Angst about the U.S. retirement system is widespread. Social Security has been considered underfunded since the mid-1980s, prompting some groups to want to trim benefits to keep the system afloat.1 Others believe that retirees need more Social Security benefits, not less. They contend that Social Security isn’t generous enough, and that other retirement income sources — such as employer plans and individual retirement accounts (IRAs) — are unreliable or skewed toward higher-income workers and retirees.2

The evidence is overwhelming — or is it?

The Social Security Administration’s (SSA’s) survey evidence suggests that Social Security is the financial bedrock of most American retirees. Since the early 1990s, income from personal assets and employer-sponsored pensions has played a diminishing role as the elderly have become more dependent on wage and salary income (Figure 1).

Figure 1. Income sources for people age 65+, 1976 – 2010
Towers Watson Media

Source: Social Security Administration

According to the SSA, nearly one-quarter of families with a member older than 65 in 2010 relied entirely on Social Security, and nearly half (43%) received 80% or more of their income from the program. In 2010, Social Security provided 84% of income to those in the bottom 20% of the income distribution, 83% of income to those in the next 20% and 66% of income to those in the middle third.3

Policy discussions about retirement income often rely on Social Security’s analysis of the Current Population Survey (CPS) or on other researchers’ analyses based on the CPS. The Center on Budget and Policy Priorities, for example, used Social Security’s 2010 analysis of income sources of the elderly to report that Social Security provides the majority of income for 65% of the elderly and provides more than 90% of income for 36% of them.4

Selena Caldera, writing for AARP, tallied the CPS herself to show that 13.7% of older Americans live on Social Security alone, 9.9% rely on Social Security for 90% to 99.9% of their income, and 25.5% receive between 50% and 89.9% of their income from the program.5

The importance of Social Security as the primary income source for the elderly is often used to document the need for bigger benefits. Virginia Reno, vice president for income security at the National Academy of Social Insurance, recently argued for more generous Social Security benefits before the Social Security Subcommittee of the House Ways and Means Committee, using the SSA’s 2008 analysis to support her position.6

Michael Lind, a policy director at the New America Foundation, and two colleagues cited 2010 data from Social Security showing that the elderly receive only 18% of their income from traditional pensions and defined contribution plans, and that among the middle 20% of the income distribution, fewer than half of retirees receive pension income from employer-sponsored retirement plans. They propose scrapping current tax preferences for retirement savings and financing a new tier of Social Security benefits with the resulting tax revenues the government would collect.7

CPS measures overlook substantial retirement income

To estimate retiree income, the SSA uses data from the Current Population Survey, which is developed by the Census Bureau based on monthly surveys of representative U.S. households. Each March, as part of the CPS, the Census Bureau administers a special “Annual Social and Economic Supplement,” which gathers detailed data on income sources, levels and labor force activities over the previous year for household members 15 or older. The March 2011 survey included approximately 97,000 households.

Figure 2 compares aggregate income paid from employer-sponsored retirement plans in the CPS with comparable data in National Income and Product Accounts (NIPA), which are produced by the Bureau of Economic Analysis of the Department of Commerce. The employer plans include defined benefit pensions, other annuities and individual account plans. Distributions from private tax-qualified plans can be tracked by Form 5500 reports submitted to the IRS, and those from public plans can be accumulated from sponsors’ disclosure or budget reports.

Figure 2. Estimated pension and defined contribution plan income from CPS March Income Supplements and National Income and Product Accounts, 1980 – 2010 ($ billions)
Towers Watson Media

Sources: Authors’ tabulations of various years of the March Supplement to the CPS and Bureau of Economic Analysis, U.S. Department of Commerce, National Income and Product Accounts

In developing National Income and Product Accounts, the Bureau of Economic Analysis pulls data from various sources. In 1990, the National Income and Product Accounts estimates of benefits paid from employer-sponsored retirement plans were 60% greater than all non-Social Security retirement income reported in the CPS, and by 2010, the estimates were twice those from the CPS.

While the CPS clearly undercounts retirement income from employer-sponsored retirement plans, some retirement benefits in the National Income and Product Accounts may include preretirement distributions from 401(k)-type retirement savings accounts. Where these distributions are rolled over into an IRA, CPS respondents would probably not consider the rollovers as income at the time and it is likely the National Income and Product Accounts aggregates do not either. But benefits paid before retirement that are not rolled over into an IRA may not be reported consistently as income across the various sources. Thus, the comparison — while useful — is not strictly apples to apples.

For a more accurate comparison, we turn to sample tax files released by the Internal Revenue Service (IRS) with reported sources and amounts of income. We compare IRS reported income for 1988, 2000, 2007 and 2008 (the most recent year available) with the CPS March Income Supplements for those years.

The 2008 tax files report that an estimated 31.6 million filers received either pension/annuity income or taxable income from a tax-favored IRA. According to the CPS, however, only 17.3 million participants received such income. Figure 3 shows distributions from pension/annuity plans and from IRAs as reported on federal income tax forms versus those reported in the CPS. Theoretically, the CPS should show more people receiving retirement income than the tax filings, because some relatively low-income retirees who don’t file income tax returns receive pension income.

Figure 3. Reported pension/annuity income and distributions from IRAs in tax files and the CPS, 1988 – 2008 ($ millions)
Towers Watson Media

Source: Authors’ tabulations of the sample of Form 1040 filings and the CPS for selected years

For filers receiving Social Security benefits, the 2008 tax files show $110.9 billion in IRA distributions while the CPS shows only $5.6 billion. An earlier analysis found that, for 1990, pension and IRA income was $112 billion in the tax files versus $88 billion in the CPS.8 The underreporting of IRA income seems to be worsening over time.

The CPS captures very little of the IRA income paid to retirees in tax filings: 3.5% in 2000 and roughly 5% to 6% in 2007 and 2008. The CPS has done a better job of capturing pension and annuity income but still comes up short by around half, with the underreporting becoming more pronounced in recent years.

Employer plans and IRAs: Not for fat cats only

Many policymakers and analysts claim that the income and tax advantages of employer pensions and IRAs fall mostly to wealthier workers and retirees. To see where those benefits are going, our analysis divides the analysis population into 10 income deciles. 

Figure 4 shows income levels reported in the 2008 tax files and the CPS. In the tax files, Social Security recipients had higher income at every decile than the total CPS population. For example, the third decile includes those whose incomes fall between the 30th and 40th percentiles of the income distribution. Among the tax filers reporting Social Security income, incomes in the third decile ranged from $25,967 to $32,449, while the range for those in the total population reporting Social Security income on the CPS was $16,599 to $24,010.

Figure 4. Income ranges within income deciles from tax files and CPS, 2008
Towers Watson Media 

Source: Authors’ tabulations of the sample of Form 1040 filings and the CPS for 2008

In 2008, an estimated 23.4 million tax filers reported receiving Social Security benefits versus 32.9 million CPS participants. To reflect the fact that many low-income retirees do not file income taxes and thus make the data more comparable, we eliminated the bottom 20% of CPS participants based on income, leaving an estimated 24.5 million units with Social Security income for 2008.

Figure 5 shows median reported Social Security benefits for each income decile from the 2008 tax files and the CPS for the top 80% of participants based on income. Those in the first four deciles have somewhat higher median Social Security benefits in the CPS than in the tax files, and the overall distribution is flatter in the CPS.

Figure 5. Income in 2008 tax files versus CPS by deciles (excluding bottom 20% from CPS)
Towers Watson Media 

Source: Authors’ tabulations of sample Form 1040 filings and the CPS for 2008

According to both sources, almost 30% of Social Security beneficiaries in the bottom income decile received some annuity/IRA income. In the second decile, the prevalence of reported annuity/IRA income among Social Security beneficiaries was 1.6 times higher in the tax files than in the CPS.

Moreover, across much of the income distribution, annuity/IRA income was significantly higher in the tax files than in the CPS — 30% to 40% higher in the third through sixth income deciles and nearly 50% higher in the seventh and eighth deciles. In the tax files, while lower-income retirees received more income from Social Security than from annuities and IRAs, the latter income was still significant. Even in the second to lowest income decile, retirees’ annuity/IRA income equaled half their Social Security benefits.

In the tax files, most filers who reported they were receiving Social Security income but no pension or IRA income had substantial earnings, including roughly 75% of those in the fourth through 10th deciles.

The data do not identify how many Social Security recipients with private retirement plans or savings are delaying private plan distributions until they need them. In fact, many healthy older people might prefer to work past retirement age and let their retirement savings continue to accumulate. In other words, the nonreceipt of private retirement benefits, especially among younger retirees, may reflect financial prudence rather than a failure of private retirement plans.

Reading the fine print

Reading the text accompanying the numbers in Social Security’s Income of the Population 55 or Older, 2010 begins to explain why so much income seems to be missing from the CPS. In the 2010 report, Brad Trenkamp writes:

We do not publish statistics differentiating between DB and DC pensions because a significant portion of payments from DC plans are not collected in the Current Population Survey. The Census Bureau only includes “regular payments” from retirement, survivor and disability income in its definition of total money income. Many people do not choose to annuitize their pension accounts and instead make withdrawals from their pension accounts on their own. These withdrawals are not part of total money income, and data are not collected on withdrawals from pension accounts in the March Supplement to the Current Population Survey.9

Analysts in the SSA unit that publishes Income of the Population 55 or Older are aware of the underreporting of retirement income on the CPS. Writing in the March 2012 Social Security Bulletin, Chris Anguelov, Howard Iams and Patrick Purcell conclude:

The shift toward greater distributions from DC plans and IRAs raises important questions about the accuracy of the CPS measures of the number of households that take such distributions and the proportion of household income derived from such accounts…The major nationally representative surveys of household income must accurately measure annual distributions from retirement accounts in order to provide a complete picture of the economic well-being of the aged and the general U.S. population.10

Despite these acknowledgements that the CPS and Social Security’s publication of summaries of its data do not include significant amounts of retirement income coming out of private retirement plans, many analysts continue to use the data files and published summaries as the true picture of how the U.S. retirement system now works.

Viewing retirement benefits from the wrong end of the telescope

The majority of retirees receiving annuities from employer-sponsored plans are those still covered by defined benefit plans. But many traditional pension plans have been converted into hybrid plans. In 2010, 37.3% of private pension plan participants were in hybrid plans, up from 20.5% in 2001, according to the Pension Benefit Guaranty Corporation.11

Hybrid plans generally offer lump sum payments to terminating workers, and evidence suggests that most of them opt for the lump sum rather than the annuity. Even in traditional defined benefit plans, at least a quarter of workers have a lump sum option at retirement,12 and some plan sponsors have recently allowed retirees already drawing an annuity to cash out the lump sum value of future benefits.

As measured by assets in the private, tax-qualified retirement system, defined benefit plans have lost a lot of ground to defined contribution plans and IRAs. Most of the money in IRAs, however, started off as benefits in defined contribution and defined benefit plans, which workers rolled over when they stopped working for the plan sponsor.13

Figure 6 shows accumulated asset balances in private defined benefit plans, defined contribution plans and IRAs separately, along with the sum of defined contribution and IRA balances for 1985 through 2010. In 1985, 54% of all private plan assets were in defined benefit plans, most of them traditional plans that paid out annuities at retirement. By 2010, only 20% of assets were in defined benefit plans and at least half the plan participants were eligible for lump sum cash-outs upon termination.

Figure 6. Assets in private tax-qualified retirement plans, 1985 – 2010 ($ billions)
Towers Watson Media

Source: Compiled by the authors from Federal Reserve Board, Flow of Funds data from various years found at

In IRAs, on the other hand, assets ballooned from 16% to 44% of all private retirement assets between 1985 and 2010. IRAs have emerged as the single largest component of today’s private retirement system, and most accumulations in employer-sponsored defined contribution plans and a considerable share of those in defined benefit plans are destined for IRAs when workers retire.14

The expanding role of defined contribution plans and IRAs in providing retirement income is indisputable, yet almost none of the income paid from IRAs is counted by the SSA in its assessments of income levels of the elderly. The shift is particularly consequential for the baby boomers, who are much more likely than earlier generations to have a defined contribution plan.

Even if we started counting dollars flowing out of IRAs more accurately, however, we could still miss some important pieces of the puzzle. Evidence suggests that withdrawals from IRAs and 401(k) accounts are strongly skewed by age. A 2006 study found that only 22% of 65- to 69-year-old account holders took such distributions compared with 72% of those 70 or older.15 The fact that relatively young retirees might not start drawing on accumulated private retirement resources until reaching their 70s does not mean that these resources are not adding to their retirement security.


There is room for concern about workers’ retirement preparations, and there is little doubt that the shift from defined benefit to defined contribution plans has made retirement more financially precarious for many. But the evidence that the majority of retirees rely mostly on Social Security benefits is dramatically underreporting the income provided by employer plans and IRAs. 

For 2008, distributions from annuities and IRAs to retirees reported in the tax files were vastly higher than comparable amounts reported on the CPS: $568.2 billion versus $227.8 billion — more than double and a difference of more than $340 billion! Moreover, some retirement benefits paid to low-income retirees do not show up in tax filings, suggesting that as little as one-third of the retirement benefits supplementing Social Security show up as retiree income in evaluations of current policies.

Concerns about the accuracy of self-reports of sources and levels of income on surveys that most people have never heard of might seem like some medieval theological debate over how many angels can fit on the head of a pin. But ignoring an ever-growing component of retirees’ private retirement savings misstates the role of Social Security in Americans’ retirement security. The CPS should either be revised to yield more accurate estimates of retirees’ income sources and amounts or be replaced. An effective U.S. retirement policy cannot be built on faulty measures of retirement income. 


1. See, for example, The National Commission on Fiscal Responsibility and Reform, Moment of Truth, at; and Bipartisan Policy Center, Debt Reduction Task Force, Restoring America’s Future, at

2. See, for example, Michael Lind, Steven Hill, Robert Hiltonsmith and Joshua Freedman, “Expanded Social Security: A Plan to Increase Retirement Security for All Americans,” Washington, D.C., New America Foundation, 2013; and Nari Rhee, “The Retirement Savings Crisis: Is It Worse Than We Think?” Washington, D.C., National Institute on Retirement Security, 2013.

3. Brad Trenkamp, Income of the Population 55 or Older, 2010, Social Security Administration (2012).

4. Center on Budget and Policy Priorities, “Top Ten Facts About Social Security,” found on July 16, 2013, at

5. Selena Caldera, “Social Security: Who’s Counting on It?” Washington, D.C.:  found on July 16, 2013 at AARP Public Policy Institute, found on July 16, 2013 at

6. Virginia P. Reno, Testimony before the Subcommittee on Social Security, Committee on Ways and Means, U.S. House of Representatives, found on July 15, 2010, at

7. Michael Lind et al., “Expanded Social Security: A Plan to Increase Retirement Security for All Americans,” pp. 3 –4, Washington, D.C., New America Foundation, 2013.

8. Sylvester J. Schieber, “Why Do Pension Benefits Seem So Small,” Benefits Quarterly (Fourth quarter, 1995), pp. 57 – 70.

9. Brad Trenkamp, Income of the Population 55 or Older, 2010, p. 14.

10. Chris E. Anguelov, Howard M. Iams and Patrick J. Purcell, “Shifting Income Sources of the Aged,” Social Security Bulletin (March 2012), vol. 72, no. 3, p. 67.

11. Pension Benefit Guaranty Corporation, Pension Data Book, 2011.

12. Bureau of Labor Statistics, U.S. Department of Labor, Employee Benefits Survey, found on August 2, 2013 at

13. See Craig Copeland, “Individual Account Retirement Plans: An Analysis of the 2010 Survey of Consumer Finances,” EBRI Issue Brief (September 2012), no. 375, for estimates that 44.5% of IRA assets are from rollovers, 44.1% from regular IRA accumulations and 11.4% from Roth IRAs. These balances are shifting relatively rapidly because rollover accumulations have been flowing into IRAs much faster than contributions in recent years. For example, the Employee Benefit Research Institute (EBRI) tracks a massive sample of IRA participants over time to evaluate their IRA participation. In 2011, they had records on 16.6 million unique individuals with $1.5 trillion in assets in their plans. Among their sample population for 2011, total contributions to IRAs were $5.96 billion compared with $78.57 billion in rollovers. This pattern has persisted for some time. See Craig Copeland, “Individual Retirement Account Balances, Contributions, and Rollovers, 2011: The EBRI IRA Database,” EBRI Issue Brief (May 2013), no. 386.

14. Craig Copeland at EBRI analyzed data for a representative sample of U.S. workers from the Survey of Income and Program Participation conducted by the U.S. Census Bureau. He estimated the percentage of individuals who had a retirement plan at a previous job and took a lump sum distribution versus leaving the assets in the prior employer’s plan. At every balance level, a significantly larger share took lump sums than left their money in the prior plan. For example, where balances were $5,000 to $9,999, 78.8% took a lump sum and 21.2% left the money in the prior plan. At $20,000 to $49,999, 69.5% took a lump sum and 30.5% left their assets in the prior plan. At balances of $50,000 and above, 63.2% took a lump sum and 36.8% left their balances behind. Most of the rollovers — 69.5% of the assets — were reinvested in IRAs. Subsequent employers received 16.4% of the assets. See Craig Copeland, “More Detail on Lump-Sum Distributions of Workers Who Have Left a Job, 2006,” EBRI Notes (July 2009), vol. 30, no. 7.

15. Chris E. Anguelov, Howard M. Iams and Patrick J. Purcell, “Shifting Income Sources of the Aged,” Social Security Bulletin, Vol. 72, No. 3, p. 65.