replacement rates In this exhibit, CBO shows replacement rates under different measures of shared benefits (numerators). The multiyear benefits are using payable benefits are lower for the later cohorts constructed using the average of all shared benefits received over the course of retirement, from age 62 until death, adjusted for because of benefit reductions following the projected changes in prices. Initial benefits are benefits received at age 65, if first claimed at that age. The earnings measure (the denominator) is based on the last 20 years of shared price-adjusted earnings, including years with no or very low earnings. All values are before depletion of the combined Social Security trust taxes. funds.  Scheduled benefits are benefits as calculated under the Social Security Act, regardless of the balances in the trust funds. Payable benefits are benefits as calculated under the act, reduced as necessary to ensure that outlays do not exceed the Social Security system’s revenues once the balances in the combined trust funds are exhausted, which is projected to occur in 2031. Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. In addition, the sample shown here for both initial and multiyear replacement rates is restricted to workers who are projected to receive benefits in at least one year between age 62 and death; it is, therefore, 1 percent to 2 percent smaller, on average and across cohorts, than the one used in previous exhibits. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. 27 MULTIYEAR SHARED REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 13 . Median shared replacement rates in widowhood are Median Replacement Rates Before and After Widowhood for Long-Career Workers substantially lower than replacement rates during Born in the 1960s, Using Shared Benefits and the Last 20 Years of Shared Earnings marriage. For example, for married workers born in the 1960s who become widowed in retirement, Percent the median replacement rate falls by 10 percentage Both Sexes points. (The replacement rates before the loss of the 100 spouse are computed using the shared benefit in Before the last full year of marriage, when both spouses are Widowhood After Widowhood receiving benefits. The subsequent replacement rates 50 87 are computed for the same group of workers on the 70 basis of the benefit in the first full year of widowhood. 53 43 31 26 The last 20 years of shared earnings, including years 0 with no or very low earnings, serve as the denomina- 100 Men tor for both sets of replacement rates so that only the numerator varies as a result of widowhood.) The reduction in replacement rates in widowhood is 50 85 71 largely attributable to the loss of economies of scale 51 43 within the household. Because a newly widowed 29 25 person can no longer share common household 0 expenses, that person’s cost of living increases in 100 Women widowhood, reducing the amount of goods and services that can be afforded using the recalculated benefit. (The economies-of-scale adjustment used in the calculation of replacement rates implies that 50 90 a married person would need about 30 percent less 70 55 income than a person living alone to maintain the 44 33 26 same general standard of living.) 0 All Quintiles Lowest Quintile Highest Quintile Although female workers have higher replacement rates than men before widowhood, they experience Source: Congressional Budget Office. a larger reduction after widowhood, ending up Before-widowhood replacement rates are based on the shared benefits received by both spouses in the last full year before with rates that are very similar to those of men. The widowhood. After-widowhood replacement rates are based on the benefit received by the surviving spouse in the first full year of median replacement rate for women is 11 percentage widowhood. (In this report, widowhood refers to the state of being a widow or widower.) points lower in widowhood, whereas the median The denominator of the replacement rate is the same in both before- and after-widowhood replacement rates and reflects average replacement rate for men is 8 percentage points lower. shared price-adjusted earnings in the last 20 years, including years with no or very low earnings. All values are before taxes. Because of the structure of spousal and survivor’s ben- Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62, who are not receiving any efits, the reduction is generally greater for households benefit at age 61, who survive long enough to receive benefits for at least one year, and who become widowed after the age of 62. with two earners compared with households with one Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit earner, and female workers are more likely to have a the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In working spouse.  addition, workers who receive Disability Insurance benefits are excluded. 28 MULTIYEAR SHARED REPLACEMENT RATES Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Exhibit 14 . Although long-term projections of Social Security Uncertainty Surrounding Shared Multiyear Replacement Rates for Long-Career Workers replacement rates are inherently uncertain, changing Born in the 1960s key economic assumptions does not have a large effect on the findings shown in this report. Specifically, Percent CBO considered two alternative projections of the Both Sexes, All Quintiles labor force participation rate—the percentage of Labor Force Participation Productivity Growth people in the civilian noninstitutionalized population 50 50 who are age 16 or older and either working or actively seeking work—and of the growth rate of productivity. (Productivity growth is calculated as the growth of 45 45 total factor productivity, which is the growth of real 46.3 47.5 46.1 45.8 46.1 output that is not explained by growth in labor or 44.7 capital.) Those alternative projections reflect values 0 0 that are 3 percentage points (for the labor force Men, Lowest Quintile participation rate) and 0.5 percentage points (for the Labor Force Participation Productivity Growth growth rate of productivity) higher or lower than 75 75 what currently underlies CBO’s extended baseline. The labor force participation rate and the growth rate of productivity affect the projected earnings that 70 70 73.0 ultimately determine benefits and replacement rates. 72.0 70.7 70.7 70.1 68.8 For workers born in the 1960s, alternative projec- 0 0 tions of the labor force participation rate result in Low Extended Baseline High Low Extended Baseline High multiyear replacement rates that are less than 0.5 per- Source: Congressional Budget Office. centage point higher or lower than is projected based In CBO’s 2018 extended baseline projection, which generally reflects current law, the labor force participation rate is 59.5 percent on CBO’s extended baseline. Alternative projections in 2048, and the growth rate of productivity is about 1.2 percent annually over the next 30 years. In the alternative scenarios, the of the growth rate of productivity result in replace- labor force participation rate in 2048 is projected to be, on average, about 3 percentage points higher or lower than is projected in ment rates that are 1 to 2 percentage points higher CBO’s extended baseline; the growth rate of productivity is projected to be about 0.5 percentage points higher or lower than in the or lower than is projected based on the extended extended baseline. (The labor force participation rate is the percentage of the civilian noninstitutionalized population who are age baseline. For specific groups of workers, the differ- 16 or older and either working or actively seeking work. Productivity growth is calculated as the growth of total factor productivity, which is the growth of real, or inflation-adjusted, output that is not explained by growth in labor or capital.) For more information ence attributable to alternative projections is some- on how CBO constructed those alternative projections, see Congressional Budget Office, The 2016 Long-Term Budget Outlook what larger. For men in the lowest earnings quintile, (July 2016), Chapter 7, www.cbo.gov/publication/51580. for instance, median multiyear replacement rates Multiyear replacement rates are constructed using the average of all shared benefits received over the course of retirement, from age would change by slightly over 1 percentage point if 62 until death, and average shared price-adjusted earnings in the last 20 years, including years with no or very low earnings. All values labor force participation was lower and by slightly are before taxes. over 2 percentage points if productivity growth was Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving higher. (Although alternative economic projections any benefit at age 61. In addition, the sample is restricted to workers who are projected to receive benefits in at least one year do not have a large effect on the findings, other between age 62 and death. It is, therefore, 1 percent to 2 percent smaller, on average and across cohorts, than the one used in the sources of uncertainty, particularly behavioral factors analysis based on initial benefits. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than and alternative methodological approaches, may have 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. considerably larger effects. See Appendix B.)  29 APPENDI X A CBO’s Analytical Approach The Congressional Budget Office’s projections How CBO Projects Social Security work histories, which is necessary for the calcula- of Social Security replacement rates and other Benefits and Preretirement Earnings tion of shared benefits. benefit measures are based on its projections of a To develop its projections, CBO uses its long-term host of demographic and economic variables. (The microsimulation model.2 The model starts with Although the long-term model projects earnings projections underlying this report—including data on a 1 in 1,000 sample of the U.S. popu- and benefits, it does not currently project pen- projections of labor force participation, wage lation and projects demographic and economic sion income, balances in retirement accounts, or growth, and mortality—are the same as those outcomes for that sample through time. For each other assets that retirees may have. As a result, this underlying The 2018 Long-Term Budget Outlook.)1 person in the sample, the model simulates birth, report presents measures of Social Security benefits In this appendix, CBO provides more detail death, immigration and emigration, marital status rather than measures of overall retirement income about its analytical approach and the sources of and changes to that status, fertility, labor force and assets. data that are particularly important to projections participation, hours worked, and earnings from of Social Security benefit measures. In addition, labor. The resulting individual earnings histories Sources of Data the appendix describes how CBO’s analytical are then used to determine the retired-worker (or For each person in the starting sample, CBO uses choices and the resulting findings compare with disabled-worker) benefits for which an individual data from the Continuous Work History Sample those of the Office of the Chief Actuary at the is eligible.3 Because the model keeps track of links (CWHS), an administrative data set provided by Social Security Administration, which has been between individuals and their spouses (current, the Social Security Administration. Covering the producing measures of replacement rates for former, and deceased), it also computes the period from 1951 to 2013, those data contain almost 30 years. dependent benefits that the individual or his or her information on individual earnings and Social spouse may receive on the basis of either of their Security benefit amounts, the types of benefits (for example, retired-worker or spousal benefits), and 2. See Congressional Budget Office, An Overview of the dates benefits were claimed. For years beyond CBOLT: The Congressional Budget Office Long-Term Model 2013, CBO projects those variables on the basis of (April 2018), www.cbo.gov/publication/53667. historical administrative data (which are collected 3. The long-term model projects labor earnings both above by government agencies) and publicly available 1. For a detailed description of demographic and economic and below the taxable maximum—that is, the maximum survey data. projections underlying the analysis in this report, see amount of annual earnings to which the payroll tax Congressional Budget Office, The 2018 Long-Term is applied. However, only earnings up to the taxable Budget Outlook (June 2018), Appendix A, www.cbo.gov/ maximum are used to determine the Social Security Because data from the CWHS lack certain publication/53919. benefits. In 2018, the taxable maximum was $128,400. demographic components, such as information APPENDIX A Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis on spouses and children, CBO uses public surveys are projected to have at least 20 years of earnings Uncertainty in Microsimulation Models to impute, or fill in, that information for each above 10 percent of the average wage index (AWI) Long-term projections of individual earnings and individual in the sample. For instance, to impute in each year. To focus on retired workers, workers benefits are inherently uncertain. The findings marriage formation and dissolution, CBO uses who receive Disability Insurance (DI) benefits at shown in this report are affected by many sources data from the Survey of Income and Program age 61 or at any time afterward are excluded. of uncertainty, including CBO’s projections of key Participation (SIPP) that are matched to admin- economic, demographic, and behavioral factors. istrative earnings records. To project individual As an illustration, CBO shows how much a earnings beyond 2013, CBO uses data from the representative sample of the 1960s birth cohort is Because the agency uses a microsimulation administrative records together with the Current reduced because of restrictions used in this report model—that is, a model that projects outcomes Population Survey (CPS).4 Information from the (see Table A-1). First, including only workers who for individual people through time—another public data sources is not merged directly into the survive through age 62 reduces the starting sample important source of uncertainty that could affect individual records in the long-term model. Rather, by about 15 percent. Second, requiring that work- the findings is known as Monte Carlo variation.5 CBO uses those sources to estimate statistical rela- ers be eligible for OAI benefits on the basis of their CBO’s long-term microsimulation model uses tionships on the basis of individual characteristics earnings through age 61 reduces the sample by an random numbers and probabilities to determine that already exist in its model. Data from the SIPP additional 21 percent. Third, limiting the analysis annual outcomes for people in the sample. For and CPS are also used to estimate people’s annual to workers with long careers reduces the sample by example, suppose that on the basis of statistical probabilities of experiencing demographic and 9 percent. Finally, excluding workers who meet the probabilities, a 25-year-old married woman has a economic changes, such as the birth of children first three criteria but receive DI benefits after age 15 percent probability of having a child in a year. and changes in labor force participation. 61 has a small effect on the sample, reducing it by To determine whether a specific 25-year-old mar- 2 percent. Overall, the sample used in this report ried woman in the model will be assigned a child, Sample Selection is representative of slightly more than half of the a random number between zero and 1 is drawn for This report presents Social Security benefit mea- 1960s birth cohort. her, and a child is assigned if the random number sures for a sample of workers with long careers is lower than 0.15. Because the model is dynamic, who are projected to survive long enough to be Because of the restrictions applied to the sample, having a child assigned at age 25 affects not only eligible for retired-worker benefits. Specifically, the the workers that CBO analyzed differ somewhat the likelihood of children at each subsequent age sample includes only workers who are projected from OAI-eligible workers in the overall popula- but also future earnings and other outcomes for to survive through age 62, who are eligible to tion. For example, for long-career workers born that person, as well as her spouse. Consequently, receive Old-Age Insurance (OAI) benefits using in the 1960s, the highest 35 years of earnings the effect of random numbers propagates and their individual earnings through age 61, and who are about 15 percent higher than the average for becomes amplified in the model over time. all OAI-eligible workers. In addition, men are 4. For more information on CBO’s projection of earnings, marginally more likely to have at least 20 years 5. For more information on Monte Carlo variation, see see Jonathan A. Schwabish and Julie H. Topoleski, of significant earnings than women, with men Michael Simpson, “Investigating Monte Carlo Variation Modeling Individual Earnings in CBO’s Long-Term accounting for about 52 percent of workers in the in a Dynamic Microsimulation Model” (presentation Microsimulation Model, Working Paper 2013-04 to the Fifth World Congress of the International long-career sample compared with about 50 per- (Congressional Budget Office, June 2013), www.cbo.gov/ Microsimulation Association, September 2, 2015), publication/44306. cent of all OAI-eligible workers. www.cbo.gov/publication/50736. 31 APPENDIX A Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Table A-1 . Monte Carlo variation, it mitigates the variation’s Illustration of Restrictions Used to Produce the Sample for the 1960s Birth Cohort effect on the findings shown in this report. Plus the Plus the Long- Plus the Adjusting for Taxes With the Survival OAI-Eligibility Career Worker DI Worker CBO computes measures of after-tax replacement Starting Sample Restriction Restriction Restriction Restriction rates by adjusting both for payroll taxes on earn- Number of Workers (Thousands) ings and for income taxes on earnings and bene- Both Sexes 50 42 32 27 26 fits. To account for payroll taxes—taxes designated Men 25 21 16 14 14 for Social Security and Medicare—CBO deducts Women 24 22 16 13 13 the employee’s share of payroll taxes from each year of preretirement earnings. For Social Security, Percentage of the Starting Sample that share is 6.2 percent of earnings applied up to Both Sexes 100 85 64 55 53 a maximum amount of $128,400 in 2018. For Men 100 82 63 56 54 Medicare, that share is 1.45 percent of earnings Women 100 89 65 54 52 applied up to $200,000 ($250,000 for two-person Source: Congressional Budget Office. households) plus an additional 0.9 percent applied to earnings in excess of $200,000 ($250,000). The starting sample includes all people in CBO’s long-term model. “With the Survival Restriction” represents how the sample is reduced if workers are excluded because they did not live through age 62. The “OAI-Eligibility Restriction” shows how the sample is affected if workers are excluded because they were not eligible for OAI benefits on the basis of their earnings through age 61 or because To calculate income taxes on earnings, CBO they had already received benefits at age 61. The “Long-Career Worker Restriction” illustrates how the sample is further reduced applies the federal tax law in place in each year a by excluding workers who had fewer than 20 years of earnings above 10 percent of the average wage index. Finally, the “DI Worker worker had earnings.6 Those calculations are based Restriction” shows how the sample is reduced when workers who receive DI benefits after age 61 are excluded. on the demographic characteristics and earnings DI = Disability Insurance; OAI = Old-Age Insurance. of the worker and his or her spouse. Because the earnings records that form the basis of this analysis do not contain all the information necessary to estimate income taxes, CBO uses several simpli- Monte Carlo variation is the change in outcomes variation because of large sample sizes and the fications. To compute income taxes on preretire- that arises from the assignment of random num- relatively short period for which outcomes need to ment earnings, CBO includes only income from bers. Each time a model is run using different be projected for workers in the 1960s cohort, the labor earnings for each family. In addition, CBO random numbers, it produces slightly different estimates that are based on smaller samples show applies the same percentage of tax-deductible outcomes. The smaller the sample size for which some variation between model runs. Therefore, to an outcome is observed, the larger the variation obtain more reliable estimates for smaller groups, 6. Jon Bakija of Williams College created the federal tax in those outcomes between different runs of the CBO follows a common technique of running calculator that CBO used to model taxes in years prior model. multiple simulations that are based on different to 1986. For more details, see Jon Bakija, Documentation random numbers and averaging the values from for a Comprehensive Historical U.S. Federal and State Although most of the estimates shown in this those simulations. Although that technique does Income Tax Calculator Program, Working Paper (Williams College, August 2009), http://tinyurl.com/bakija report are not significantly affected by Monte Carlo not eliminate the uncertainty that is attributable to (PDF, 486 KB). 32 APPENDIX A Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis spending on state and local taxes, charitable con- To test the sensitivity of the findings to those besides Social Security benefits, the resulting tributions, and mortgage interest to all families. methodological choices, CBO computes federal after-tax replacement rate is 5 percentage points (Previously, CBO estimated that percentage to income taxes under two alternative specifications: higher than when those factors are included (see be 19 percent of earnings, using the most recent excluding the effect of children on taxes and Figure A-1). year of available data from the Internal Revenue excluding non-Social Security income sources Service’s Statistics of Income, 2014.) when calculating income taxes in retirement.9 How CBO’s Analytical Approach The median after-tax replacement rates using Compares With That of the Office of CBO also accounts for common deductions and those specifications are somewhat higher than the Chief Actuary at the Social Security credits that workers are projected to receive both the findings shown in this report. For example, Administration before and after retirement. For example, people for long-career workers born in the 1960s, a The Office of the Chief Actuary (OCACT) at the in eligible households with children would receive median after-tax replacement rate that excludes Social Security Administration has been projecting the earned income tax credit while working, and the effect of children on taxes is 1 percentage Social Security replacement rates for hypothetical elderly tax filers would claim an additional elderly point higher than the after-tax rate that includes workers since 1989.10 Although the OCACT’s standard deduction. the effect of children. When federal income taxes analytical method differs from the approach CBO are computed without an adjustment for chil- used for this report, the OCACT’s and CBO’s Up to 85 percent of a recipient’s Social Security dren and without including retirement income replacement rates would be similar if computed benefits are subject to income tax.7 However, the for workers with similar average earnings and CWHS data do not contain all the information taxation of benefits in this report are broadly consistent using the same specifications of numerators necessary to compute income taxes on benefits, with the findings in that publication. (Because the (benefits) and denominators (earnings). which are based on total income in retirement. findings in that publication are primarily based on cross- Therefore, CBO imputes unearned income from sectional data, whereas the analysis in this report is based The OCACT produces replacement rates for pensions and other sources before applying federal on panel data, the two sets of results are not directly five types of hypothetical workers characterized comparable.) tax law to total income in each year that benefits by their career-average earnings: very low, low, are received. (With those imputations, estimated 9. Accounting for children could have a substantial effect medium, high, and earnings at the taxable maxi- taxes on Social Security benefits approximate on both before-tax and after-tax replacement rates. mum level.11 (Career-average earnings are defined For simplicity, in this report, CBO does not include the amounts reported in data from the Internal children when applying the adjustment for household Revenue Service.)8 size to preretirement earnings and does not include child 10. For the OCACT’s most recent projected replacement rates, see Social Security Administration, Office of the benefits when computing household-level benefits. If, in Chief Actuary, “Replacement Rates for Hypothetical the computation of preretirement earnings, household 7. For more information on income taxes on benefits, see Retired Workers,” Actuarial Note 2018.9 (June 2018), size was adjusted for children, the resulting after-tax Congressional Budget Office, “The Taxation of Social www.ssa.gov/OACT/NOTES/ran9/an2018-9.pdf replacement rates based on initial shared benefits and Security Benefits,” CBO Blog (February 12, 2015), (137 KB). the last 20 years of shared earnings would be about www.cbo.gov/publication/49948. 5 percentage points higher than the rates that account 11. The taxable maximum is the maximum amount of annual 8. To impute pensions and other income sources, CBO for children in the computation of taxes but not in the earnings to which the payroll tax is applied ($128,400 in used 2015 data from the Internal Revenue Service’s computation of earnings. For a brief discussion of the 2018). The taxable maximum increases annually with Statistics of Income, analogous to the data used in effect of children on replacement rates, see Congressional average earnings; in years without a cost-of-living Congressional Budget Office, “The Taxation of Social Budget Office, Measuring the Adequacy of Retirement adjustment (as in 2010, 2011, and 2016), the taxable Security Benefits,” CBO Blog (February 12, 2015), Income: A Primer (October 2017), www.cbo.gov/ maximum does not increase. The taxable maximum does www.cbo.gov/publication/49948. The patterns in the publication/53191. not decrease when average wages decline. 33 APPENDIX A Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Figure A-1 . as the average of the highest 35 years of earnings, earnings, CBO then computes median values for Sensitivity of Shared After-Tax Replacement adjusted for growth in average wages to the year replacement rates for the sample as a whole and by before the worker retires.) The OCACT constructs the quintile of shared lifetime earnings to reflect Rates for Long-Career Workers Born in the hypothetical earnings histories using scaled factors the distribution of replacement rates. (Shared 1960s to Alternative Specifications of that are applied to the average wage index in each lifetime earnings are defined as the present value of Federal Income Taxes year. Those scaled factors are based on the average inflation-adjusted shared earnings over a lifetime. A Percent work and earnings of a sample of actual insured quintile is a division of the sample into one of five 60 workers. The final earnings patterns are adjusted groups ranked according to those earnings.) Before Taxes After Taxes so that the career-average earnings levels for the 50 hypothetical workers are set at 25 percent, 45 per- In this appendix, CBO illustrates how the earn- 40 cent, 100 percent, and 160 percent of the AWI ings of the OCACT’s hypothetical workers and for the hypothetical workers with very low, low, the sample of long-career workers analyzed in this 30 medium, and high earnings, respectively. (For the report compare with the AWI (see Figure A-2). 47 51 45 46 20 hypothetical worker with earnings at the taxable To make the comparison, CBO computes a maximum level, career-average earnings are about measure of the highest 35 years of wage-indexed 10 250 percent of the average wage index.)12 earnings—a measure similar to the OCACT’s 0 career-average earnings—for workers born in the With Adjustment Removing Plus Removing In this report, CBO shows projected median 1960s, showing the median findings within each for Children and Adjustment Non-Social Non-Social Security for Children Security Social Security replacement rates for a sample of quintile of shared lifetime earnings.14 (The quin- Retirement Income Retirement Income long-career workers. CBO begins its analysis with a tiles used in Figure A-2 are the same as those used sample of real, rather than hypothetical, U.S. work- throughout the report.) Source: Congressional Budget Office. ers, uses the earnings histories for those workers For this figure, replacement rates were computed using initial until 2013, and simulates their earnings for years shared benefits if claimed at age 65 and the last 20 years of shared price-adjusted earnings, including years with no or very low beyond 2013 on the basis of various demographic earnings. The sample used is the same as that used for Exhibit 11. characteristics and projected trends.13 Using those the basis of his or her work history and other individual The after-tax replacement rates shown here account for federal characteristics.) The advantage of holding the age of income taxes only. After-tax replacement rates that included a 12. For additional information about the OCACT’s scaled claiming constant at a hypothetical age is that it takes out tax adjustment for children and non-Social Security income were factors, see Social Security Administration, Office of the the effect of changes in average claiming patterns. See computed in the same way as in Exhibit 11: A tax adjustment for Chief Actuary, “Scaled Factors for Hypothetical Earnings Appendix B for a comparison of replacement rates that are children was made when taxes on preretirement earnings were Examples Under the 2018 Trustees Report Assumptions,” based on the hypothetical and projected ages of claiming. computed, and an imputation of other retirement income was Actuarial Note 2018.3 (June 2018), www.ssa.gov/OACT/ made when income taxes on benefits were computed. 14. Although CBO’s measure of career-average earnings NOTES/ran3/an2018-3.pdf (152 KB). comes close to the OCACT’s measure, several small Long-career workers are workers with 20 or more years of 13. Although CBO uses age 65 as the hypothetical age for differences remain. For example, the OCACT uses earnings above 10 percent of the average wage index in each claiming benefits in the single-year analysis, that choice earnings from age 21 through age 64 for its calculation, year. To limit the focus to individuals with significant attachment yields replacement rates that are similar to those based whereas CBO uses earnings from age 16 through age to the labor force, workers with fewer than 20 years of earnings on the projected age of claiming. (The projected age of 61. The OCACT indexes earnings to the year before are excluded. In addition, workers who receive Disability claiming is the age at which CBO’s long-term model retirement (the age of retirement is 65 in this example), Insurance benefits are excluded. projects a future beneficiary would first claim benefits on whereas CBO indexes earnings to age 65. 34 APPENDIX A Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Figure A-2 . The differences between the earnings levels of the Comparison of Earnings Levels for Long-Career Workers With OCACT’s Hypothetical hypothetical workers in the OCACT’s analysis and the median values within quintiles of lifetime Scaled Workers in the 1960s Cohort earnings in CBO’s analysis lead to differences OCACT’s Earnings Levels for Hypothetical Scaled Earners in estimated replacement rates. To illustrate those differences, CBO presents findings from Scaled Scaled Scaled a configuration of replacement rates that closely Very Low Low Medium Scaled Maximum Earnings Earnings Earnings High Earnings Earnings resembles a specification used by the OCACT (see Figure A-3). Specifically, for the numerator, CBO 0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 uses a hypothetical retired-worker benefit that is based on the assumption that benefits are claimed CBO’s Median Earnings Levels for Long-Career Workers at age 65; for the denominator, CBO uses the highest 35 years of wage-adjusted earnings. Lowest Second Third Fourth Highest Quintile Quintile Quintile Quintile Quintile Because the long-career workers in CBO’s analysis 0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 tend to have higher earnings than workers in the Percentage of the Average Wage Index overall population, CBO’s projections of median earnings in the bottom two quintiles of shared life- Source: Congressional Budget Office, using data from Social Security Administration, Office of the Chief Actuary. time earnings are higher than the OCACT’s earn- The OCACT’s earnings levels for hypothetical scaled workers are taken from Social Security Administration, Office of the Chief Actuary, ings levels for hypothetical workers with very low “Replacement Rates for Hypothetical Retired Workers,” Actuarial Note 2018.9 (June 2018), https://www.ssa.gov/OACT/NOTES/ran9/ and low earnings. Consequently, CBO’s median an2018-9.pdf (137 KB). Scaled workers are hypothetical workers whose annual earnings are obtained by multiplying the average wage replacement rates for the bottom two quintiles index by the scaled factors in corresponding years. Maximum earnings reflect earnings at the taxable maximum amount at each age. are lower than the OCACT’s replacement rates CBO’s median earnings levels for long-career workers reflect the highest 35 years of wage-adjusted individual earnings for workers for hypothetical workers with very low and low born in the 1960s and the average wage index for 2018. Long-career workers are workers with 20 or more years of earnings above earnings because of the progressive nature of the 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. Social Security benefit formula, which means that the ratio of benefits to earnings is lower for people OCACT = Office of the Chief Actuary of the Social Security Administration. with higher average earnings than for people with lower average earnings. CBO’s median earnings levels for the top two Overall, median earnings in CBO’s five quin- from approximately 40 percent of the AWI to quintiles of lifetime earnings are lower than the tiles of lifetime earnings reflect a tighter range of nearly 210 percent of the AWI. The OCACT’s OCACT’s earnings levels for hypothetical workers earnings levels. Nonetheless, that range is roughly scaled workers reflect a range from 25 percent of with high earnings and earnings at the taxable comparable with the range of earnings of the the AWI to about 250 percent of the AWI. maximum. Consequently, CBO’s median replace- OCACT’s five types of hypothetical workers. For ment rates are higher than the respective rates CBO, the median earnings of the groups range produced by the OCACT. (For the top quintile, 35 APPENDIX A Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Figure A-3 . CBO’s median replacement rate is 24.4 percent; Replacement Rates for OCACT’s Hypothetical Workers and CBO’s Long-Career Workers the OCACT’s replacement rate for a worker with maximum earnings is 23.7 percent.) The Born in the 1960s, Using Wage-Adjusted Earnings earnings for the middle quintile of shared life- Percent time earnings in CBO’s projections are similar to those of the OCACT’s hypothetical worker with 80 medium earnings, resulting in similar replacement rates. Despite the differences in the analytical 60 approaches, the OCACT’s and CBO’s replace- OCACT CBO ment rates are roughly comparable for similar 40 earnings levels and specifications of numerators 67 and denominators. 49 49 20 39 36 35 30 32 24 24 0 Very Low Earnings/ Low Earnings/ Medium Earnings/ High Earnings/ Maximum Earnings/ Lowest Quintile Second Quintile Third Quintile Fourth Quintile Highest Quintile Source: Congressional Budget Office, using data from Social Security Administration, Office of the Chief Actuary. OCACT replacement rates reflect benefits if claimed at age 65 and career-average wage-adjusted earnings for workers born in 1965. Those rates are taken from Social Security Administration, Office of the Chief Actuary, “Replacement Rates for Hypothetical Retired Workers,” Actuarial Note 2018.9 (June 2018), https://www.ssa.gov/OACT/NOTES/ran9/an2018-9.pdf (137 KB). CBO’s median replacement rates reflect initial benefits, if claimed at age 65, and the average of the highest 35 years of wage-adjusted earnings for workers born in the 1960s. CBO’s replacement rates were computed for all individuals who were eligible to claim retirement benefits at age 62 and who were not receiving any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. All values are before taxes. OCACT = Office of the Chief Actuary of the Social Security Administration. 36 APPENDI X B Sensitivity of Findings to Certain Analytical Choices In calculations of Social Security replacement • Using a traditional shared-earnings approach the official federal poverty threshold. (Additional rates, the findings are highly sensitive to the to account for household size rather than results based on different multiples of the thresh- specification of the numerator (benefits) and adjusting for economies of scale within the old are included in the supplemental data and the denominator (earnings), as demonstrated household. interactive tool that accompany this report.) in Exhibits 6 and 7 of this report. In addition, the measures of Social Security benefits that the In evaluating whether Social Security benefits Although the analysis that is based on 100 percent Congressional Budget Office developed for the enable beneficiaries to meet their basic needs in of the federal poverty threshold shows that the report are sensitive to certain choices the agency retirement, the findings are highly sensitive to the vast majority of long-career workers have initial made. In this appendix, CBO presents findings choice of the basic needs threshold. Initial shared worker benefits that exceed the poverty threshold using benefit measures that are based on four alter- benefit measures are also sensitive to the way in (as shown in Exhibit 1), increasing that threshold native analytical choices.1 Those include: which differences in spouses’ projected claiming to 200 percent reveals that most of those workers ages are treated. In contrast, multiyear shared ben- have benefits that are below the higher threshold • Using a higher threshold for meeting basic efit measures are considerably less sensitive to the (see Figure B-1).2 More than four times as many needs in retirement than 100 percent of the differences in spouses’ claiming ages. workers born in the 1940s have benefits that fall official federal poverty threshold, below 200 percent of the poverty threshold than Certain analytical choices do not have a significant 100 percent of the threshold. Among women, • Using a projected age for claiming retired- impact on the results. Those include whether to nearly 90 percent of workers born in the 1940s worker benefits in the computation of initial use benefit amounts at the hypothetical claiming have retired-worker benefits that fall below the replacement rates rather than a hypothetical age of 65 or at the projected claiming age in the higher threshold, whereas about 26 percent of age of 65, computation of retired-worker replacement rates those workers have benefits below the lower and whether to adjust the findings to reflect econ- threshold. • Using alternative methods to account for omies of scale in the computation of initial shared differences in spouses’ projected claiming ages, replacement rates. and Basic Needs Threshold To demonstrate the sensitivity of findings to alter- native measures of basic needs, CBO projected the 2. In 2018—the most recent year for which Census poverty 1. Additional alternative specifications are shown in the percentage of workers with initial retired-worker thresholds are available—the poverty threshold for one supplemental data posted on CBO’s website along with person age 65 or older was $12,043, and the threshold this report at www.cbo.gov/publication/55038. benefits that are below a specified multiple of for two people was $15,178. APPENDIX B Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Figure B-1 . Retired-Worker Benefits at the Percentage of Long-Career Workers With Initial Individual Benefits Below 200 Percent of the Hypothetical Claiming Age of 65 Versus Poverty Threshold the Projected Claiming Age The majority of the exhibits in this report show Percent Social Security benefit measures that are based Both Sexes on the assumption that benefits are first claimed 100 Scheduled Payable at the hypothetical age of 65. However, it is also Benefits Benefits informative to look at replacement rates that 50 are based on projected claiming patterns. (The 69 69 66 72 81 58 projected age for claiming benefits is the age at 0 which CBO’s long-term model projects a future beneficiary would first claim benefits on the basis 100 Men of his or her work history and other individual characteristics.) 50 76 53 53 56 63 54 Historically, a substantial share of eligible retired 0 workers has claimed benefits at one of two ages: the earliest eligibility age (EEA) or the full retire- 100 Women ment age (FRA). Although the EEA has remained constant at 62 (for people born in 1899 and later), 50 88 88 the FRA has been rising over time. Under current 78 83 87 64 law, for workers born before 1938, the FRA is 65. 0 For workers born between 1938 and 1943, the 1940s 1960s 1980s FRA increases by two months for each successive Birth Cohort birth year, until it reaches 66 for people born in Source: Congressional Budget Office. 1943. The FRA remains at 66 for workers born The basic needs threshold used in this figure is 200 percent of the official federal poverty threshold for one person age 65 or older, between 1943 and 1954, and then, starting with adjusted for growth in prices over time. people born in 1955, it increases by two months Initial individual benefits are based on the assumption that workers first claim benefits at age 65. Benefits are computed for all for each successive birth year, until it reaches 67 people who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. All benefit amounts for people born in or after 1960. are before taxes. Scheduled benefits are benefits as calculated under the Social Security Act, regardless of the balances in the trust funds. Payable benefits are benefits as calculated under the act, reduced as necessary to ensure that outlays do not exceed the About one-third of eligible workers born in the Social Security system’s revenues once the balances in the combined trust funds are exhausted, which is projected to occur in 2031. 1940s are estimated to have claimed benefits at the Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit EEA, and less than one-fifth are estimated to have the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In claimed benefits at the FRA. Historically, however, addition, workers who receive Disability Insurance benefits are excluded. labor force participation at older ages has been 38 APPENDIX B Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis increasing, and the average age at which benefits Table B-1 . are first claimed has been rising.3 For workers born Median Initial Replacement Rates for Long-Career Workers, Using Different Numerators and the in the 1980s, CBO projects that less than a quar- Last 20 Years of Earnings, Including Years With No or Very Low Earnings ter of eligible workers will claim benefits at the EEA, whereas about one-third will claim benefits Percent 10-Year Hypothetical Benefit at Hypothetical Benefit at Benefit at Projected Hypothetical Benefit at at the FRA. FRA Age 65 Claiming Age Age 62 Birth Cohort Scheduled Payable Scheduled Payable Scheduled Payable Scheduled Payable Consequently, for long-career workers born in the Both Sexes 1940s, replacement rates that are based on benefits 1940s 48 48 46 46 44 44 37 37 being claimed at a projected age (which averages 1960s 50 45 44 41 43 40 35 35 a bit less than 65) are slightly lower than the rates 1980s 54 39 47 34 48 37 38 27 based on benefits being claimed at the hypotheti- Men cal age of 65. For example, for workers born in the 1940s 46 46 43 43 41 41 35 35 1940s, median replacement rates based on the last 1960s 48 42 41 39 41 38 33 33 20 years of earnings and benefits at the projected 1980s 53 38 46 33 48 37 37 27 age of claiming are about 2 percentage points Women lower than the rates based on benefits claimed 1940s 51 51 49 49 48 48 39 39 1960s 53 48 46 44 45 42 37 37 at the hypothetical age of 65 (see Table B-1). In 1980s 55 40 48 34 47 37 39 28 contrast, with scheduled benefits, for workers born in the 1980s, the median replacement rate based Source: Congressional Budget Office. on a projected claiming age is slightly higher than For this table, individual replacement rates are computed under four alternative choices for individual benefits (numerators): the the replacement rate based on the hypothetical hypothetical benefit if claimed at the full retirement age (FRA), the hypothetical benefit if claimed at age 65, the benefit at the projected age of 65 but considerably lower than the replace- claiming age, and the hypothetical benefit if claimed at age 62. ment rate based on hypothetical claiming at the Scheduled benefits are benefits as calculated under the Social Security Act, regardless of the balances in the trust funds. Payable FRA. That finding is a result of workers in the benefits are benefits as calculated under the act, reduced as necessary to ensure that outlays do not exceed the Social Security 1980s cohort remaining in the labor force longer system’s revenues once the balances in the combined trust funds are exhausted, which is projected to occur in 2031. and claiming benefits at later ages than those in The earnings measure (the denominator) is based on the last 20 years of any individual price-adjusted earnings, including years with the 1940s cohort; however, because a substantial no or very low earnings. All values are before taxes. share of workers are expected to continue to claim Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving benefits before the FRA, projected replacement any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. 3. For more information, see Alicia H. Munnell and Anqi Chen, Trends in Social Security Claiming, Issue Brief 15-8 (Center for Retirement Research at Boston College, May 2015), http://crr.bc.edu/briefs/ trends-in-social-security-claiming/. 39 APPENDIX B Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis rates do not reach the hypothetical rates based on some cases, one person may claim retired-worker of claiming enters the multiyear benefit calcula- benefits being claimed at the FRA. benefits while the spouse does not and continues tion, the period during which both spouses receive to work or relies on other income sources that are benefits together is generally longer than the initial Overall, replacement rates based on benefits being not projected in CBO’s long-term model. When period when only one spouse receives benefits. claimed at the hypothetical age of 65 provide a the spouse ultimately claims benefits, the person reasonable alternative to replacement rates based who claimed first can apply for spousal benefits, if To show the sensitivity of findings to the treat- on the projected claiming age, while offering the eligible, resulting in that person’s getting a larger ment of the difference between spouses’ projected advantage of taking out the effects of changes in benefit than the retired-worker benefit he or she claiming ages, CBO has computed both initial average claiming patterns over time. received when benefits were first claimed. Thus, and multiyear shared replacement rates using the total benefit amount received by the worker alternative specifications. For initial rates based on Accounting for Differences in Spouses’ and his or her spouse can change substantially the projected claiming age, CBO has assessed the Projected Claiming Ages When after the first year in which benefits were received. effects of three alternative approaches: Computing Shared Benefit Measures Although shared benefits that are based on the For computing multiyear shared benefits in this ■ First, the agency has constructed initial shared hypothetical claiming age of 65 provide a simple report, CBO defines shared benefits at the pro- benefits using the first year in which both approximation for household-level benefits, they jected claiming age as benefits in the first year in the worker and his or her spouse receive a do not capture observed benefit-claiming patterns which either the worker or the spouse receives benefit, if married, rather than the first year among spouses. For the initial shared benefit benefits. Considering that the other member of a that either person receives a benefit. Because measures shown in this report, CBO summed two-person household often continues to work, that alternative specification captures both all retired-worker benefits as well as any spousal applying an adjustment for household size to one spouses’ benefits, it yields a significantly or survivor’s benefits that both the worker and person’s benefit could significantly understate the higher shared benefit at the projected claiming his or her spouse would be eligible to receive at resources available to the married couple in that age and, consequently, a higher replacement the hypothetical claiming age.4 However, many year as the other spouse’s ongoing earnings would rate—a 49 percent replacement rate compared married long-career workers do not claim ben- not be included. Therefore, in CBO’s calculations, with the 35 percent rate that is based on the efits in the same year that their spouses do. In an adjustment for household size (or economies specification of shared benefits used in this of scale) is not made when only one member of a report. (The denominators of the replacement 4. For example, for a married worker, the shared benefit at two-person household receives a benefit. rates are based on the last 20 years of shared the hypothetical age of 65 is the sum of retired-worker earnings, including years with no or very low benefits and any spousal benefits, if applicable, that Because initial replacement rates that are based earnings.) It is important to note, however, both the worker and his or her spouse could receive if each claimed at age 65 (adjusted for economies of scale on the projected age of claiming are sensitive to that the resulting benefit is not the benefit and household size). If the worker is older or younger the way in which the difference between spouses’ received in the first year of claiming but rather than the spouse, that measure would sum benefits from projected claiming ages is treated, the results the benefit received in the first year that both different calendar years, applying the appropriate price throughout this report are based on the hypotheti- spouses are projected to receive benefits. adjustment to keep benefits in constant dollars. The main cal claiming age of 65 for both spouses. In con- advantage of that approach is its conceptual simplicity. However, it is important to note that the resulting benefit trast, multiyear replacement rates are considerably ■ To incorporate years of individual benefits amount reflects the hypothetical shared benefit when less sensitive to alternative treatments: Although before both spouses receive benefits together, both the worker and the spouse receive benefits together. the single-year shared benefit at the projected age the second approach involves computing 40 APPENDIX B Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis initial shared benefits at the projected CBO has computed alternative shared bene- in other settings—to develop the official federal claiming age by applying the adjustment for fit measures for multiyear replacement rates by poverty thresholds and the supplemental poverty household size and economies of scale in all applying the latter two approaches. Because measure, for instance—a traditional computation years in which the worker is married and multiyear benefits tend to include a large share of shared resources within a household accounts receives a benefit. That approach implicitly of years when both spouses are receiving benefits, for household size without adjusting for econo- assumes that the spouse who did not claim the alternative specifications of shared benefits do mies of scale. benefits does not work and does not receive not make a substantial difference in the multiyear any other income that could be used to help results. Using the average of the last 20 years of The adjustment for economies of scale, while pay his or her share of the household expenses. shared price-adjusted earnings as the denominator important conceptually, has almost no effect Because that approach understates resources and the average multiyear benefit computed by on initial Social Security replacement rates. For available to the worker’s household, the applying the adjustment for economies of scale in example, for workers born in the 1960s, median resulting replacement rate—25 percent—is all the years in which the worker is married yields replacement rates that are based on shared benefits substantially lower than the 35 percent rate a shared replacement rate for long-career workers if claimed at age 65 and the last 20 years of shared based on the CBO’s preferred specification. born in the 1960s that is only 1 percentage point price-adjusted earnings (including years with no lower than the multiyear replacement rate shown or very low earnings) remain essentially the same ■ To better adjust for the shared resources in this report—45 percent compared with 46 per- with and without the adjustment for economies of and lower per capita living expenses within cent. Using the same denominator and a multi- scale (see Figure B-2). Removing the adjustment the household, the third approach involves year benefit computed by augmenting individual lowers both the numerator and the denominator multiplying the initial benefit by the square benefits in years when only one spouse receives of the replacement rates, resulting in nearly the root of two when only one spouse has claimed a benefit yields a replacement rate of 48 percent, same ratio. benefits. (When both spouses are projected which is only 2 percentage points higher than the to receive benefits together, the sum of their multiyear replacement rate shown in this report. For multiyear replacement rates, however, the benefits is divided by the square root of two.) economies-of-scale adjustment plays a bigger That approach accounts for economies of scale Adjustment for Economies of Scale role. When household composition changes over within a two-person household, which enable When analyzing Social Security replacement rates, the course of retirement, failing to account for a married worker’s Social Security benefit CBO computes shared resources while accounting the corresponding changes in economies of scale to cover a larger share of per-person living for both household size and economies of scale. within the household may produce misleading expenses than an unmarried worker’s benefit. The economies-of-scale adjustment reflects the results. For example, without accounting for However, that approach also assumes that the fact that two spouses generally share expenses for economies of scale, median replacement rates for spouse who did not claim benefits continues housing, utilities, and other items and therefore married workers born in the 1960s would increase to earn wages or to receive nonwage income need less than twice the income that each person after widowhood, in contrast with the findings that is sufficient to pay his or her share of the would need if living separately. The square-root shown in Exhibit 13. Consider a household in household expenses, which is not always the adjustment used in this report implies that a which a worker and his or her spouse received a case. Accounting for economies of scale in married person would need about 30 percent less retired-worker benefit of $1,000 per month and that way results in a shared replacement rate of income than a single person living alone to main- $800 per month, respectively, resulting in a shared 42 percent, which is higher than the 35 percent tain the same general standard of living. Although benefit without an economies-of-scale adjustment rate based on CBO’s preferred specification. an economies-of-scale adjustment is widely used of $900. After the spouse dies, the worker would 41 APPENDIX B Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Figure B-2 . receive only his or her own retired-worker benefit Difference in Median Initial Shared Replacement Rates for Long-Career Workers Born in the of $1,000. Consequently, the worker’s shared ben- efit would increase from $900 before widowhood 1960s Resulting From Adjustments for Economies of Scale Within the Household to $1,000 after widowhood, causing an increase in his or her replacement rate. In contrast, if the Percent square-root adjustment for economies of scale was Replacement Rates 100 No Adjustment With Adjustment applied, the worker’s shared benefits before and for Economies for Economies after widowhood would be $1,273 (that is, $1,800 of Scale of Scale 50 divided by the square root of two) and $1,000, 74 73 respectively, resulting in a decrease in replacement 46 45 26 25 rates in widowhood. 0 Thousands of 2018 Dollars 40 Numerators (Benefits) 20 36 26 28 21 15 13 0 Thousands of 2018 Dollars 150 Denominators (Earnings) 100 142 50 109 43 56 17 21 0 All Quintiles Lowest Quintile Highest Quintile Source: Congressional Budget Office. In this figure, “With Adjustment for Economies of Scale” indicates that a square-root adjustment is applied to account for economies of scale within the household. “No Adjustment for Economies of Scale” indicates that the calculation adjusts for household size by dividing benefits and earnings in each year of marriage by two, rather than the square root of two. The earnings measure (the denominator) is based on the last 20 years of shared price-adjusted earnings, including years with no or very low earnings. All values are before taxes. Replacement rates are computed for all individuals who are eligible to claim retirement benefits at age 62 and who are not receiving any benefit at age 61. Long-career workers are workers with 20 or more years of earnings above 10 percent of the average wage index in each year. To limit the focus to individuals with significant attachment to the labor force, workers with fewer than 20 years of earnings are excluded. In addition, workers who receive Disability Insurance benefits are excluded. 42 Definitions Types of Benefit Measures Presented in years when the worker is not married, his or her already incorporate an adjustment for household This Report earnings enter the calculation without the adjust- size and economies of scale. ment for economies of scale.) Shared benefits are Individual Measures: These measures are gener- computed as the sum of all retired-worker, spou- Scheduled Benefits: Scheduled benefits are ally based on individual earnings and the Social sal, and survivor’s benefits that both the worker benefits as calculated under current law, regardless Security benefits that workers can receive on the and his or her spouse are eligible for on the basis of the amounts available in the combined Social basis of their own work history. To capture all ben- of either of their earnings histories, divided by the Security trust funds. efits that retired workers are eligible to receive, the square root of two. For example, if the worker is individual measure of benefits can be expanded to eligible to receive a $2,000 monthly benefit, and Payable Benefits: Social Security benefits are include spousal or survivor’s benefits that workers his or her spouse is eligible for a $1,000 monthly paid from designated Social Security trust funds. can receive on the basis of their spouse’s earnings. benefit, the worker’s shared benefit would equal Payable benefits are benefits as calculated under However, for simplicity, individual measures in $2,121—that is, $3,000 divided by the square current law and reduced as necessary to conform this report typically include only retired-worker root of two. (Note that the shared benefit credited to the limits imposed by the trust funds’ balances. benefits. to the spouse would also be $2,121, provided that If the trust funds’ balances declined to zero and the spouse is an eligible long-career worker who is current revenues were insufficient to cover benefits Shared Measures: These measures are based on included in the studied population.) If that worker specified in law, the Social Security Administration the earnings and Social Security benefits that a then becomes a widow or widower, the shared would no longer be permitted to pay full benefits worker and his or her spouse can receive, adjusted benefit would be reduced to $2,000. when they were due. The manner in which outlays for economies of scale and household size. In the would be reduced is not specified in law. For this analysis underlying this report, the Congressional Household Measures: These measures are based report, CBO assumed that benefits paid to existing Budget Office applied an economies-of-scale on the earnings and Social Security benefits that and new beneficiaries would be reduced by the adjustment to account for the lower per-person the worker and his or her spouse are eligible percentage necessary to make the program’s total living expenses in households with two members to receive. For this report, household measures annual outlays equal its total available revenues relative to single-person households and to make differ from shared measures in that the household (about a quarter, by CBO’s estimate) once the the shared benefit measures directly comparable measures do not include an adjustment for econ- combined trust funds were exhausted. (The com- with the individual measures. Specifically, shared omies of scale or an adjustment for the size of the bined Social Security trust funds are projected to earnings are computed as the sum of earnings of household. Household measures were used only be exhausted in 2031.) both spouses, divided by the square root of two to compare Social Security benefits of workers (about 1.4), in each year during the marriage. (In with poverty thresholds because those thresholds DEFINITIONS Social Security Replacement Rates and Other Benefit Measures: An In-Depth Analysis Time Spans for Benefit Receipt Used in threshold (or a multiple of the poverty threshold). benefits by quintile of shared lifetime earn- This Report As defined by the Census Bureau, poverty thresh- ings, which are defined as the present value of olds vary by family size and composition. If a price-adjusted shared earnings over a lifetime. Initial (Single-Year) Measures: These measures family’s total income is less than the family’s corre- When the sample is divided into five groups that are based on Social Security benefits generally sponding threshold, then that family is considered are ranked according to those earnings, a quintile measured at age 65, under the assumption that to be in poverty.1 In 2018, the poverty threshold is one of those five groups. (A present value is a workers first claim benefits at that age. This is a for a single person age 65 or older was $12,043; single number that expresses a flow of current and hypothetical measure of benefits that does not for two people, it was $15,178.2 future income or payments in terms of an equiva- account for variation in the observed timing of lent lump sum received or paid at a specific time. retirement and, therefore, does not equal the pro- Preretirement Earnings: Researchers who seek The present value depends on the rate of interest, jected first-year benefit that workers receive. The to determine whether retirees would be able to known as the discount rate, that is used to trans- advantage of holding the age at which benefits are maintain the standard of living they attained late future cash flows into current dollars.) claimed constant at 65 is that it removes the effect before retirement compare retirement income of changes in benefit-claiming patterns over time. with preretirement earnings, calculating a replace- Birth Cohort: A birth cohort is a group of people ment rate—the amount of income in retirement born in a given period. In this report, CBO Multiyear Measures: These measures are based expressed as a percentage of income before projects outcomes for people placed into one of on the average shared benefits received from the retirement. The analysis described in this report three 10-year birth cohorts: the 1940s cohort first year of projected claiming (which may not considers only the portion of retirement income (people born between 1940 and 1949); the 1960s be 65) until death, adjusted for changes in prices that comes from Social Security benefits and the cohort (people born between 1960 and 1969); or over time. Although retired-worker benefits portion of preretirement income that comes from the 1980s cohort (people born between 1980 and remain generally the same over that period, shared earnings. 1989). measures that incorporate benefits of the workers and their spouses may change. In the construc- Other Frequently Used Terms Price-Adjusted Earnings: Price-adjusted earnings tion of multiyear benefit measures, CBO projects account for inflation and reflect the purchasing all benefits received by retired workers or their Social Security Replacement Rate: The Social power of earnings over time. In this report, CBO spouses from the workers’ earliest eligibility age of Security replacement rate is the amount of Social uses the price index for personal consumption 62 until death, including spousal and survivor’s Security benefits received in retirement, expressed expenditures as the measure of inflation. benefits, if any. as a percentage of earnings before retirement. Wage-Adjusted Earnings: Wage-adjusted earn- Benchmarks Used in This Report for Shared Lifetime Earnings Quintile: In this ings account for changes in the average wage index Evaluating Benefits report, CBO describes measures of Social Security (AWI) and reflect the increase in workers’ average standard of living over time. The AWI is based on Official Federal Poverty Threshold: Researchers 1. See Census Bureau, “How the Census Bureau Measures compensation subject to federal income taxes and who seek to determine how many retirees are able Poverty” (accessed February 22, 2017), https://go.usa. contributions to deferred compensation plans. to afford essential living expenses often compare gov/xRGnv. retirement income with the official federal poverty 2. See Census Bureau, “Poverty Thresholds” (accessed February 22, 2017), https://go.usa.gov/xEXMu. 44 About This Document This Congressional Budget Office report was prepared at the request of the Chairman of the Senate Committee on the Budget. In keeping with CBO’s mandate to provide objective, impartial analysis, the report makes no recommendations. Marina Miller of CBO’s Health, Retirement, and Long-Term Analysis Division wrote the report with guidance from Julie Topoleski and David Weaver. Ed Harris, Nadia Karamcheva, Noah Meyerson, Kevin Perese, Charles Pineles-Mark, Adam Russell, Kurt Seibert, Michael Simpson (formerly of CBO), John Skeen, and Jeff Werling provided useful comments on various drafts of the report. Jimmy Chin provided fact-checking. In addition, Andrew G. Biggs of the American Enterprise Institute, Peter J. Brady of the Investment Company Institute, Michael D. Clingman and Patrick J. Purcell of the Social Security Administration, and Alicia H. Munnell of the Center for Retirement Research at Boston College provided helpful comments. (The assistance of external reviewers implies no responsibility for the final product, which lies solely with CBO.) Wendy Edelberg, Mark Hadley, Jeffrey Kling, and Robert Sunshine reviewed the report, Loretta Lettner edited it, and Jorge Salazar prepared it for publication. The report, supplemental data, and an interactive tool are available on CBO’s website (www.cbo.gov/publication/55038). CBO continually seeks feedback to make its work as useful as possible. Please send any feedback to communications@cbo.gov. Keith Hall Director April 2019 45