House Ways and Means Subcommittee on Social Security Hearing

Mr. Chairman and members of the Social Security Committee, thank you for inviting me here to testify before you today regarding the role of Social Security in providing retirement security to U.S. workers and retirees. This is an extremely important matter pertaining to our citizens' future physical wellbeing and peace of mind.

I am not representing any specific individuals or organizations regarding the matters I address in the following discussion. I have worked more than 40 years on both practical and policy issues related to retirement security. I began my career working for the Social Security Administration, spent some time as the research director at the Employee Benefit Research Institute and have spent 30 years in the consulting industry during which I have worked with many plan sponsors on the design and related issues pertaining to their retirement plans. I have focused on both Social Security and supplemental retirement provisions in my employment, service on various commissions, advisory councils and in extensive research and writing related to the U.S. retirement system.

Summary

There are many technical elements to determining whether workers are adequately preparing to meet their retirement income needs. How various analysts deal with these can play a significant role in their conclusions about the retirement income security prospects of workers and the current welfare of retirees.

. Retirement income replacement models are used widely in retirement plan design and communications for broad participant populations because generalized rule-of-thumb directions are the best vehicles available to plan sponsors and administrators.

. More microscopic lifecycle and similar models are beyond most workers' reach but can help policymakers and analysts understand whether the cruder measures are effectively helping workers achieve reasonable retirement savings goals.

. Social Security earnings replacement rates presented by the Trustees are not equivalent to conventional replacement rate measures and may be confusing to many participants and analysts.

. Conventional measures of Social Security replacement of earnings result in replacement rates 15 to 20 percent higher than those presented by the Trustees for full-career workers retiring at normal retirement age (see Table 1).

. Some analysts even apply wage indexing of all preretirement income in defining retirement income targets suggesting that workers should increase saving beyond what is needed to maintain preretirement living standards in retirement with the implication that normal working-period living standards would be reduced in order to finance much higher living standards after retirement. This problem may be accentuated over time (see Table 2 with accompanying discussion).

. The Social Security Administration regularly publishes summary survey data gathered by the U.S. Bureau of the Census in its Current Population Survey each year shows that employer-sponsored retirement plans and individual retirement savings make little contribution to the retirement security of most of the elderly.

. Comparison of the reported income provided by pension/annuity plans and IRAs to Social Security from IRS tax filings proves that as much as 60 percent of the pension and IRA income paid to Social Security beneficiaries is not being captured by the Current Population Survey (see Table 3) a problem the Census Bureau and the Social Security Administration has been aware of for some time.

. It is impossible to clearly understand who is doing well and who is doing poorly under the current arrangements if official government reports on the income status of retirees ignore hundreds of billions of dollars of their income.

Retirement Income Adequacy

Over the years, retirement researchers, policy analysts and retirement plan designers have come to think of an adequate retirement income as the level of income needed to allow retirees to maintain their preretirement standard of living. For a small segment of the elderly population, even this level of income is considered to be inadequate. Some workers live on an income for extended periods considered to be less than necessary to sustain even a basic living. For these people, there are assistance programs that help with needed food purchases, housing, health care and the like. For such individuals, it is likely that assistance outside of the benefits provided by contributory pensions and retirement savings is necessary to support them in retirement. For the remainder, an adequate retirement income is generally considered to be one that allows them to maintain their career living standards.

Economists often focus on whether workers are going to accumulate adequate resources to meet their retirement needs in the context of a life cycle model. The adaptations of these models are highly variable but can allow for borrowing early in life for education, as households are established, children are born, and so forth. As the career progresses, debts are paid off and savings for retirement accumulated. Then in retirement, the combination of Social Security, defined benefit pensions and retirement savings finance consumption. These models can account for expenses that are unique to both the working and retirement periods. Some analysts who use this approach to assess whether workers are saving properly take into account whether households have children or not during the working period. Evidence suggests that such households have elevated consumption while children are present and higher savings rates after the children move out. For workers, wages are subject to payroll taxes while retirement benefits are not and income taxes may be higher during the working period. Retirees have risks of health and long-term care expenses that can be higher or more variable than incurred when younger.

Retirement planners and consultants who design retirement plans for individuals and employers generally rely on an alternative model where preretirement spendable income is determined by subtracting work-period taxes and work-related expenses from gross earnings. To maintain the preretirement standard of living, a retiree's spendable income level post career must equal the level achieved before retirement. This spendable income is divided by gross earnings to calculate a target earnings "replacement rate" that will fulfill the goal of leveling consumption levels over the pre- and post-retirement periods.

The Retirement Security Projection Model developed by Jack VanDerhei and Craig Copeland at the Employee Benefit Research Institute is not a lifecycle model but it has many stochastic features that are similar to those found in lifecycle models. In their modeling, VanDerhei and Copeland estimate retirees consumption needs and then assess whether the combination of annuity pensions including Social Security and employer-sponsored plans plus other assets will be sufficient to cover the expected consumption needs of workers at various income levels. My personal view is that VanDerhei and Copeland have made particularly valuable contributions by more realistically modeling a much broader spectrum of uncertainties that workers face in preparing for retirement than any of the other models widely reported.

Each of the approaches to considering whether workers are preparing adequately for retirement or retirees have adequate income has strengths and weaknesses. The Retirement Security Projection Model and the lifecycle model recognize that every household faces a unique set of circumstances as it progresses through its employment and retirement periods. The amount that each household should save, what its consumption should be in the preretirement period and during retirement are unique to that household. The replacement rate models are typically applied on a generalized basis and are often estimated on an individual worker basis instead of at the household level. For example, it is common for the consulting firms that help employers sponsoring plans with plan designs and assessments to estimate target earnings replacement rates for workers at various specified earnings levels. In developing these models, the analysts project earnings and retirement benefit accumulations to retirement age, estimate taxes unique to the working period and other work-related expenses to estimate spendable income which is then used to estimate the replacement of gross earnings required to maintain living standards in retirement. The adequacy of workers' retirement preparation is assessed by determining whether the combination of Social Security and retirement plan benefits will be sufficient to allow retirees to maintain spendable income during retirement.

There are a number of methodological problems with each type of model that are relatively technical and which are beyond the scope of the current discussion. n1 Beyond these, the problems with the models that focus on individual household units are that the assessments of their retirement income preparations result in a one-size-fits-one answer to whether workers are adequately preparing for retirement whereas the replacement rate assessments result in a one- size-fits-many answer. The Retirement Security Projection Model and lifecycle models are superior to the replacement rate model in assessing how individual households are doing but are of little practical use to a large sponsor of a retirement plan with many participants where the sponsor has limited information beyond the participants' own earnings levels, age, expected retirement age and other salient information. The structuring of generalized retirement plans fitting large populations or the generalized educational support that goes along with them has to utilize generalized rules-of-thumb to help workers assess whether they are accumulating the needed resources to meet their retirement needs. This is true for Social Security as well as for employer-sponsored retirement plans. The more microscopic analysis used in lifecycle model assessments can help policymakers and other interested parties understand whether the cruder measures used by plan sponsors and advisors are effectively helping workers along the way.

Career Income and Replacement Rate Measurement Issues

Because Social Security is a national system that applies to millions of workers at any point in time, the Trustees regularly include estimates of the extent to which the program's retirement benefits will replace preretirement earnings. The 2013 Trustees Report estimated that a worker with career-average medium earnings who retired at his or her normal retirement age in 2013 received benefits equal to 44.5 percent of "career average earnings, indexed by national average wage growth to the year prior to retirement." n2 Despite the fact that the replacement rates are calculated using the average of the high-35 years of wage-indexed earnings based on the average wage index in which the worker turned age 64, the benefits in the table presenting the rates are shown in CPI-indexed dollars. Regarding the replacement rates that are presented in the annual report, the trustees noted, "this method of calculation produces percentages that may differ significantly from those that would be produced by comparing benefits to these representative workers' recent average earnings levels or to other more common measures of pre-retirement income." n3 There is no explanation in the report on how or why the replacement rates presented by the trustees "may differ significantly" from "more common measures" of such rates. The differences are significant.

The "more common" measures of earnings replacement rates that the Social Security trustees refer to in the 2013 report are the measures typically used by employer retirement plan designers and sponsors and by retirement advisors and those used by lifecycle modelers focusing on retirement accumulations and provisions. Employers sponsoring retirement plans and their advisors virtually always consider the earnings replacement of their retirement plans on the basis of projected final earnings of the workers participating in their plans. In some instances this will be the annual wage or earnings level at the time of retirement and in others an average of the last three or five years of full earnings before retirement. But the earnings projections that they use in estimating replacement rates generally assume relatively stable wage growth trajectories over workers' careers. n4 Among representative samples of retirees receiving Social Security benefits, however, covered earnings records show that typically even nominal wages tend to decline, on average, as workers approach retirement n5 which suggests that earnings levels immediately before retirement may not be the best indicator of the standards of living workers have achieved during their careers. An alternative measure that both Scholz and Seshadri n6 and Pang and Schieber conclude is a better indicator of peak earnings late in the career is the average of fifth through ninth years of positive earnings prior to retirement indexed by the CPI. The lifecycle model premises that households base their consumption over their lifetimes on the basis of their real earnings over the full period of retirement and often use the average of inflation indexed earnings during a household's working years as the appropriate measure.

Pang and Schieber have compared the method for computing the replacement rates presented by the trustees with some of the more common measures of replacement rates used by retirement plan designers, counselors and many researchers using a sample of Social Security beneficiaries receiving benefits at the end of 2004. n7 To look at how the alternative measures of earnings might affect replacement rates for real workers, they used a sample of beneficiaries and their earnings records from the Benefits and Earnings Public-Use File, 2004, released by the Social Security Administration in 2005. This file contains information about beneficiaries of the OASDI program from Social Security's Master Beneficiary Record (MBR) file. n8

In order to compare alternative measures of earnings replacement provided by Social Security to real workers, Pang and Schieber focused on individuals receiving "retired worker" benefits from the 1931 through the 1942 birth cohorts. For these cohorts, the historical earnings records covered retirees at least from age 20 through their retirement. The average indexed monthly earnings (AIME) on which each person's benefits were based was indexed to 2004 using the national average wage index to put them on a consistent basis. Then the sample of retired workers was distributed into 10 equal sized groups (deciles) based on their AIME ranking. In the segment of the analysis focused on the earnings replacement capacity for workers roughly matching the trustee's hypothetical workers, long-career workers with at least 35 years of covered earnings were included which comprised 64 percent of all beneficiaries receiving retired benefits at the end of 2004. n9 The remainder had shorter periods of covered employment.

Pang and Schieber calculated replacement rates two ways for comparative purposes in the analysis. In the first, they used actual benefits at retirement compared to a range of alternative measures of preretirement earnings. In the second, they recalculated benefits that would have been paid at normal retirement age assuming the long-career workers had not taken benefits until then. These latter calculations might have resulted in slightly lower benefits in some cases than retirees would have received had they deferred retirement until then but it is likely the differences are minimal since all the retirees considered had at least 35 years of covered earnings when they actually retired. Because the trustees' hypothetical workers' replacement rates are shown at age 65 and normal retirement age, for comparison purposes, only the replacement rate calculations for retirements at normal retirement age are considered in the following discussion.

Table 1 shows four sets of replacement rates each based on a different definition of pre-retirement earnings. The rates are shown for each of the AIME decile groupings described earlier. The first column of replacement rates was developed using the method used in deriving the rates shown by the trustees in their reports, namely, the rates are calculated using the average of the high-35 years of wage-indexed earnings based on the average national wage index in which the worker turned age 64. The second column of replacement rates is based on the benefit paid at retirement divided by the average of the nominal wages paid the workers in their last five years of positive earnings. The third column of rates is based on benefits divided by the average of fifth through ninth year's positive earnings prior to retirement indexed for price inflation. While this measure does not literally correspond with the measures of final earnings that retirement plan sponsors and their advisors typically consider in estimating replacement rates, I believe this is a better implicit proxy for their final-salary measures of replacement rates when applied to a general population like that of Social Security participants and beneficiaries. The fourth column is the average of the high-35 years of price indexed earnings over the workers' careers reflecting the average real lifetime earnings of the worker that is often used in lifecycle analyses. I believe this is the best measure of the average earnings capacity over workers' lifetimes that can be used for leveling consumption in the lifecycle context.

Table 1: Median Re-estimated Social Security Replacement Rates for Beneficiaries at the End of 2004 Who Had Long Careers Assuming They Had Been Paid Normal-Retirement-Age Benefits

Measures of preretirement income used to estimate replacement rates

AIME deciles SS Trustees Age-64 based AIME Final 5-year average Average of years 5-9 pre-retirement CPI-W indexed Hi-35 CPI-W indexed average

1 85.8 115.5 105.3 103.7

2 71.0 101.4 93.2 86.0

3 57.6 82.9 71.2 69.8

4 50.6 70.1 59.7 60.6

5 46.0 64.4 54.2 54.9

6 42.9 60.9 50.7 50.9

7 40.6 58.6 47.9 48.1

8 38.3 52.9 42.4 45.2

9 34.8 43.3 34.8 41.1

10 31.7 31.3 29.4 36.6

Source: Gaobo Pang and Sylvester J. Schieber, "Understanding Social Security's Income Replacement Measurements," Social Science Research Network (2014) available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2433181, Table 8.

Considering the measures derived using the Social Security Trustees' method shown in Table 1, the median of the distribution of replacement rates would be the average of the rates shown for the fifth and sixth deciles or 44.5--(46.0 + 42.9) = 2 = 44.45.--almost precisely equal to the rate shown for the medium worker in the 2013 Trustees Report. The median for the distribution of replacement rates using the average of the final-five years of positive earnings is 62.7 percent, nearly 20 percentage points higher than the trustees' measure. The medians for the average of real earnings in years five-to-nine of positive earnings prior to retirement is 52.5 and that for the average of real career earnings is 52.9 percent. The latter two medians are about 15 to 20 percent greater than the average posited using the trustees' estimates.

According to the Social Security Administration's official website, "Most financial advisors say you will need about 70 percent of your pre-retirement earnings to comfortably maintain your pre-retirement standard of living. If you have average earnings, your Social Security retirement benefits will replace only about 40 percent. The percentage is lower for people in the upper income brackets and higher for people with low incomes. You'll need to supplement your benefits with a pension, savings or investments." n10 Some retirement plan designers and advisors suggest that workers need to shoot for an 85 replacement of preretirement earnings n11 or even higher. Whatever the target deemed appropriate, the challenge to workers in accruing adequate resources to maintain living standards is one of filling the gap not covered by Social Security. In order for them to do so effectively, it is important that they have a reasonable understanding of the gap that they need to fill. If the total target that an average earner is 70 percent of preretirement earnings and Social Security covers 52 percent of real earnings replacement, then the worker must come up with resources to cover the other 18 percent. However, if Social Security only covers 44 percent of the earnings replacement target, then the worker has to come up with the residual 26 percent--44 percent more than the balance needed to replace the real average career earnings. For families trying to act responsibly regarding saving for their retirement security, the difference in what is commonly thought to be reasonable savings patterns and what is implied by the Social Security Trustees is not trivial.

Some research and policy analysts take the Social Security Trustees' approach to estimating preretirement earnings replacement rates and even apply it to the supplemental savings targets that workers should be aiming to fill. For example, Boston College's Center for Retirement Research wage-indexes all preretirement income, including asset income, in calculating replacement rate targets in developing the widely publicized National Retirement Risk Index. The underlying model used to develop this risk index is supposedly based on the lifecycle model of consumption and saving used by many economists. n12 I believe that workers saving to meet the targets devised by wage indexing all preretirement income in this fashion almost certainly would be acting irrationally in the context of the lifecycle model.

Consider a worker at the beginning of her career who anticipated earnings that will average $50,000 per year in real terms over her working life. Assume she calculates that her combined payroll tax payments and savings should be $12,500 per year and that these will finance a retirement income of $37,500 per year all in real terms--75 percent of her preretirement earnings. Further assume that if she realized economy-wage growth, her wage-indexed career-average earnings would be $60,000 per year. If she aspires to having a total retirement income based on the $60,000 average, she will have to increase her retirement contributions to $15,000 per year. This would reduce her working-consumption levels to $35,000 per year to finance retirement consumption of $45,000 per year. For her to save for retirement as though her wages grow more rapidly than they do in real terms will result in an irrational outcome in the lifecycle context.

The practical situation that workers face during their working lives is that when they go to the grocery store, buy clothes, purchase a house or get their car repaired they have to spend real dollars not wage-indexed dollars. The real dollars they earn while working define the standard of living they can achieve prior to retiring unless they are getting welfare transfers which most career workers do not. Why their historically wage-indexed earnings levels should define their consumption targets in retirement is a question not explained by either the Social Security Trustees or the authors of models like the National Retirement Risk Index.

The situation of the hypothetical worker in the example would likely be exacerbated across time if workers are expected to maintain reasonable steady replacement rates relative to wage-indexed average earnings. Using the data files that Pang and I used in our analysis cited earlier, we compared the rates of growth of wage-indexed and real wages over the high-35 year career average earnings for Social Security beneficiaries who were from the 1931 and 1938 birth cohorts. The results of the exercise are shown in Table 2.

Clearly, the wage-indexed career average earnings were growing much more rapidly than workers' real earnings capacity across much of the earnings spectrum. In the top two deciles the opposite was true, but it was here, especially the top decile, where real wage growth was dragging up the national average wage index benefitting workers further down the distribution who were benefitting from wage growth (for which they were not accountable) used to determine their Social Security benefits. I am not disputing the current method of calculating Social Security benefits here, but I do believe that wage indexing total income for purposes of setting retirement replacement rate targets will lead to the same irrational result for the majority of workers as the example above showed it did for the exemplary hypothetical worker. If workers are going to be expected to save on their own so they can keep up with wage growth across the economy at retirement from one birth cohort to the next, each successive cohort will be expected to give up an increasing portion of their career earnings so they can live a higher lifestyle in retirement than they could achieve while working.

Table 2: Percent Increase in the Average of High-35 Years of Earnings Wage Indexed with the National Average Wage Index at Age 60 and CPI-W Indexed to Age 64 for the 1931 to the 1938 Birth Cohorts

AIME Wage-indexed Price-indexed

deciles AIME AIME

1 14.4% 3.9%

2 27.4 16.1

3 26.8 15.9

4 34.3 23.0

5 33.1 22.0

6 32.3 21.4

7 30.4 20.0

8 31.4 21.1

9 4.0 20.5

10 5.2 31.5

Source: Derived by the author as described in the text.

Retirement Savings and Retirement Income

Today there is a somewhat pervasive perception that we face a "retirement crisis" in the United States. One reason is the notion that a significant majority of retirees receive most if not all of their retirement benefits from Social Security and that supplemental plans make little contribution to the retirement security of most of the elderly. A major piece of evidence pointing to this conclusion is widely used estimates published by the Social Security Administration (SSA) of the prevalence and level of benefits provided to the retiree population based on a periodic Census Bureau Survey, the Current Population Survey (CPS)

According to SSA, among families with a person 65 years of age or older in 2010, 39.7 percent of households received regular income or annuities from a retirement plan other than Social Security. Among all families with an elderly person, public pensions accounted for 5.5 percent of total money income and private pensions 7.6 percent. By comparison, Social Security benefits comprised 58.5 percent of the elderly's income. For the elderly receiving retirement benefits, they were a much more important component of retirement income than the population averages suggest. Those receiving public pensions received 44.5 percent of their total income from their retirement plan; for those receiving private pensions it was 28.4 percent; and for those receiving Social Security benefits, it accounted for 66.0 percent of their income. n13

In order to assess the reporting of pension and other retirement income on the CPS, Miller and Schieber compared the survey results with data from a representative sample of federal income-tax filers for whom the IRS releases a data file each year. n14 They assumed that federal income-tax filings were likely less prone to reporting errors than survey responses. Miller and Schieber compared various components of reported income from the IRS tax files for 1988, 2000, and 2008, the latter being the most recent year available, with the same income elements from the CPS Income Supplement files for the same years.

According to the tax files, in 2008 an estimated 23.4 million filing units reported Social Security income. According to the CPS, 32.9 million family units received Social Security. The number of units receiving Social Security was higher in the CPS than in the tax files because the IRS tracks a population subset, tax filers, while the CPS supposedly represents the entire civilian, non-institutionalized population.

The IRS files do not include indicators of filers' ages so Miller and Schieber used the reporting of Social Security income on both the CPS and tax filings as an indicator that someone in a filing unit was retired either due to age or disability although they could not distinguish for which reason in the IRS files. They compared the reporting of pension/annuity or retirement saving income received by filing units from the two sample files for the various years they had data from both sources. The results are shown in Table 3.

The CPS captures very little of the IRA income being reported on tax filings: 3.5 percent in 2000 and roughly 6 percent in 2008. The CPS has done a better job of capturing reported pension and annuity income but still comes up short by around half, and the underreporting appears to have worsened considerably in recent years. In aggregate, the CPS for 2008 captured only 40 percent of the pension/annuity or IRA income received by family units that were receiving Social Security benefits. Considering that the tax filings do not capture all income paid by employer-sponsored retirement plans and IRAs--because low-income recipients do not have to file tax forms and Roth-type benefits are not taxable--aggregate benefits from these plans likely exceeded the Social Security benefits retirees received in 2008. This is a far cry from the relative levels of retirement income reported by Social Security as noted above.

Recognizing that retirees at the bottom of the income distribution were not included, n15 Miller and Schieber used the 2008 tax files to evaluate the distribution of income derived from pensions and retirement savings. They split the tax-filing units into deciles (ten equal groups) based on total income. Both the prevalence and level of benefits were quite low in the bottom decile. By the second decile, 60 percent of the filing units reported receiving pension income or income from an IRA or similar individual retirement savings plan. For those receiving the benefits, at the median their pension/IRA income equaled 50 percent of the median Social Security benefits, a nontrivial supplement to their retirement income. Above the second decile, three quarters of filing units or more reported annuity or individual account income and from the fifth decile and above, the median income levels from the retirement plans consistently exceeded median Social Security benefits.

Miller and Schieber also found that among tax filers reporting Social Security income but no other retirement income that 69 percent overall and 74 percent of the top 80 percent of tax filers by income were still reporting wage or salary income in 2008. For many of these people, the lack of reported pension or retirement saving income may be more of an indication of prudence in retirement planning than that they will not benefit from supplemental retirement benefits to their Social Security once they fully retire.

Table 3: Reported Pension/Annuity Income and Distributions from IRAs on Federal Income Tax Forms and the CPS for Social Security Beneficiaries for Selected Years

IRA distributions Pensions/annuities Total benefits (Dollar amounts in millions) 1988 IRS$4,788$74,376$79,164

1988 CPS $597$79,683$80,280

CPS/IRS 12.5% 107.1% 101.4%

2000 IRS$59,358$258,764$318,122

2000 CPS $2,083$143,909$145,992

CPS/IRS 3.5% 55.6% 45.9%

2008 IRS$110,920$457,311$568,231

2008 CPS $5,564$222,248$227,812

CPS/IRS 5.0% 48.6% 40.1%

Source: Billie Jean Miller and Sylvester J. Schieber, "Contribution of Pension and Retirement Savings to Retirement Income Security: More than Meets the Eye," Journal of Retirement (Winter 2014), vol. 1, no. 2, p. 19.

A major reason the CPS does such a poor job reporting IRA and other individual retirement account distributions is because the Census Bureau does not believe that distributions that are not paid on a regular periodic basis should be counted as income. The idea that retirees who only tap these funds as needed or as required by law are not receiving economic benefits is inconsistent with virtually all economic modeling of retirement preparedness which routinely considers accumulated wealth as a source of economic security without regard to how people choose to distribute it in retirement.

There are legitimate reasons to believe that some workers are not adequately preparing for retirement and that some retirees have inadequate resources to meet their economic needs, and the shift from defined benefit to defined contribution plans is widely considered to have made retirement more financially precarious for workers and retirees. It is impossible, however, to clearly sort out who is doing well and who is doing poorly under the current arrangements if official government reports on the income status of retirees ignore hundreds of billions of dollars of their income.

n1 Many of the methodological problems with the models are explored in Gaobo Pang and Sylvester J. Schieber, "Why American Workers' Retirement Income security Prospects Look So Bleak: A Review of Recent Assessments," Journal of Retirement (Summer 2014), vol. 2, no. 1, forthcoming

n2 The 2013 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, p. 145.

n3 Ibid., p. 143.

n4 For example, see Aon Hewitt, 2012, "The Real Deal: 2012 Retirement Income Adequacy at Large Companies," found on October 29, 2013 at: http://www.aon.com/human-capital-consulting/thought-leadership/retirement/survey_2012_the-real-deal.jsp.

n5 Gaobo Pang and Sylvester J. Schieber, "Understanding Social Security's Income Replacement Measurements," Social Science Research Network (2014) available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2433181.

n6 Scholz, John Karl and Ananth Seshadri, 2009b, "What Replacement Rates Should Households Use?" University of Michigan Retirement Research Center, Working Paper WP 2009-24.

n7 Pang and Schieber, "Understanding Social Security's Income Replacement Measurements."

n8 Social Security Administration, "Benefits and Earnings Public-Use File, 2004," (October 2005), found at: http://www.ssa.gov/policy/docs/microdata/earn/.

n9 Pang and Schieber also analyzed beneficiaries with shorter covered careers in their analysis not included in the current discussion because they do not align with the trustees' hypothetical workers' career profiles.</p>

n10 Social Security Administration, "Retirement Planner: Learn about Social Security Programs," found on July 23, 2014 at: http://www.ssa.gov/retire2/r&m6.htm.

n11 Fidelity Investments, 2013, "Fidelity Outlines Age-Based Savings Guidelines to Help Workers Stay on Track for Retirement," found on August 7, 2013 at: http://www.fidelity.com/inside-fidelity/employer-services/age-based-savings-guidelines and Aon Hewitt, 2012, "The Real Deal: 2012 Retirement Income Adequacy at Large Companies," found on October 29, 2013 at: http://www.aon.com/human-capital-consulting/thought-leadership/retirement/survey_2012_the-real-deal.jsp.

n12 Center for Retirement Research at Boston College, 2006, "Retirement at Risk: A New National Retirement Risk Index," found at: http://crr.bc.edu/wp-content/uploads/2011/09/NRRI1.pdf and Munnell, Alicia H., Anthony Webb and Francesca Golub-Sass, 2012, "The National Retirement Index: An Update," Issue Brief, Boston: Center for Retirement Research at Boston College, no. 12-20.

n13 Trenkamp, Brad, Income of the Population 55 or Older, 2010, Social Security Administration (2012), found at: www.socialsecurity.gov/policy/docs/statcomps/income_pop55/2010/incpop10.pdf, pp. 50, and 235-236.

n14 Billie Jean Miller and Sylvester J. Schieber, "Contribution of Pension and Retirement Savings to Retirement Income Security: More than Meets the Eye," Journal of Retirement (Winter 2014), vol. 1, no. 2, pp. 14-29.

n15 The number of income-tax filers reporting Social Security income in 2008 was about 25 percent smaller than the number of households reporting they received Social Security benefits in that year.

Read this original document at: http://waysandmeans.house.gov/UploadedFiles/7_29_14_SS_Testimony_Sylvester_Schieber.pdf