FROM ISSUE NUMBER 20 ~ SUMMER 2014 GO TO TABLE OF CONTENTS
Is There a Retirement Crisis?
If you have followed the headlines in recent years on the question of Americans' retirement savings, you could be forgiven a bit of panic. "Our next big crisis will be a retirement crisis," read one headline in MarketWatch earlier this year. "The Greatest Retirement Crisis in American History," read another in Forbes. A study by the National Institute on Retirement Security put its warning in the form of a question: "The Retirement Savings Crisis: Is It Worse Than We Think?"
These predictions of doom typically point to one or another recent study prepared by important-sounding groups. The Center for Retirement Research at Boston College estimates that more than half of working-age households are at risk of having inadequate retirement resources. The National Institute on Retirement Security goes further, claiming that at least 65% of workers are saving less than required to meet their retirement income needs. The New America Foundation reports that, among middle-income retirees, "[f]ewer than half...have any form of pension income, and only a slim majority have any form of asset income." Unsurprisingly, 92% of Americans believe that we face a retirement crisis, according to one recent survey conducted for PBS. "And they're right," the PBS web site notes in reporting the finding.
But the facts are not nearly so simple, and the story of the retirement crisis has often been sold as much as told. Sometimes this selling is entirely well-intentioned, moved by a desire to get Americans to save more. In other cases, there may be selfish motives: Ask yourself, what are the chances the financial-services industry will tell me I'm saving too much for retirement? In yet other cases, the motives are political: By arguing that America's private retirement-savings system has failed, progressives pave the way to eliminate tax incentives for retirement saving and expand Social Security to take its place.
Policy proposals often reveal the political motivations of their authors. Massachusetts senator Elizabeth Warren has become the darling of the left for supporting a larger Social Security system. Iowa senator Tom Harkin, chairman of the Committee on Health, Education, Labor and Pensions, calls for enhancing Social Security benefits and creating a new USA Retirement Fund system to automatically enroll workers not covered by existing retirement plans. President Obama has offered proposals to set up "myRAs" for lower earners and to limit accumulations for higher earners in pensions and retirement savings plans. The New America Foundation study calls for eliminating most of the tax preferences for existing retirement savings plans and substantially increasing Social Security benefits. It is not hard to see where these ideas point: to the vast majority of Americans receiving the vast majority of their retirement incomes from the government. The so-called "three-legged stool" of Social Security, employer retirement plans, and personal saving may end up with one leg significantly longer than the others.
Dire tales of crisis serve this goal quite well, but are they true? A review of the key sources of data and analysis behind the crisis narrative raises some grave questions about their accuracy and adequacy. The most commonly cited figures and articles overstate what workers need to provide themselves with a secure retirement income, underestimate what they have been accumulating for that purpose, and ignore much of the income paid to retirees out of their pensions and retirement savings plans. This combination results in deeply misleading conclusions about the state of Americans' retirement savings, and may be leading us toward much greater dependence on public retirement systems that are already badly underfunded and threaten government finances at every level.
QUANTIFYING RETIREMENT NEEDS
In considering the validity of the available research regarding retirement savings, it is worth starting with a basic sense of how most economists and policy analysts think about retirement.
The dominant approach in the field is the so-called "life-cycle model" developed by Franco Modigliani and Richard Brumberg in the 1950s. In this view, individuals seek to rearrange their resources through borrowing and saving in order to maximize the well-being — or "utility" — that they derive from them. In simple terms, the life-cycle model explains why individuals often borrow when young, save during middle age, and live off those savings in retirement. While income varies dramatically from year to year, consumption is far smoother.
Adequate retirement income, in this sense, is an income that allows retirees to maintain their pre-retirement standard of living. It is not an income that makes a household rich in retirement; indeed, it is possible for a household to smooth its consumption perfectly between work and retirement and yet have a very low level of consumption in both periods. If a household lives in poverty during its working years and maintains that poverty-level standard of living in retirement, that household's problem is not inadequate retirement planning but the simple fact that it is poor. But if a household of any income level finds that inadequate resources force upon it a dramatic reduction in its standard of living after retirement, that is evidence of inadequate retirement planning.
Simply put, to understand retirement preparation we need to know how much Americans are setting aside for retirement and how much they will need. Unfortunately, many of the popular and influential studies pointing to a retirement crisis get one or both of these answers wrong. Some studies ignore Social Security's progressive benefit formula, which replaces a far larger percentage of pre-retirement earnings for low and middle earners than for high earners. As a result, a lower earner need not save nearly as much to supplement Social Security retirement benefits. Other studies ignore the fact that workers with advanced degrees may have student debts that slow early-career saving relative to those with less education. But the advanced degrees often result in higher earnings that give workers greater saving capacity later. Finally, most studies completely ignore the effect of children on household consumption and saving patterns.
The nuts and bolts of the various studies predicting a retirement crisis are beyond the ability of most Americans to understand, much less replicate. At the same time, however, the models used in these studies are often crude relative to those employed in academic studies and the far more sophisticated models developed by government agencies. There is nothing wrong with shorthand calculations and rules of thumb, so long as they generally point toward the same conclusions as more sophisticated modeling. But in many cases this isn't true with regard to the retirement-crisis literature.
With these facts in mind, then, let us review the key studies that make up the case for a coming retirement crisis.
THE INCOME OF THE AGED
A major theme of the retirement-crisis story is that the private retirement saving system — which for most individuals means employer-sponsored 401(k) plans and individual retirement accounts — has been ineffective for large segments of the population and is becoming less effective as time passes.
The Social Security Administration has for several decades presented evidence to this effect, which critics of private retirement savings employ in making their claims. The shift by private employers from traditional defined-benefit plans to defined-contribution plans, moreover, has heightened criticisms of this segment of the retirement system.
Since the mid-1970s, the Social Security Administration has released a series of reports called "Income of the Population 55 or Older" that purport to show the sources and levels of income received by people in or near retirement. According to their most recent estimates, around 86% of those aged 65 or older receive Social Security benefits, which make up roughly 35% of seniors' incomes. Income from employer-sponsored retirement plans or individual retirement accounts, by contrast, is both less common and less adequate: The SSA reports that only 39% of elderly people receive pensions and those pensions comprise only 17% of retirees' total income. The New America Foundation's Michael Lind and his co-authors explicitly cite these data in calling for expanding Social Security and eliminating tax incentives for private retirement saving.
But there is a very simple problem with these SSA figures, a problem the agency has been aware of for almost two decades: Their data source, the Census Bureau's Current Population Survey, ignores almost all of the income that retirees receive from 401(k)s and IRAs. In the CPS, pension benefits are counted as income only if they are paid on a regular basis. So, for instance, a monthly benefit check from a traditional pension is considered income. But occasional, as-needed withdrawals from an IRA or 401(k) — which is how most people use these accounts — are not counted as income.
Internal Revenue Service data show the scale of this omission. Since most withdrawals from IRAs and 401(k)s are taxable, they must be reported to the IRS. The SSA reports that retirees who were collecting Social Security benefits in 2008 also collected $5.6 billion in income from IRAs. The IRS data, by contrast, show that Social Security beneficiaries had $111 billion in IRA withdrawals that year. Similarly, the SSA reports $200 billion in employer pensions and annuities going to households receiving Social Security, less than half the $457 billion reported in IRS data. In total, the SSA reports that retirees collected only $228 billion from employer-sponsored pensions and individual retirement savings plans — just 40% of the $568 billion total reported by the IRS.
Put another way, the CPS data that SSA relies upon state that, on average, retirees' pension benefits are equal to 58% of their Social Security benefits. But economists Al Gustman, Thomas Steinmeier, and Nahid Tabatabai, using the Health and Retirement Study — which does count irregular pension withdrawals — find that pension income is 82% as valuable as Social Security. Combined pension and Social Security income is, on average, 15% higher in the HRS than in the CPS, a figure that is consistent with IRS data.
Moreover, these ignored pension withdrawals are not isolated at the top of the income distribution. Income-tax filings show that across much of the income distribution, significantly more retirees receive income from pensions or retirement savings plans and receive higher benefits than the SSA reports.
The Social Security Administration has known since the mid-1990s that the CPS omitted much of the income that seniors derive from retirement plans. And the errors in the agency's reports have only grown larger since then, as traditional defined-benefit plans have given way to 401(k)s and IRAs. Between 2011 and 2012, the balances in private retirement accounts increased by nearly seven times the amount that private defined-benefit assets grew. It is not difficult to conclude that the private retirement system is not working when seemingly authoritative data systemically exclude large portions of the income derived from these plans. But ignoring this part of the retirement system makes no sense in discussions about retirement income security.
While understating what Americans have saved for retirement, the Social Security Administration also overstates how much individuals will need. In a publication designed to educate workers nearing retirement, the SSA states that "[m]ost financial advisors say you'll 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." These statistics, precisely because they come from such an authoritative source, have made their way into countless discussions of retirement security. But the agency's figures are misleading.
The problem is that financial advisors measure replacement rates relative to final earnings — that is, they measure the first year of retirement income as a percentage of the final year of working income — while the Social Security Administration measures replacement rates relative to the wage-indexed average of lifetime earnings. These are entirely different numbers.
"Wage indexing" adjusts past earnings years to reflect the growth of average wages in the economy. Say you are retiring at age 65 this year and earned $20,000 in 1985. Adjusted for price inflation, that 1985 salary was worth $43,640 in 2014 dollars. But in calculating replacement rates, SSA wage indexes that $20,000 for the growth of the economy, generating a value of $53,281. Not surprisingly, replacing 70% of $53,281 is a lot more difficult than replacing 70% of $43,640.
For SSA's hypothetical medium earner, the wage-indexed figure of average lifetime earnings used in SSA's replacement-rate denominator is 35% higher than the average of his final five years of earnings. It is easy to make Social Security benefits appear meager if the denominator in the replacement-rate calculation is increased by 35% compared to the rule of thumb used by financial advisors. Even when compared to the inflation-adjusted average of an individual's total lifetime earnings, the wage-indexed denominator is still 10% higher. There is almost nothing in economic theory that explains why an individual should want to replace some percentage of his wage-indexed pre-retirement earnings, nor is it common to do so in financial planning.
It is possible to construct an ex post rationale for SSA's approach, though not one that fits very well with either the life-cycle model or financial advisors' advice. But the origin of SSA's wage-indexed replacement-rate denominator is in fact much simpler: Using this measure allowed the agency to maintain its talking points even as it improved certain methods used to calculate retirement benefits. Up through 2000, SSA calculated replacement rates the same way financial advisors do: relative to earnings immediately preceding retirement. The agency illustrated these replacement rates using hypothetical individuals who, in each year of their working lives, earned some percentage of the national average wage. The hypothetical low-wage worker earned 45% of the average wage each year; the medium-wage worker earned 100% of the average wage, and the high-wage worker earned 160% of the average wage. Based on these assumptions, low-, medium- and high-wage workers received Social Security replacement rates of about 55%, 41%, and 34%, respectively. These stylized facts on the adequacy of Social Security benefits, based on income immediately before retirement, were well-known and oft-repeated.
But beginning in 2002, the SSA updated its hypothetical earners to more realistically estimate the effects of personal retirement accounts on benefits. Instead of earning a steady percentage of the average wage each year, SSA's revised earnings followed a hump-shaped pattern in which earnings rise to a peak in the worker's fifties, then stabilize and even decline slightly as the worker nears retirement.
While more accurate, these new hypothetical workers presented a problem: Measured relative to final earnings, their Social Security replacement rates would be higher. Instead of the 55%, 41%, and 34% replacement rates calculated for the SSA's old hypothetical workers, these more realistic earners would receive replacement rates of approximately 68%, 51%, and 42%. No government agency likes to change its talking points, and in this case doing so would make Social Security seem more generous than before. So the SSA altered the way it measured replacement rates from final earnings to wage-indexed average lifetime earnings that, along with a little extra adjustment, produced the desired result: Replacement-rate figures published by SSA didn't change. In fact, the 2002 Social Security Trustees Report, which introduced replacement rates based on these new stylized earners, explicitly acknowledges that its methods were designed and calibrated to maintain the same replacement figures the agency had been publishing for years.
Using publicly available data, it is possible to measure the replacement rates Social Security actually pays relative to the 70% rule of thumb used by financial advisors. The Social Security Administration's Benefits and Earnings Public-Use File contains thousands of individuals' work histories culled from SSA's databases, along with the benefits received by each individual. Using these data, we calculate that the typical long-term worker who works until the age that unreduced benefits can be claimed receives a Social Security benefit equal to about 62% of his final earnings — defined here as the average of non-zero earnings in the five years prior to claiming Social Security benefits — and 52% of his inflation-adjusted average lifetime earnings. Among lower earners, replacement rates are higher. For instance, a long-term worker in the third earnings decile would receive a replacement rate of 75% of final earnings and 64% of his real average lifetime earnings.
Many workers choose to retire early and receive reduced benefits. Including these workers reduces the median benefit paid to long-term workers to 53% of final average earnings and 43% of real average lifetime earnings. These benefits are considerably higher than Social Security's presentations imply. Moreover, while early retirees may receive lower Social Security replacement rates, replacement-rate targets for early retirees also are lower: An individual who seeks to retire early must save a greater portion of his pre-retirement earnings, leaving less for consumption. Thus, an early retiree can match his pre-retirement consumption while receiving a lower replacement rate from Social Security and other forms of retirement income.
Given these facts, the adequacy of Social Security retirement benefits, which serve as the foundation for overall retirement preparation, can be seen in a different light. Unfortunately, some other studies, such as those from Boston College's Center for Retirement Research, have used the wage-indexed replacement-rate denominator for studying overall retirement-income adequacy. As far as we are aware, though, neither these groups nor the SSA itself have ever explained why this measure makes sense as a way of determining whether retirees can maintain the standards of living they achieved while working. The point is not that 70% is a perfect replacement rate or that final earnings are what replacement rates should be measured against. As we will see, neither is necessarily true. But if we do choose to follow financial advisors' widely accepted rules of thumb in measuring retirement-income preparedness, Americans are far better off than we are being led to believe.
THE SOCIAL SECURITY STATEMENT
The Social Security Administration builds this same flaw into its "Social Security Statement." The Statement, available to every working-age individual not yet collecting Social Security benefits, includes a record of the individual's earnings and contributions to the program, as well as an estimate of the benefits he may receive upon retirement or disability. In 2009, the Social Security Advisory Board stated that "it is imperative that the Social Security Statement provide the most accurate information possible and that information be communicated in a clear and objective manner." But the Statement fails to meet these goals.
The problem is not in the way the SSA projects the benefits that workers will receive. The agency assumes that individuals will continue to work at their current wages and receive small annual raises until retirement age, and then it uses these projected earnings to estimate future benefits. While any such estimate involves errors, the SSA's approach predicts future benefits fairly well. Where the agency goes wrong, however, is in how those future benefits are expressed in the Statement. Ordinarily, a future dollar amount would be expressed either in nominal dollars (meaning the actual dollar amounts that will be written on benefit checks) or in inflation-adjusted dollars, which would express the current purchasing power of those future benefits. Either approach can be useful, and ideally both figures would be provided.
But the SSA's Social Security Statement does neither. While until recently the Statement declared that benefits were expressed "in today's dollars" (which suggests they are adjusted for inflation), the Statement in fact expresses benefit amounts in "wage-indexed dollars." Just as wage indexing of past earnings increases their value in the replacement-rate calculation, wage indexing of future Social Security benefits reduces their value.
The effects on an individual's estimated Social Security benefits can be large. For example, a typical worker retiring 30 years from today will receive a nominal Social Security benefit of about $64,750 per year. That sounds like a lot, until you realize how big a role inflation plays. Adjusted for inflation, that future benefit will be just $27,683. That's a figure that a person planning for retirement can more easily understand. But that's not the figure he'll see on his Social Security Statement. Rather, because the SSA wage-indexes his future benefits, the figure he would see on his Statement will be just $17,982, which is 35% lower than what the true purchasing power of his benefits is expected to be.
When critics pointed this out to the SSA over the last decade, the response was not to correct the way projected benefits are expressed in the Statement; it was to remove the phrase "in today's dollars." Workers are now left with no explanation of what the dollar amount on the Statement means. A Frequently Asked Questions document on the SSA's web site says, in reference to this figure in the annual Statement, "[we] show the resulting estimates in today's dollar amounts (rather than in 'future dollars' adjusted for assumed inflation) so you can compare them with today's living costs." But that is not correct; the Statement doesn't express future benefits indexed for changes in the cost of living but rather with a greater adjustment that can significantly understate the benefits individuals are actually entitled to receive.
It's logical to point out, of course, that Social Security is underfunded and full promised benefits may not be paid. But even if Congress addressed the entire Social Security shortfall by reducing benefits through a method referred to as "price indexing" of the benefit formula, most individuals would still receive higher benefits than are expressed on their Social Security Statement. It is no wonder that many Americans fear for their retirement security when they are being told they will receive so little.
CRUDE RULES OF THUMB
If the amount of press ink used to discuss a study is any indicator of its contribution to public understanding, the National Institute on Retirement Security's 2013 report entitled "The Retirement Savings Crisis: Is It Worse than We Think?" might be considered the gold standard on the subject of retirement security.
The report, authored by NIRS analyst Nari Rhee, concludes that 84% of workers are not meeting reasonable retirement-savings targets and, even when total net worth is considered, two-thirds are not on a sound path to retirement security. Aggregate saving may be as much as $14 trillion below what workers should have socked away, according to the analysis. Economist Nancy Folbre, writing in the New York Times, concluded that "[t]he report lends weight to the longstanding criticisms of the increased reliance on individual savings in the United States retirement system."
But looking behind the fancy name and membership of the organization that published the report — NIRS membership includes a long list of public-employee retirement plans, unions, financial-management and insurance companies, some pension consultants, and the AARP — the substance of the study should give pause to anyone who understands the nature of the analysis behind the numbers.
The NIRS study uses Fidelity Investments's 2012 release, "Fidelity Outlines Age-Based Savings Guidelines to Help Workers Stay on Track for Retirement," as a starting point. Fidelity's baseline, in turn, assumes that workers will stay in the labor force until age 67 and that their retirement incomes should replace 85% of their pre-retirement earnings. Fidelity's hypothetical workers start saving at age 25 and save continuously until they retire at age 67. They contribute 6% of pay at age 25, rising to 12% by age 31 and remaining constant thereafter.
Based on these assumptions, to hit the 85% replacement-rate target a worker should have saved an amount equal to his current annual salary by age 35, three times his annual salary at 45, five times salary at 55, and eight times final salary in the year before retirement. Using data from the Federal Reserve's Survey of Consumer Finances, NIRS compares individuals' reported retirement savings to Fidelity's savings goals. To the degree that actual savings fall short of NIRS's benchmarks, an individual is considered to have inadequate retirement savings.
But there are a number of serious problems with NIRS's application of Fidelity's rule-of-thumb savings path. For one, Fidelity's 85% replacement-rate target is higher than many other estimates — note that the SSA cites a 70% recommended target. More important, NIRS either ignores or does not understand that Fidelity's rule-of-thumb savings milestones are for an average earner. Because Social Security replaces a higher percentage of lower earners' pre-retirement earnings, low earners would have to save less than the average earner to hit the Fidelity replacement income target of 85% of pre-retirement earnings. The opposite is true of higher earners — they would have to save more. And NIRS is also totally silent about the different consumption patterns and savings requirements of people who have children versus those who do not.
Finally, despite the fact that Fidelity clearly labels its suggestions as based on a "rule of thumb," NIRS applies its savings milestones as universal requirements, even though there is evidence that many workers save comparable amounts over their lifetimes with savings patterns different than Fidelity's linear approach. In essence, NIRS requires that workers contribute 6% of pay to their retirement savings plan at age 25 and then increase it by one percentage point of salary each year up to 12% of pay, and then contribute that amount annually until retirement. If workers instead contributed 6% per year from age 25 to age 40 and then increased the contribution rate by one percentage point per year up to 20%, for instance, at age 67 they would have the same accumulated savings as NIRS workers. Under this alternative savings pattern, workers fall behind the NIRS benchmark savings levels by 21% at age 30, 34% at age 40, and 23% at age 50. NIRS judges these workers as having inadequate retirement savings in every year through age 65 even though they end up exactly on target by retirement age. After getting early-life debts and start-up expenses under control, many workers can ramp up their contributions during peak earning years and as children leave home. Yet at every point except for retirement, NIRS would judge these workers at risk of an inadequate retirement income because they deviated from a set of arbitrary guideposts.
As the University of Wisconsin's John Karl Scholz and Ananth Seshadri note in analyzing similar wealth-to-income ratios in the Health and Retirement Study, these "ratios may be consistent with problems in wealth accumulation, or may reflect precisely the pattern we would expect to see if the lifecycle model capably summarizes behavior." In other words, NIRS's basic approach tells us very little because we have no firm grasp of how much a given household should have saved by a given age.
These are hardly methodologies robust enough to support claims of a "retirement savings gap" of up to $14 trillion, even if newspapers find such claims attractive headline-bait.
ONE SIZE DOES NOT FIT ALL
One of the best known, most frequently cited, and better-researched studies of retirement preparedness is the National Retirement Risk Index, compiled by the Center for Retirement Research at Boston College. The Center's researchers are solid, and its intentions appear good. And yet the Center's researchers make a number of debatable methodological choices, most of which point in the direction of overstating the degree to which Americans are under-saving for retirement.
The NRRI begins with replacement-rate targets generated by Aon Hewitt, a benefits consulting firm. The Center uses the Federal Reserve's Survey of Consumer Finances to develop a model of the working population. For each individual or household, the Center projects forward its earnings and the development of its household wealth. At retirement age, the NRRI compares the income that can be derived from the household's wealth to its pre-retirement earnings. If the household's projected replacement rate falls short of its target replacement rate by more than 10%, that household is considered "at risk" of being unable to maintain its standard of living in retirement. Using this approach, the Center's researchers found in 2010 that 44% of working people in their fifties, 55% of those in their forties, and 62% of those in their thirties were at risk. Overall, the NRRI found that 53% of all working Americans were at risk of having seriously insufficient incomes in retirement. When retiree health costs and long-term health expenses are considered, 64% are estimated to be at risk.
In many ways the NRRI is the best of the popular studies. For instance, it includes household wealth along with explicit retirement saving to align more closely with a traditional life-cycle model of saving and consumption. The NRRI also avoids the more egregious methodological errors of studies such as that from NIRS and the overstatements based on faulty data made by the New America Foundation researchers.
And yet there are a number of questionable methodological choices that push the NRRI's "at risk" figures upward. For instance, while the NRRI adopts replacement-rate targets that are consistent with financial advisors' rules of thumb — such as 70% for a single male with average earnings — the NRRI calculates replacement rates relative to wage-indexed average lifetime earnings, even though financial advisors calculate replacement rates relative to final earnings. As we have seen, a wage-indexed measure of lifetime earnings is generally substantially higher than a worker's final earnings. As a result, the NRRI increases households' income targets and thereby increases the share of households deemed "at risk" in retirement.
The NRRI also measures pre-retirement income differently than most other studies. Most studies focus on earned income, which is by far the most important component for most working-age households. The NRRI, by contrast, includes capital income, which for most households will principally be composed of interest earned on retirement accounts such as 401(k)s. But, since this is money set aside for retirement, the vast majority of this capital income isn't consumed. Including capital income raises the bar for middle- and upper-income households and causes more of them to be judged to be saving inadequately.
The replacement-rate targets the NRRI adopts also may overstate the number of households at risk of an inadequate retirement income by inadequately accounting for household composition. The NRRI adjusts its target replacements based on income; for instance, a low-income couple has a target replacement rate of 81% while a high-income couple has a target of just 67%. And the NRRI's target replacement rates make small adjustments for couples versus singles.
However, the NRRI's target replacement rates — along with most other popular depictions of the retirement crisis — make no adjustment for whether the individual or couple has or had children. University of Wisconsin economists John Karl Scholz and Ananth Seshadri conclude that "children have a substantial effect on the level and dispersion of wealth and thus should be accounted for in typical retirement planning advice." Dartmouth economist Jonathan Skinner states the logic in humorous terms: "[P]arents are already used to getting by on peanut butter, given that a large fraction of their preretirement budget has been devoted to supporting children, so it's not difficult to set aside enough money to keep them in peanut butter through retirement. By contrast, childless households with the same income but accustomed to caviar and fine wine must set aside more assets to maintain themselves in the style to which they have become accustomed."
Now, many parents understandably might wish to shift from peanut butter to caviar once their nest is empty, but this generally appears not to happen: Households with children save less for retirement than those without kids, and this lower saving is entirely consistent with a life-cycle model in which households smooth consumption, be it of peanut butter or caviar.
While the NRRI does not account for the effects of children on households' retirement-saving needs, it is possible to do so. Consumer Expenditure Survey data show that children consume around 70% as much as adults in the same household. Based on these data, the National Academy of Sciences constructed a formula by which to adjust consumption both for the size of the household — single or couple — and the number of children present. A similar scale is currently used by the Census Bureau in calculating poverty thresholds for different household types in the recently released Supplemental Poverty Measure.
The effects of children on desired replacement rates can be significant. For simplicity, consider a single adult with no children and an income of $50,000. Since there is no one else in the household, that adult consumes most of that $50,000, with the remainder left for taxes, mortgage payments, saving, and other related costs. If we assume that he requires a 70% replacement rate in retirement, he could get by adequately with an income of $35,000. Now imagine he has one child: According to the NAS calculations, this single parent's share of his $50,000 gross income is $34,483; 70% of that amount is just $24,138. Put another way, his target replacement rate relative to his gross pre-retirement household income of $50,000 would be just 48%. Add another child and the target falls to 38%.
These results may sound extreme, but they are borne out in peer-reviewed economic research, which receives little attention in the public discussion of retirement preparedness. The University of Wisconsin's Scholz and Seshadri calculated target replacement rates using a life-cycle economic model that accounts for a range of factors, including children. They calculated a median target replacement rate of 75% of inflation-adjusted average lifetime income; this value is itself below the NRRI's targets, because the NRRI wage-indexes lifetime income. But even more important, nearly half of households had optimal target replacement rates below 65%.
Similarly, a 2009 study (conducted by one of us, Andrew Biggs, for the American Enterprise Institute) found that adjusting for family size and composition effectively increases the realized replacement rates of households by approximately 15 percentage points. Both this study and the one by Scholz and Seshadri indicate that a large number of households that may well be saving optimally for retirement would be judged "at risk" by the NRRI and other studies because their targets ignore the presence of children. One size does not fit all.
Finally, the NRRI — like the NIRS study — utilizes the Federal Reserve's Survey of Consumer Finances as an integral part of its calculations. The SCF is a fine data set, but due to small sample sizes it is not perfectly suited for this type of detailed analysis. The 2010 SCF, which was the basis for the NRRI, has a total sample size of about 6,500. Of this sample, only around 60% are in the 30 to 60 age group that the NRRI analyzed. And this remaining sample is divided three ways to examine Early Boomers, Late Boomers, and GenXers, meaning that each group may potentially be studied using a sample size of around 1,300 individuals. Moreover, the SCF oversamples high-income households, meaning that the sample for low- and middle-income households is relatively smaller. As William Emmons and Bryan Noeth of the Federal Reserve Bank of St. Louis note, "Recent SCF sample sizes have been increased, but they remain small for performing detailed examinations — particularly of the undersampled groups who are not wealthy, such as young, minority, and less-educated families." These under-sampled groups are presumably those most at risk of an inadequate income in retirement.
Honest analysts can differ on some of these methodological questions. But if the NRRI were recalculated using final income, as financial advisors do, or real lifetime income, and if it defined income excluding investment earnings, as most financial advisors and analysts do, and if it adjusted its replacement-rate targets for the costs of raising children, its results regarding the number of Americans at risk of an inadequate retirement income would likely change dramatically. And this in turn might dramatically alter the debate about retirement security in the United States.
READY OR NOT?
Every study of retirement income security must build a theoretical model and an empirical estimate. The theoretical model calculates how much households should save for retirement, while the empirical estimate determines how much households actually have saved. As we have shown, many of the most prominent and frequently cited retirement-adequacy models have significant, perhaps fatal, methodological flaws. They ignore the income derived from personal retirement plans; they overstate the income households need in retirement; they set arbitrary saving goals with no foundation in economic theory; they define pre-retirement income in odd ways; and they fail to make necessary adjustments for household composition.
So where do things actually stand? How bad is it? We are not prepared to point to a single figure, given all the uncertainties involved in these calculations. But we are confident in saying that things aren't nearly as bad as some people claim. There are a number of ways to demonstrate this.
We can begin with what current retirees are saying. Ohio State University economist Jason Seligman finds, using Health and Retirement Study data, that two-thirds of current retirees deem themselves "very satisfied" in retirement, with another quarter calling themselves "moderately satisfied." Just 14% report that their retirement years are "not as good" as the years immediately preceding retirement. Likewise, RAND economists Michael Hurd and Suzann Rohwedder, also using HRS data, find that most retiree households end up with higher incomes than they had anticipated. "If anything," they conclude, "households seem to be pleasantly surprised by their level of resources" in retirement.
We can also turn to a number of high-quality academic studies on retirement preparation. These studies are conducted by academic economists with no apparent bias and are not funded by industry groups, and they generally are peer reviewed or presented at academic conferences. Unfortunately, these studies are often highly technical, rendering them beyond the easy understanding of typical reporters and newsreaders.
RAND's Hurd and Rohwedder, for instance, have published widely with regard to many aspects of retirement security. In a highly detailed 2008 study of recent retirees funded by the Social Security Administration, for instance, they find that 83% of married couples and 70% of singles are adequately prepared for retirement. Given that married retirees outnumber singles by roughly two to one, their overall population results imply that 79% of new retirees are adequately prepared. Interestingly, they find that perhaps the biggest threat to retirement security is taxes paid in retirement: Before accounting for taxes, 84% of recent retirees were well prepared.
A second high-quality study is one conducted by Wisconsin's Scholz and Seshadri, along with the Brookings Institution's William Gale. The authors use HRS data to construct a life-cycle model that accounts for a wide variety of factors, including marriage and children. The study concludes that around 26% of households are currently under-saving for retirement. For those who are under-saving, the median shortfall is $32,000, or 17% of the median optimal wealth level. The authors conclude, "While the results suggest that some households will need to ratchet their living standards downward in retirement, most Americans are, by in large, preparing sensibly, given the existing generosity of social security, Medicare, and pension arrangements." Again, this is not to deny that some Americans, even a significant number, are falling short in preparing for retirement. But high-quality studies tend to find that fewer Americans are falling short, and by a smaller amount, than the cruder, more attention-grabbing studies conclude.
Probably the most detailed and best-vetted computer model for retirement purposes is maintained by the Social Security Administration in cooperation with the Urban Institute and other research organizations. "Modeling Income in the Near Term" — MINT for short — was developed beginning in the late 1990s and is currently in its sixth iteration. MINT simulates individuals over their full working lives, incorporating education, work, marriage, divorce, and saving — practically the full range of factors that affect individuals' preparations for retirement. The MINT model reports not simply Social Security benefits, but pensions, welfare benefits, housing equity, and other potential sources of retirement income. In terms of sophistication, MINT is a quantum leap beyond the models used in the popular retirement-crisis literature.
In a 2012 study, SSA analysts used the MINT model to project retirement income for four groups: "depression babies," born from 1926–1935; "war babies" (1936–1945); "leading boomers" (1946–1955); "trailing boomers" (1956–1965); and "GenXers" (1966–1975). For each group, the study calculated replacement rates relative to inflation-indexed average lifetime earnings. The median, or typical, replacement rate for Depression Babies was 109%, rising to 119% for War Babies, and then gradually declining to 116% for Leading Boomers, 113% for Trailing Boomers, and 110% for GenXers. These figures indicate both that future generations of retirees typically will have incomes substantially exceeding the real incomes they enjoyed while working, and that replacement rates for future retirees will not be dramatically lower than for Americans retired today. These figures hardly support political scientist Jacob Hacker's dire contention that "[w]e live in the waning days of the Golden Age of Retirement."
Nor do the MINT model's projections show an emerging underclass in terms of retirement security, a group that lives in poverty even as others do well. For instance, MINT estimates that only 26% of Depression Babies had replacement rates below 75% of their average pre-retirement earnings, and only 8% had replacement rates below 50%. MINT's comparable figures for the supposedly dramatically under-saving GenXers are 25% and 8%. In other words, despite significant changes in the composition of retirement income — future retirees will rely more on asset income and less on traditional defined-benefit pensions than do present retirees — the overall level and distribution of retirement income will remain roughly the same.
IMPROVING RETIREMENT SECURITY
Any model of retirement saving involves speculation, error, and uncertainty. Yet the types of results discussed above — from respected academic economists and government models that have been continuously vetted and improved over time — are almost never reported. Instead we hear claims that up to 90% of Americans are at risk of an inadequate retirement income, coupled with demands to dismantle current tax incentives for private retirement saving and substantially boost the size of a Social Security program that is already underfunded by $10 trillion or more.
Following the advice of these purveyors of gloom would be a mistake. This is not to deny that some Americans are underprepared for retirement. Inadequate retirement preparation is less prevalent and smaller in degree than the panic-stricken headlines generated by interest-group studies suggest, but it exists and, in a country as large as the United States, will affect millions of citizens. But a more limited and isolated retirement-saving problem demands different solutions than would be called for if the vast majority of Americans were under-saving by significant margins.
The progressive left is today almost completely united in supporting an expanded Social Security program. From a fiscal perspective this seems absurd, given that the program is already significantly underfunded. To make Social Security sustainably solvent while paying full promised benefits would require an immediate and permanent one-third increase in revenues. The proposals to expand Social Security continue Congress's habit of focusing more on promising benefits than paying for them. Iowa senator Tom Harkin's plan would eliminate the $117,000 cap on Social Security taxes, which would have the effect of raising top federal marginal tax rates by 12.4 percentage points. If Social Security benefits remained constant, at least, this tax increase would be sufficient to keep Social Security solvent for almost 75 years. But Harkin "spends" half the increase on increasing both initial Social Security benefits and the annual cost-of-living adjustments. As a result, despite a truly massive tax increase that effectively taps out high earners for any additional tax revenue, Harkin's plan extends Social Security's solvency by only 16 years, from 2033 to 2049.
Moreover, Social Security expansions would still leave many of the most vulnerable elderly in poverty in retirement. The reason is that Social Security is an earnings-based program: To qualify, an individual needs at least ten years of work history, and his benefits are based upon his wages and payroll taxes. In other words, individuals with sporadic attachment to the labor force — precisely those who are most likely to reach retirement age without significant savings — receive little or nothing from Social Security. The Social Security Administration estimates that 4% of the elderly will never receive benefits from the program. This so-called "never-beneficiary" population has, according to the SSA, "lower education levels and higher proportions of women, Hispanics, immigrants, the never-married, and widows than the beneficiary population. Never-beneficiaries have a far higher poverty rate (about 44 percent) than current and future beneficiaries (about 4 percent). Ninety-five percent of never-beneficiaries are individuals whose earnings histories are insufficient to qualify for benefits." And under proposals to expand Social Security, these individuals would remain never-beneficiaries because the left's goal is to make Social Security bigger, not to make it work better.
That is one reason why we have elsewhere proposed Social Security reforms that couple a universal retirement benefit, paid without reference to work history, with individual retirement savings accounts. This approach, which is similar to the retirement plans in place in Australia, New Zealand, and the United Kingdom, would provide a stronger safety net against poverty while increasing saving for retirement. And it could do so at far lower cost than an expanded Social Security program.
Even with a strong Social Security safety net, there are surely many middle- and even upper-income Americans who are not saving enough for retirement. But in assessing that problem, it is essential that we start with solid data. For instance, it is common to hear, in NIRS's phrasing, that "only half of private sector employees have access to workplace retirement benefits." President Obama himself made this claim in his 2014 State of the Union address. This figure derives from the Current Population Survey, in which individuals are asked whether their employer offers them a retirement plan. But as research from the Social Security Administration shows, many individuals answer incorrectly. The SSA's examination of tax records found that 72% of all workers in 2006 were offered a retirement plan by their employer; among firms with 100 or more employees, 84% of workers were offered a retirement plan.
Moreover, it is difficult to isolate individuals who are under-saving merely by looking at how many Americans participate in 401(k) or IRA plans. Economists Peter Brady and Stephen Sigrist show that many employees who are not offered pensions are younger or lower-income individuals who rationally may wish to devote their limited resources to other priorities. As these individuals age or their incomes rise, retirement saving becomes a higher priority, and they shift toward jobs where pensions are offered.
Yet simple policies such as automatic enrollment in 401(k) plans can and already have increased participation in employer-sponsored pension plans. While employees may opt out of participation if they wish, studies have shown that automatic enrollment can cut the number of pension non-participants in half. Some have proposed that all employers offering retirement plans be required to use automatic enrollment. Similarly, an increased focus on simple, low-cost investment offerings, perhaps packaged as life-cycle funds that automatically shift from stocks to bonds over time, could help all savers amass funds for retirement. These are simple policy changes with large potential benefits, which don't demand dismantling our current private savings system or expanding the underfunded Social Security program.
But given the broad disparities in optimal replacement rates that households would rationally pursue, it is extremely difficult to dictate a single saving rate or retirement-income target that will work for all households. A better approach is for public policy to aim for minimum levels of retirement income that are sufficient to avoid poverty and destitution, while allowing individuals and households — who know their needs and preferences better than a government planner — to decide how much to save on top of that minimum.
Whatever course we take, however, we must do so with our eyes open and with a clear grasp of reality. Talk of a massive retirement crisis lacks such a grasp, and so is more a hindrance than a help to improving retirement security.
Andrew G. Biggs is a resident scholar at the American Enterprise Institute and former principal deputy commissioner of the Social Security Administration.
Sylvester Schieber is an independent pension consultant and former chairman of the Social Security Advisory Board.