The Importance of Social Security Benefits to the Income of the Aged Population

Introduction

The traditional major sources of retirement income in the United States—often called the three-legged stool or the three pillars—are Social Security benefits, employer-provided pensions (including retirement accounts), and income from assets or savings. Social Security is a social insurance program that provides an inflation-indexed lifetime annuity to aged beneficiaries. In addition to enjoying the protection provided by indexing, a prospective beneficiary who delays claiming Social Security benefits essentially purchases additional longevity insurance—reducing the risk of “running out of savings”—by raising his or her lifetime monthly benefit (Shu, Payne, and Sagara 2014; Shoven and Slavov 2012). Many observers regard Social Security benefits as the base of retirement income, particularly because benefits are a steady and reliable resource for almost all aged households (Brady, Burham, and Holden 2012; American Council of Life Insurers, American Benefits Council, and Investment Company Institute 2013; Poterba 2014). Because Social Security benefits represent a substantial portion of the income of Americans aged 65 or older (Social Security Administration [SSA] 2002, 2012, 2014, 2016a, 2016b), accurate measurements of that portion are important to researchers and policymakers (Banerjee 2013; Employee Benefit Research Institute 2013; Miller and Schieber 2013, 2014). Using data from the Current Population Survey (CPS), SSA estimates that in 2014, about 84 percent of people aged 65 or older received Social Security benefits; and among those in the bottom 40 percent of the income distribution, benefits accounted for an average of around 84 percent of total income (SSA 2016b).

Some analysts have criticized the use of CPS data to underlie such estimates. Research has suggested that the CPS does not adequately measure income from certain sources—in particular, income from retirement accounts, such as individual retirement accounts (IRAs) or defined contribution (DC) plans (Miller and Schieber 2014). Specifically, researchers have argued that estimates based on CPS data were likely to overstate older Americans' reliance on Social Security benefits and to understate their reliance on income from retirement accounts, particularly among lower-income respondents. In response, the Census Bureau changed the income questions in the 2015 CPS, aiming to account more accurately for income drawn from retirement accounts. In addition, trends in recent decades in employer-provided pension offerings, societal changes, and Social Security program rule changes may have affected the relative importance of different income sources for older Americans, particularly that of Social Security. Thus, it is important for policymakers to have an accurate picture of the composition of retirement income so that any proposed changes to Social Security may better address the needs of the aged.

This article assesses the extent to which Americans aged 65 or older rely on Social Security benefits. We use data from the 2015 CPS, which incorporates the revised income questions and may therefore provide more accurate results than were provided in previous survey years. We compare 2015 CPS results with those from the 2013 CPS to assess the effect of the survey revisions. We also attempt to validate the 2015 CPS results by comparing them with those from the 2008 panel of the Survey of Income and Program Participation (SIPP) and the 2012 wave of the Health and Retirement Study (HRS). Unlike CPS results on Social Security benefit income, which are based solely on the survey reports, data from the latter two surveys can be augmented with verifiably accurate information from Social Security administrative records, which contain data on Social Security benefits that respondents received in a given year. In addition, the latter two surveys provide (or allow us to calculate) information on income from retirement accounts.

To examine the extent to which persons aged 65 or older rely on Social Security, we estimate the proportions of aged Americans for whom Social Security benefits account for (1) at least 50 percent and (2) at least 90 percent of their family income.¹ Interestingly, the estimates are quite similar, despite design differences across the three surveys. We find that about half of the population aged 65 or older live in households that receive at least 50 percent of their family income from Social Security benefits and about 25 percent of aged households rely on Social Security benefits for at least 90 percent of their family income.

In the next three sections, we discuss findings from previous research, explore possible reasons for changes in recent decades in the relative importance of certain income sources for the aged, and describe the data and methods we use. In the final two sections, we present our findings and conclude with a discussion of policy implications.

Previous Research

SSA has published statistics on the income of the aged population based on CPS data since the 1970s. From 1976 through 2006, about 90 percent of people aged 65 or older lived in households receiving income from Social Security.² Over that period, the average share of income from Social Security was always substantial (between 66 percent and 84 percent in any given year), particularly for households in the bottom half of the income distribution. Poterba (2014, Table 6), using data from the 2013 CPS, also finds wide variation across total-income quartiles in the distribution of income by source for individuals aged 65 or older.

Using data from the redesigned March 2015 CPS, the Federal Interagency Forum on Aging-Related Statistics (2016, Table 9a) reports the percentage distribution of per capita 2014 family income by source, overall and in each total-income quintile for persons aged 65 or older.³ Social Security benefits were the primary income source, accounting for an average of about 49 percent of total family income for aged individuals. Combined income from annuities and pensions (including distributions from retirement accounts) amounted to 16 percent of family income, and income from assets accounted for 6 percent. Beyond the three traditional pillars, earnings—now often considered a fourth pillar of retirement security—accounted for about 24 percent of family income, reflecting an increase in continued employment among the aged or the presence of younger workers in the family, or perhaps both. Public assistance and “other” sources respectively accounted for 2 percent and 3 percent of per capita family income of the aged population.

The table also shows wide variation in income distribution by source across family-income quintiles. Social Security benefits in 2014 accounted for between 54 percent and 72 percent of family income in the three lowest income quintiles, compared with 18 percent to 34 percent of family income in the two highest quintiles. Furthermore, for aged individuals in the lowest two income quintiles, the share of family income received from private and public pensions was trivial (less than 8 percent), compared with around 25 percent for those in the highest two income quintiles. Similarly, the share of income from earnings was less than 14 percent among aged individuals in the lowest two income quintiles; but for those in the highest income quintile, it was much more important (40 percent). In addition, assets were a minor source of income for aged individuals in all income quintiles except the highest, for whose members they provided on average 13 percent of income. Together, those findings show that aged individuals in the lower income quintiles rely much more on Social Security benefits than their counterparts in the highest quintile do and that pensions, earnings, and assets are not very important income sources for aged persons with lower incomes.

Changes in Retirement Income by Source

The share of income from Social Security among persons aged 65 or older may have changed over time because of trends in pension offerings, societal changes, and program rule changes. We examine each factor in turn.

Pension Offerings

Employer-provided pensions are an important source of U.S. retirement income (Hardy and Shuey 2000; Herd 2009; O'Rand 2011; Poterba 2014; Shuey and O'Rand 2004; Warner, Hayward, and Hardy 2010). Over the last three decades, the dominant pension offering changed dramatically, from the defined benefit (DB) type to the DC type (Costo 2006; Mackenzie 2010; Wiatrowski 2012; Anguelov, Iams, and Purcell 2012). In DB plans, employees are enrolled automatically, and employers fund most or all of the pension plan. Employers also bear the capital-market and longevity risks related to providing income (in the form of an annuity) to retired workers. Because DB plans lack portability of pension accruals across jobs, they are risky for workers with high job turnover. DC plans are more attractive to those workers because DC accrued account balances are portable. However, DC plan participation is voluntary and employees bear all the investment risks.⁴ In addition, their account balances—and consequently the income such accounts generate in retirement—depend not only on the amounts contributed over time, but also on whether those contributions were subject to earnings and employment shocks during the working years (Dushi and Iams 2015). As Johnson (2016, 63) observes of DC pensions: “These do-it-yourself retirement plans can generate substantial retirement income only if workers choose to make significant contributions to their accounts each pay period, resist the temptation to dip into their accounts before they retire, and manage funds wisely after they retire.”

Predictably, the changing landscape of pension offerings led to a dramatic shift in aggregate asset holdings from traditional DB plans toward tax-advantaged DC plans or IRAs. In 1981, Americans held $673 billion in DB plan assets, $174 billion in DC plan assets, and $38 billion in IRAs. By 2014, traditional retirement plans held $8.0 trillion in assets, DC retirement plans held $6.3 trillion, and IRAs held $7.4 trillion (Federal Interagency Forum on Aging-Related Statistics 2016, Table 11c). Notably, most IRA assets reflect transfers from tax-advantaged DC retirement plans (Holden and Schrass 2016).

Despite the growth in recent decades in aggregate retirement assets and holdings, estimates based on CPS data show that the percentage of income received from public and private pensions (including IRAs) among aged units (single persons aged 65 or older or couples with at least one member aged 65 or older) increased from 18 percent in 2000 to 21 percent in 2014. Over the same period, the share of their total income attributed to assets decreased from 18 percent to 10 percent (SSA 2002, Table 7.1; 2016b, Table 10.1); interest rate changes over that span may have contributed to the latter trend.

Although the above-noted changes in income by source seem muted, the outlook for future retirees seems uncertain as studies continue to document low retirement savings among American workers (Munnell 2014; Ghilarducci 2014; Knoll, Tamborini, and Whitman 2012). Estimates based on the 2013 Survey of Consumer Finances, for example, indicate that 41 percent of American households headed by individuals aged 55–64 have no savings in retirement accounts. Even more striking is the sharp variation by household income. Among households headed by individuals aged 55–64, the proportion with any retirement savings ranges from 9 percent in the lowest income quintile to 68 percent in the middle quintile and to 94 percent in the top income quintile. Among the 59 percent of households that have some retirement savings, the median amount saved is about $104,000, and one-quarter of those households have saved less than $26,000 (Government Accountability Office 2015, Tables 1–3). Such savings may not provide an adequate annuity payment over the period of retirement. In sum, the shift in the dominant type of pension plan offerings and the resulting shift in the income they can generate is likely to have influenced not only the proportion of the retired population that draws pension income but also the composition and importance of such income in retirement.

Societal Changes

Increasing labor force participation among women and among older workers of either sex, particularly during the last decade, has led to an increase in earned income among people aged 65 or older. From 1999 to 2014, the labor-force participation rate of men aged 65–69 increased from 29 percent to 36 percent; for men aged 70 or older, it increased from 12 percent to 16 percent. Over the same period, the labor-force participation rate of women aged 65–69 increased from 18 percent to 28 percent, and for women aged 70 or older it rose from 6 percent to 9 percent (Federal Interagency Forum on Aging-Related Statistics 2016, Table 12). As a result, the percentage of family income derived from earnings among aged households increased from 23 percent in 2000 to 32 percent in 2014 (SSA 2002, Table 7.1; 2016b, Table 10.1).

Facing scarce employment opportunities during the Great Recession of 2007–2009, some older unemployed workers claimed early benefits (Haaga and Johnson 2012). As a result, their monthly Social Security benefit amounts were reduced relative to the benefits they would have received if they had claimed at full retirement age (FRA). Furthermore, the changing marital histories, educational attainment, and patterns of lifetime labor-force attachment of women have generally increased the retired-worker benefits to which they are entitled based on their own earnings while reducing their auxiliary (wife or survivor) benefits (Iams and Tamborini 2012; Butrica and Smith 2012; Iams 2016). These trends have reduced the Social Security benefits of many couples (Sass 2016).

Programmatic Changes

Social Security program changes, such as claiming-age rule changes, can strongly affect the level of Social Security benefits (Shoven and Slavov 2012). The 1983 Amendments to the Social Security Act stipulated that the FRA of 65 for individuals born before 1938 would be adjusted upward for those born in later years. The FRA increases by 2 months for members of each successive birth cohort from 1938 through 1942, reaching 66 for those born in 1943.⁵ Because monthly benefits are permanently reduced for individuals claiming before reaching their FRA, the increased FRAs for members of later birth cohorts affect their Social Security income. For example, the monthly benefit of a person who claims at age 65 is reduced by 6.7 percent if her or his FRA is 66 (born 1943–1954) versus no reduction for a person whose FRA is 65 (born in 1937 or earlier). Likewise, a person whose FRA is 65 and who claims benefits at age 66 receives a delayed retirement credit of as much as 6.5 percent, versus no upward adjustment for someone whose FRA is 66.⁶

The Senior Citizens' Freedom to Work Act of 2000 instituted another programmatic change by eliminating the retirement earnings test (RET) for working beneficiaries who had reached FRA. Prior to that law's enactment, benefits were reduced for working beneficiaries with earnings above given thresholds. After the RET elimination, retired-worker benefit claims spiked in 2000, particularly among workers who had reached FRA. In addition, increasing shares of claims were delayed in subsequent years among those workers who had not reached their increased FRA (Purcell 2016; Song and Manchester 2007).

In sum, the changes mentioned above led to a decrease in Social Security benefits as a percentage of total family income for aged households, from 38 percent in 2000 to 33 percent in 2014 (SSA 2002, Table 7.1; 2016b, Table 10.1).

Data and Methods

For this analysis, we use data from three nationally representative surveys: the 2015 CPS, sponsored jointly by the Census Bureau and the Bureau of Labor Statistics; wave 11 from the 2008 panel of the Census Bureau's SIPP; and the 2012 wave of the University of Michigan's HRS.⁷ The CPS income questions refer to amounts received in the calendar year preceding the survey year; in the HRS, they refer to income received in the prior month, and respondents' answers are annualized for the survey year; in the SIPP, the questions refer to income in the survey months, and responses likewise are annualized. Hence, our variables measure income in 2014 for the CPS and in 2012 for the SIPP and the HRS. We use the 2012 data for the HRS and the SIPP (and not more recent data) because when this article was written, information on Social Security benefits from administrative records were not available beyond 2012.

Each survey provides information on socioeconomic characteristics (such as sex, marital status, race, Hispanic origin, education, income, and age group) and on income by source (such as Social Security, pensions, assets, earnings, and welfare programs). We estimate the proportion of income that comes from each source. In particular, we examine how reliance on Social Security benefits varies across socioeconomic subgroups.

The sample for this analysis consists of individuals aged 65 or older. For each individual, we define his or her total family or household income as the sum of all income from all members of the family or the household. Similarly, the total income from Social Security is the sum of benefits received by all family members. We then calculate the share of total family or household income received from Social Security and estimate the percentage of the aged population that relies on Social Security benefits as a primary source of income in retirement. We examine two thresholds of reliance on Social Security benefits. Specifically, we calculate the percentage of aged individuals who live in a household that derives (1) at least 50 percent and (2) at least 90 percent of family income from Social Security benefits.

The CPS

We use data collected in the March 2015 CPS Annual Social and Economic Supplement. Respondents from a nationally representative sample of the U.S. population were asked detailed questions about income in the previous year for each person in the household, including whether they received any Social Security income, the amount of any such income, and the benefit type (retired worker, disabled worker, or dependent/survivor). Respondents were also asked if Medicare premiums were deducted from their Social Security benefits and, if so, how much. Based on responses to these questions, the Census Bureau calculated the Social Security income (which includes Medicare premiums) for each family member and then calculated the total family income. Because our sample is restricted to people aged 65 or older, Social Security income mostly comes from retired-worker, spouse, and survivor benefits.

Critics have claimed that the CPS inadequately measures asset income and distributions from tax-advantaged retirement accounts such as 401(k) plans or IRAs (Iams and Purcell 2013; Fisher 2008; Davies and Fisher 2009; Munnell and Chen 2014). Hence, previous research has argued that CPS-based estimates of the distribution of income of the aged population are likely to overstate reliance on Social Security benefits and understate reliance on retirement accounts (Miller and Schieber 2014). To address those criticisms, the Census Bureau redesigned the CPS for 2015 to improve the collection of data on sources of income received in the reference year (2014 in this case), particularly for the aged population. The redesign sought in part to reduce respondents' query fatigue by omitting questions for which the answer could be determined based on the response to an earlier question. Also, a “dual-pass” approach was implemented by asking first about the sources of income and then about the amount from each reported source. Additionally, the redesigned questionnaire revised the order in which the income questions appear, to better capture accurate information on the most likely sources of income among three types of respondent households: (1) those with a member aged 62 or older, (2) those with low income, and (3) all others. More importantly, for the first time, the 2015 CPS asked separate questions about retirement-account withdrawals and distributions and collected information on property income. The Census Bureau tested the redesign on a randomly selected subsample of ⅜ of the full 2014 CPS sample; the rest of the 2014 sample replied to the traditional questionnaire (Semega and Welniak 2015).

Among aged households, the estimated real median income among redesign respondents was 4.6 percent greater than that of traditional-questionnaire respondents. Furthermore, estimates based on the redesigned questionnaire of the prevalence and the aggregate value of retirement income from sources other than Social Security were about 50 percent and 22 percent higher, respectively, than estimates based on the traditional questionnaire. By contrast, the estimated prevalence and aggregate value of Social Security income were both only about 2 percent greater under the redesigned survey (Semega and Welniak 2015, Tables 1–2). Nevertheless, even in the redesigned-CPS sample, Social Security “remains the overwhelmingly predominant source of income for those ages 65 and older” and “over 60 percent of individuals in the two lowest-income quartiles received more than 90 percent of their total income from Social Security” (Copeland 2015, 11).

As noted earlier, the 2015 CPS data should provide a more accurate picture of the Social Security share of income for aged individuals. Comparing the 2015 CPS data with those from the other two surveys will provide a validity test of their accuracy.

The SIPP

The 2008 SIPP panel started in late 2008 and continued through 2013, with a new wave of interviews conducted every 4 months. For this analysis, we use the monthly SIPP data collected in 2012. The survey data are routinely matched to Social Security administrative records; 93 percent of respondents in the 2008 SIPP panel were matched to their SSA records.

SIPP respondents were asked detailed questions about the income sources and government program participation of each individual in the family. Starting with the 2015 CPS, the official Census Bureau definition of family income includes distributions from retirement accounts (which the Bureau calls lump sums). Therefore, in our definition of family income for the SIPP data, we include distributions from retirement accounts (as calculated by the Bureau for calendar year 2012). In addition, because SIPP-reported Social Security benefits do not include Medicare premiums (Iams and Purcell 2013), we replace the respondent's reported Social Security benefits with the total amount of Social Security benefits (calculated as the sum of the amounts of benefits paid by check or deposited to a bank account and the amount of Medicare premiums) using data from SSA's Payment History Update System.⁸

The HRS

The HRS is the most comprehensive national longitudinal survey of Americans aged 51 or older. It began in 1992 and follow-up interviews have been conducted every other year since then. The purpose of the HRS is to provide high-quality data to examine “the ways in which older adults' health interacts with social, economic, and psychological factors and retirement decisions.” By conducting “unique and in-depth interviews,” it also provides a comprehensive inventory of the income and wealth of the population aged 51 or older (National Institute on Aging 2007).

The HRS is more systematic than the CPS and the SIPP in collecting information on retirement-plan account balances and distributions. If income or wealth information is missing, HRS respondents are asked follow-up questions about the dollar amount using an “unfolding brackets” approach to identify the range limits of the missing data item.

Czajka and Denmead (2008) find that HRS-reported household incomes in 2002 for persons aged 51 or older were substantially higher (by 20–30 percent) than those reported in the CPS and the SIPP. The characteristics of this aged population were largely similar across the three surveys, although HRS respondents were slightly more likely to live with others and less likely to live alone than were their CPS and SIPP counterparts. The authors conclude that “HRS incomes are higher than those of the Census Bureau surveys, but resolving whether this is due to better measurement or over-representation of higher-income families must be left to future research.”

In this article, we use household income information obtained from the RAND Corporation's user-friendly HRS data file (version O). Specifically, we focus on variables that correspond to the 2012 wave. The HRS household-income measure includes earnings, private pensions, Social Security benefits, and income from welfare programs, capital, and other sources. For married respondents, total household income combines that of both spouses.

The HRS understates total Social Security benefit amounts because it asks respondents to report the net amount. Specifically, in 2012, question NQ085 asked: “About the Social Security income that you (yourself) receive, how much was the Social Security check or the amount deposited directly into your account last month? We want the amount after any deductions.” Note that the amount paid to the respondent as a check or direct deposit does not include the amount of Medicare premiums withheld from the total benefit. Given the wording of this question, it is plausible that respondents report only the amount paid to them rather than the total or gross benefits.

The true Social Security benefit amount (either gross or net) can be determined from Social Security administrative records. For about half of the respondents in the 2012 wave of the HRS, we have matched records containing information on their earnings and Social Security benefit. For those respondents, we update their self-reported benefit amount with administrative information (that is, the sum of benefits paid and of the Medicare premiums).⁹ We use data for 2012 because we do not have matched administrative data beyond 2012. For respondents without matched records we add to their survey-reported benefits an amount of $1,200, which is the average (and the median) Medicare premium observed in the Payment History Update System records. Thus, for couples, the amount of Social Security benefits is the sum of the corrected Social Security benefits received by both spouses. Then, we correct the total household income variable in the RAND HRS data file with this adjusted measure of Social Security benefits.

Although the HRS household-income measure does not include withdrawals or distributions from IRAs or 401(k)-type accounts, respondents report those account balances in either the survey's wealth module or its employment module. We use respondents' self-reported balances to calculate the stream of annual income one can withdraw from such tax-advantaged retirement accounts. Because people aged 70½ or older are legally required to take annual minimum distributions from their account balances, we assume that they have received such distributions even if they are not reported as income. To estimate the distribution amount, we use the Internal Revenue Service (IRS) required minimum distribution factor, which is based on life expectancy at a given age. Research has shown that few people draw distributions from IRAs (Holden and Schrass 2016) or from 401(k)-type accounts (Poterba, Venti, and Wise 2011a, 2011b) before they are required to do so. Nevertheless, to measure retirement-account income consistently, we extrapolate the distribution factors for persons aged 65–69 from IRS data for persons aged 70½ or older. Then, we add the estimated amount of annual withdrawals to total household income and use the corrected measure to estimate the reliance on Social Security benefits.

Results

To examine the effects of the CPS redesign, we start by comparing estimates from the 2013 CPS (covering income in 2012) with those from the 2015 CPS (covering income in 2014). For both years, Table 1 presents the estimated proportion of respondents aged 65 or older for whom Social Security provided at least 50 percent or at least 90 percent of their family income. In 2014, about half (52 percent) of aged persons lived in families that derived at least half of their total income from Social Security benefits. As expected, that figure is lower than the estimated percentage for 2012 (56 percent), most likely because the redesigned CPS better measured income from other sources. About one-quarter of the aged population lived in families that received 90 percent or more of their family income from Social Security benefits, and the estimated percentage for 2012 (27 percent) was only slightly higher than that for 2014 (25 percent). Although the redesigned CPS income questions resulted in slightly lower estimates of reliance on Social Security benefits, the overall pattern did not change much.

Reliance on income from Social Security varies greatly by socioeconomic characteristics. Women relied on Social Security benefits more than men did. In 2014, 55 percent of women and 48 percent of men lived in families receiving at least half of their income from Social Security benefits, and the corresponding estimates for the 90 percent threshold are 27 percent and 21 percent. Similarly, nonmarried respondents in 2014 relied on Social Security substantially more than married respondents did: 60 percent versus 46 percent, respectively, at the 50 percent threshold and 33 percent versus 19 percent at the 90 percent threshold. Across race/ethnicity groups, non-Hispanic blacks were more likely to receive at least half of their income (57 percent) and at least 90 percent of their income (33 percent) from Social Security in 2014 than were respondents in other groups. Reliance on Social Security income decreases with higher education levels. Around two-thirds of aged respondents with less than a high school degree or with a high school degree relied on Social Security benefits for at least half of their income in 2014, compared with about one-third of college graduates. Furthermore, a substantial proportion (41 percent) of those who did not complete high school relied on Social Security benefits for at least 90 percent of their family income, compared with 14 percent of college graduates.

Expectedly, reliance on Social Security benefits decreases as family income increases. Differences across the income distribution are substantial: Respondents in the lowest and second-lowest income quintiles in 2014 were much more likely (87 percent and 82 percent, respectively) to receive at least half of their family income from Social Security than were those in the highest income quintile (2 percent). The corresponding estimates for those at the 90 percent threshold of reliance were 64 percent for those in the lowest income quintile and 0 percent for those in the highest income quintile. Finally, reliance on Social Security income increases with age, suggesting that as people get older they may have depleted other income sources, without which Social Security becomes even more important. In 2014, the proportions of persons receiving at least half of their income from Social Security were 42 percent at ages 65–69, 51 percent at ages 70–74, 57 percent at ages 75–79, and 61 percent at ages 80 or older. The respective proportions receiving at least 90 percent of income from Social Security were 18 percent, 23 percent, 27 percent, and 33 percent. For every socioeconomic subgroup except one, the percentages in 2014 were lower than in 2012.

Table 2 compares estimates of reliance on Social Security benefits based on data from the March 2015 CPS (covering income in 2014) with estimates based on data from the 2012 HRS and the 2008 panel of the SIPP (each covering income in 2012). Despite differences in reference years and methodologies, the three surveys produce very similar estimated percentages of aged persons who live in households that receive at least half of their income from Social Security (around 54 percent in the HRS and the SIPP and 52 percent in the CPS). Estimates of reliance at the 90 percent threshold, however, are more divergent: about one-quarter of the aged population in the HRS and the CPS, and about one-fifth in the SIPP. It is unclear why the SIPP estimate is lower than those from the other two surveys. Perhaps, because the SIPP was designed to focus on income and program participation of the low-income population, it may better reflect the composition of their income. In any event, the patterns of reliance across socioeconomic subgroups are generally consistent across the surveys. Notably, even though we correct for retirement-account withdrawals and distributions, the proportion of aged persons who rely on Social Security for at least half of their family income is for many subgroups higher in the HRS than it is in the other two surveys. In particular, HRS estimates of reliance are higher for women, the nonmarried, all race/ethnicity groups other than non-Hispanic whites, persons with no postsecondary education, all income quintiles except the lowest, and the 70–74 and 75–79 age groups. These findings suggest that Social Security remains the primary source of retirement income for substantial segments of the aged population and that retirement accounts, despite their increased prevalence, have not changed the importance of Social Security benefits for the majority of the aged population.

Summary and Conclusion

Analysis of three independent surveys—the HRS, the SIPP, and the CPS—reveals that despite their different samples, designs, and approaches to measuring income by source, they yield similar results regarding the importance of Social Security benefits to the income of the aged. They confirm the findings of previous research that Social Security benefits provide the majority of retirement income to persons aged 65 or older. Estimates based on data from the three surveys reveal that about half of the aged population live in households receiving at least 50 percent of their family income from Social Security benefits and about one-quarter live in households receiving at least 90 percent of their family income from Social Security.

In the CPS, the estimated proportion of persons aged 65 or older who relied on Social Security benefits for at least half of their family income was lower in 2014 (52 percent) than in 2012 (56 percent). Similarly, the estimated proportion receiving 90 percent or more of their family income from Social Security benefits was slightly lower in 2014 (25 percent) than in 2012 (27 percent). These seeming decreases likely reflect better measurement of asset and retirement-account income in the redesigned 2015 CPS, leading to apparent increases in estimated income for 2014 from those sources, although the increases are not substantial enough to affect the reliance on Social Security. Nevertheless, the results of even the redesigned CPS indicate that persons aged 65 or older rely heavily on Social Security benefit income.