PDFpdf_small | Flashflash_small | Watch the video

“The Importance of Social Security for Sustaining Living Standards in Retirement”

Testimony of
Dean Baker, Co-Director
Center for Economic and Policy Research

At the hearing on

The Role of Social Security, Defined Benefits, and Private Retirement Accounts in the Face of the Retirement Crisis

Before the
Subcommittee on Social Security, Pensions, and Family Policy of the
U.S. Senate Committee on Finance

December 18, 2013

I’d like to thank Chairman Brown and Ranking Member Toomey for the opportunity to speak to the subcommittee today about Social Security. I will make three main points in my testimony:

  1. Social Security is the main source of income for most of the senior population. It keeps most seniors out of poverty and ensures them at least a minimal standard of living in retirement.

  2. Social Security is likely to be more important to seniors in the decades ahead as a result of the disappearance of the traditional pensions and the inadequacy of 401(k)-type accounts as replacements. Due to the structure of Social Security, most seniors may expect to share in at least some portion of the benefits of economic growth even if the pattern of wage stagnation over the last three decades persists.

  3. The proposal to change the basis for the indexation of benefits to the CPI-E is consistent with the intent of Congress when it decided to index benefits to inflation in 1975. The goal was presumably to ensure that the purchasing power of benefits kept pace with the inflation rate experienced by seniors. An index that tracks the actual buying patterns of seniors will come closest to meeting this goal.

The Importance of Social Security to Seniors

Social Security comprises a large and growing share of the income of the senior population. In 1962 Social Security accounted for 30 percent of the income of households including someone over age 65.[1] This share was up to 36 percent in 2011. Since a substantial portion of the over-65 population is still working, the share of non-wage income is considerably larger, at 52.2 percent in 2011 compared with 41.7 percent in 1962.

Thirty five percent of senior households rely on Social Security for more than 90 percent of their income, while almost two-thirds of senior households rely it on for more than half of their income. The dependence on Social Security is even greater for unmarried seniors, most of whom are women. It accounts for more than 90 percent of the income for 45 percent of unmarried seniors and more than half of the income of 74 percent of unmarried seniors.  

The payback structure is highly progressive, so there is not a large difference between the maximum individual benefit of $2,530 a month and the average benefit of $1,300. (This is one reason that means testing would matter little to the finances of the program. The small number of wealthy seniors do not get much more in benefits than a typical beneficiary, so there is little money that can potentially be saved by reducing or eliminating their benefits.) Social Security is the main reason that poverty rates for the elderly are now somewhat below the average for the adult population as a whole, 9.1 percent for seniors compared to 13.7 percent for the population age 18-64.[2] (The gap is smaller when one looks to the Census Bureau’s Supplemental Poverty Measure, which is generally viewed as the more accurate measure of poverty. This measure shows a poverty rate of 14.8 percent for seniors compared to 15.5 percent for the aged 18-64 population.[3]) The poverty rate for seniors by either measure would be over 50 percent without income from Social Security.

Poverty rates for senior non-married women and minorities are substantially higher than the overall rate. For senior non-married women the poverty rate is 16.3 percent by the official measure, with another 11.0 percent near poor, for a total of 27.3 percent who are either poor or near poor.[4] In the case of African American seniors, 17.3 fall below the poverty level with another11.0 percent near poor. For older Hispanics the poverty rate is 18.7 percent, with another 9.5 percent near poor. For the poor and near poor populations, Social Security would provide the overwhelming majority of their income.

Clearly Social Security is a hugely important source of income for seniors. It is most important for moderate and low-income seniors, but the percentage of seniors who could get by without a Social Security check is small. In 2009, 18.5 percent of benefits went to people with non-Social Security income of more than $30,000 a year (roughly $32,000 in today’s dollars) and less than 10 percent went to people with non-Social Security income of more than $50,000 a year.[5] In short, the vast majority of Social Security benefits go to people who need it to maintain their standard of living.

The Growing Importance of Social Security through Time

As important as Social Security is for the current population of retirees, it is projected to become even more important in future decades for middle income households, primarily due to the collapse of traditional defined-benefit pensions. While future retirees are projected to receive somewhat more income from defined-contribution retirement accounts than had been the case in prior decades, this will not come close to making up for the loss of traditional pensions.

The Social Security Administration’s projections from their MINT model show that for households in the middle quintile at age 67, Social Security benefits will rise from 34.2 percent of income for the cohorts born between 1936-1945 to 37.0 percent of income for the cohorts born between 1966-1975, as shown in Table 1.[6] Since a much larger share of this younger cohort is projected to still be receiving wage income at age 67, the rise in non-wage income is considerably more striking, from 41.9 percent to 48.6 percent. Pulling out imputed rent, Social Security is projected to account for 54.8 percent of non-wage income at age 67 for the middle-income quintile in the cohort born between 1966-1975, up from 46.4 percent for the cohort born between 1936-1945.[7]

Table 1: Projected Shares of Income at Age 67

Income quintile

Social Security as a share of

Cohort birth years




Total income



Non-wage income



Non-wage and non-rent income




Total income



Non-wage income



Non-wage and non-rent income




Total income



Non-wage income



Non-wage and non-rent income



Source: Butrica et al. Table 5.


The main reason for the growing importance of Social Security for this middle cohort is the virtual collapse in income from traditional pensions. These are projected to fall from 25.0 percent of non-wage, non-rent income to just 3.2 percent of non-wage, non-rent income over this period. Defined-contribution retirement accounts are projected to make up slightly more than half of this shortfall, rising from 10.7 percent of non-wage, non-rent income to 22.6 percent of non-wage, non-rent income.  

Social Security is projected to continue to account for the bulk of the income for households in the bottom-income quintile. There is a small decline in its projected importance for households in the bottom quintile, primarily due to the growing importance of income from assets, retirement account income and imputed rental income. Social Security is projected to be a declining share of the income for households in the top quintile. This is primarily the result of a sharp projected increase in the income from assets that this group will receive.

The projected rise in the importance of Social Security for the middle-income workers points to both the need to sustain the value of benefits to ensure retirement security for this group and also the need to establish a system of supplemental retirement income to replace defined-benefit pensions. It is worth noting in this respect that the income for this middle quintile at age 67 is projected to rise from $36,000 per person for 1936-1945 cohorts to $44,000 for the 1966-1975 cohorts, an increase of 21.1 percent (both figures are in 2011 dollars). As noted earlier, much of this increase is attributable to the fact that people in the younger cohort are more likely to be working at age 67, but after even pulling out earnings there would still be an increase from $31,000 to $35,000 per person, a rise of 12.9 percent. This is entirely attributable to a projected rise in Social Security benefits.

This brings up an interesting aspect of the Social Security benefit structure. Even though the Social Security Administration’s MINT model projects the wage stagnation of the last three decades to continue, as there is a further upward redistribution of wage income, the benefit structure would still lead to an increase in the size of real benefits. This is the result of the progressive payback structure for benefits and the fact that initial benefits are indexed to average real wages.

Table 2 shows benefits, measured in 2013 dollars, for workers with average real earnings of $10,000 and $30,000. These figures are not wage indexed, so the assumption is that a worker had the same real wages over his or her lifetime. The columns show the retirement benefits he or she would receive with this earning history in 2002 and in 2032.

Table 2: Social Security Benefits for Workers Retiring at the Normal Retirement Age

(2013 dollars)

Average Lifetime Earnings

Benefits in 2002

Benefits in 2032







Source: Author's calculations

As can be seen, in both cases the benefit is higher in 2032 than 2002. In the case of the worker with average earnings of $10,000, the increase would be 5.4 percent. In the case of a worker with average earnings of $30,000, the increase would be 11.7 percent. The reason for the increase is that the bend points[8] for the payback are indexed to average wages (which rise faster than median wages). If workers do not have increases that rise in step with the average wage, then a larger share of their wage income will fall below the bend points, and therefore be subject to a higher payback rate. The benefits of the hypothetic $10,000 lifetime average earner are less than that of the $30,000 lifetime average earner because by 2032 she would already have all her income below the lower bend point. This means that she cannot benefit from further growth in the average wage.

However, for a $30,000 lifetime earner (which would be close to the actual median), this payback structure helps to ensure that he will get some of the gains from economic growth, even if he has been excluded from these gains during his working years. This is an interesting and arguably desirable feature of the Social Security payback structure. (As a practical matter, it is important to note that the age for receiving full benefits will have risen from 65 to 67 over this period, which is equivalent to a 12 percent cut in benefits.)

There are a couple of points worth noting from the SSA projections and this arithmetic exercise. First, the benefits for low-end earners are not likely to rise in step with economic growth because most of them will have lifetime earnings that already place them below the first bend point. (The projections show an average benefit for the cohort born between 1966 and 1975 of just $8,000.) If they are to see any gains from economic growth over this period, then it would be necessary to increase the payback rate.

Fortunately it is not expensive to have modest increases in benefits for those in the bottom portion of the income distribution. For example, the Institute for Women’s Policy Research and the National Committee to Preserve Social Security and Medicare Foundation have proposed an increase in the basic benefit of 5 percent, or roughly $55 a month.[9] If this increase was phased out near or somewhat above the average benefit, then it would increase costs by less than 0.3 percent of payroll. Increasing the benefit for surviving spouses to 75 percent of the joint benefit (which also can be capped at the average wage) is another relatively low-cost way to improve retirement security.

The other point is that is clear from these projections is that a new tier to the retirement system is necessary. The ideal would be a voluntary system that would be fully portable between jobs and have low administrative costs. The federal Thrift Savings Plan is an excellent model. It also could be possible to piggyback on state employee retirement systems to allow an experiment on a smaller scale. The idea would be that workers would have the ability to contribute to the system regardless of who their employers are. There would be limited choices of investment options, with a default of a diversified fund, in order to keep the plan simple and to minimize costs. Annuitization would be the default option at retirement. Such proposals have received bipartisan support at both the federal and state levels. Unfortunately, the crisis created by the collapse of the housing bubble has obstructed progress on this front.     

The Indexation of Benefits Post-Retirement

The last point that I wanted to briefly address was the indexation formula used to adjust benefits after workers retire. Currently benefits are tied to the rate of inflation as shown by the consumer price index for wage and clerical workers (CPI-W), which was the only CPI in existence at the time that Congress first voted to index benefits in 1975. Since then the Bureau of Labor Statistics (BLS) began fielding the CPI-U, which tracks the rate of price increase for the consumption basket of a typical urban household. More recently, BLS has constructed a research index, the chained CPI (C-CPI-U), which incorporates substitutions in purchases of goods and services in its measure of inflation.

Since people typically substitute from items whose prices are rising rapidly to items whose prices are rising less rapidly, the C-CPI-U has shown a rate of inflation that averages 0.2-0.3 percentage points less than the rate of inflation shown by the CPI-U or the CPI-W. Based on this difference, many political figures have advocated switching the indexation of Social Security benefits from the CPI-W to the C-CPI-U. The Congressional Budget Office has projected that this switch would save more than $120 billion over a decade as a result of lower benefit payments.

While there is little doubt that a switch to the chained CPI would reduce benefits and save money, there are two important points that Congress must recognize. First, these cuts are substantial. A cut in benefits of 0.2-0.3 percent in a single year may not be of much consequence, but the size of the cut accumulates over time. After ten years, an annual reduction in benefits of 0.3 percentage points implies a 3 percent cut in benefits. After twenty years the cut would be 6 percent and after 30 years it would be 9 percent. (These calculations ignore the impact of compounding.)

Such a pattern would imply sizable cuts in benefits to many people whose sole income is Social Security. In fact, the implied cuts would be considerably larger, measured as a share of income, than the tax increases on the wealthy that were put in place at the end of 2012. While the Obama administration has proposed ways in which the impact of these cuts could be ameliorated, there can be no doubt that they would be felt by a population that for the most part does not fit anyone’s definition of wealthy.

The other point is that there is no basis whatsoever for claiming that the C-CPI-U provides a more accurate measure of the rate of inflation experienced by the elderly. The BLS has an experimental index that tracks the specific consumption patterns of seniors, the CPI-E. This index has typically shown a rate of inflation that is 0.2-0.3 percentage points higher than the overall CPI. The reason for the difference is that medical care comprises a much larger share of the expenses of the elderly, and health care costs have typically risen more rapidly than the general rate of inflation.

If the goal is to have a more accurate measure of the rate of inflation experienced by the elderly, then we should want to see Social Security adopt the CPI-E as the basis for indexation. Many have objected to going this route based on the fact that the CPI-E is an experimental index that simply re-weights components of the CPI-U, rather than directly collecting data on prices for the specific items purchased by seniors at the outlets used by seniors. (In fact, the CPIs constructed by BLS already allow for substitutions between narrow categories of products, such as types of apples. It is possible that this overstates the extent of substitution done by seniors and for this reason understates the rate of inflation they experience.) 

While this complaint against the CPI-E is true, the fact that the CPI-E is not a full CPI is a policy choice. If the goal is to have an index that accurately measures the rate of inflation seen by the elderly, then BLS has the expertise to accomplish this task, if Congress were to instruct it to do so. A full elderly index would also allow researchers to determine if seniors substitute purchases to the same extent as the population as a whole. There are many instances where seniors find it less practical to change their consumption patterns than the population as a whole. For example, someone with a heart condition is unlikely to substitute from margarine to butter in response to a price increase in margarine. Also many of the items consumed by seniors may not lend themselves to easy substitution. It is difficult to substitute away from hip surgery or other medical procedures.

If the goal is to have an accurate measure of the rate of inflation experienced by seniors then there is little excuse not to go the route of constructing a full elderly CPI. The notion of tying post-retirement benefits to the rate of inflation certainly seemed to be the intent of Congress back in 1975 when it indexed benefits to the best measure of inflation available at the time.

Congress is free to change its approach to Social Security, but it is simply dishonest to claim that indexing benefits to the chained CPI is an effort to provide a more accurate measure of the rate of inflation. It is intended as a way to reduce benefits and should be evaluated as such.


Social Security is the main source of retirement income for the overwhelming majority of workers. Its importance is projected to grow in the years ahead largely as a result of the collapse of the traditional defined-benefit pension system. There are good arguments that Congress should be looking to expand, rather than cut, Social Security, especially for low-and moderate-income workers. It is also important that Congress understands that proposals to switch the basis for indexing post-retirement benefits to the C-CPI-U involve cuts to Social Security, not making the indexation formula more accurate. If the goal is to have an indexation formula that more closely tracks the rate of inflation experienced by seniors, then Congress should use either the experimental CPI-E already produced by BLS or instruct the agency to construct a full CPI-E.  


[1] The numbers in this paragraph are taken from Social Security Administration, 2013. “Fast Facts About Social Security,” http://www.ssa.gov/policy/docs/chartbooks/fast_facts/, accessed 12-16-2013.

[2] Census Bureau, 2013. “Income, Poverty, and Health Insurance Coverage in the United States,” http://www.census.gov/newsroom/releases/archives/income_wealth/cb13-165.html, accessed 12-16-13.

[3] Short, Kathleen, 2013., “The Research Supplemental Poverty Measure: 2012”  Census Bureau, www.census.gov/prod/2013pubs/p60-247.pdf, accessed 12-16-13.

[4] Social Security Administration, 2013. “Fast Facts About Social Security,” http://www.ssa.gov/policy/docs/chartbooks/fast_facts/, accessed 12-16-2013.

[5] Baker, Dean and Hye Jin Rho, 2011. “The Potential Savings to Social Security from Means-Testing,” Center for Economic and Policy Research, http://www.cepr.net/index.php/publications/reports/the-potential-savings-to-social-security-from-means-testing, accessed 12-17-2013.

[6] These numbers are taken from Butrica, Barbara A., Karen E. Smith, and Howard M. Iams, 2012, “This Is Not Your Parents’ Retirement: Comparing Retirement Income Across Generations,” Social Security Bulletin Vol. 72,#1.

[7] The income figures in the table include a value of imputed rent for owner-occupied housing.

[8] Social Security benefits are calculated by applying lower percentages to higher levels of earnings.  The “bend points” are the levels at which the percentages drop.  See http://www.ssa.gov/oact/cola/Benefits.html.

[9] Institute for Women’s Policy Research and National Committee to Preserve Social Security and Medicare Foundation, 2013. “Breaking the Social Security Glass Ceiling: A Proposal to Modernize Women’s Benefits,”  www.ncpssmfoundation.org/Portals/0/breaking_ss_glass_ceiling.pdf, accessed 12-17-2013.