Is the Retirement Crisis a Mirage?

At the National Academy of Social Insurance conference earlier this month, keynote speaker Michael Lind of the New America Foundation called for reducing tax subsidies for 401(k)s and other employer-based plans to help pay for expanded Social Security benefits, a view I share. Lind said it was time to move Social Security expansion to the realm of serious discussion and relegate privatization to the political fringe.

Lind is part of a growing movement that supports expanding Social Security to fill a hole caused by past benefit cuts as well as what Paul Krugman has called the “gigantic failure” of the 401(k) experiment. A bill introduced last year by Senator Tom Harkin and Representative Linda Sanchez, for example, would provide a modest across-the-board increase in benefits as well as a higher cost-of-living adjustment in recognition of the fact that seniors are more affected by rising health costs. The movement got a boost last November when Senator Elizabeth Warren made a speech on the Senate floor on the need to increase benefits to avert a retirement crisis. In her speech, Warren cited a $6.6 trillion gap between what Americans under 65 were on track to save and what they needed to maintain their standard of living in retirement.

The counterargument was presented at the conference by Edward Ferrigno of the Plan Sponsor Council of America, an industry trade association. Ferrigno argued that the current system has worked for most Americans, and that the retirement crisis is an illusion created by incomplete data. Ferrigno cited an article by Social Security Administration researchers that pointed out that Census data on senior incomes often cited as evidence that most seniors live modestly and rely heavily on Social Security benefits underreports distributions from defined contribution (DC) plans and IRAs because it doesn’t include lump sum distributions. The same issue was raised by at least two other conference participants: Andrew Biggs of the American Enterprise Institute and Karen Smith of the Urban Institute. Biggs recently coauthored a Wall Street Journal op-ed on the subject with former Social Security Advisory Board Chair Sylvester Schieber. The op-ed included the eye-popping claim that “the most commonly cited measure of retirement income ignores at least 60% of the money that seniors receive.”

Is it true that some Census data underreports distributions from DC plans and IRAs? Is this news? Does it matter? Yes, no, and not much.

It is true that Current Population Survey (CPS) income measures don’t include lump sum withdrawals from retirement plans, limiting the usefulness of that survey in assessing the well-being of prosperous seniors. This isn’t an oversight by the Census Bureau but a reflection of the complicated relationship between income and wealth, and the arbitrary distinction it is sometimes necessary to make between the two in surveys that, for historical reasons, tend to focus on one or the other.

The recent flurry of interest in this issue appears politically driven, since it has been well-documented since at least 2008, and most retirement researchers are aware of it. (Schieber himself apparently raised this legitimate point almost two decades ago in Benefits Quarterlyunfortunately not web-accessible—though much else in that paper was quickly challenged by Social Security Administration researcher John Woods.)

Here, for example, is a quote from the Retirement Inequality Chartbook I coauthored with Natalie Sabadish: “A limitation of retirement-income measures is that they include regular distributions from retirement savings accounts but not lump sum withdrawals, giving an incomplete picture of available resources. Some retirees may have low reported incomes but large defined-contribution account balances, for example.”

In the Chartbook, Natalie and I also point out that aggregate assets in these accounts are large and growing. But these assets are extremely concentrated at the top of the distribution and will not provide significant retirement income for the vast majority of Americans. An article by John Turner, David McCarthy and Norman Stein in the current Journal of Retirement, for example, suggest that there may be as many as 165,000 retirement plans with average account balances greater than $3 million, though some may not qualify for tax benefits. This includes one 401(k) plan with three participants and an average balance of $240,000,000! Similarly, in the Chartbook, we found that 1% of households had more than $1.3 million saved in retirement accounts, and households in the top fifth of the income distribution accounted for 72% of assets in these accounts.

So yes, there is enormous wealth in 401(k) and IRAs. But the existence of outsize account balances has no bearing on whether ordinary workers can rely on savings in these accounts to finance a decent retirement, and the billions missing from the CPS data don’t amount to much for most workers. For example, two of the same Social Security Administration researchers cited by Ferrigno found in a later paper that while the average disbursement for senior households receiving them averaged $8,121 in 2009, only 19% of senior households received any disbursements at all, so that the overall average disbursement received among all senior households was much lower—$1,543. These are total disbursements as measured in the Survey of Income and Program Participation (SIPP), not just what’s missing from CPS data. Not surprisingly, Ferrigno chose not to cite these figures in his remarks.

If distributions from retirement accounts are on the order of $1,500 a year, how is it possible for CPS to understate “retirement income” by 60%? This suggests retirees are even worse off than we thought! Presumably, the Wall Street Journal op-ed is referring to retirement income from sources other than Social Security, though the wording is misleading to say the least.

Longer articles by Schieber and Billie Jean Miller in the Journal of Retirement and an industry newsletter confirm this interpretation. Schieber and Miller cite SIPP data, but rely primarily on Statistics of Income (SOI) from tax returns to make their case that CPS data significantly understates the resources available to senior households. Relying on Internal Revenue Service data is problematic for a number of reasons, including the fact that many households don’t file tax returns, and taxpayers don’t report their age. Though Schieber and Miller restrict their sample to taxpayers reporting Social Security income (not necessarily retirement benefits), they acknowledge that their measure includes some pre-retirement disbursements from retirement accounts.

Potentially more troubling, Schieber and Miller claim that rollovers “are not considered income for tax filing purposes thus are not included in taxpayers’ 1040 filings.” This seems to be contradicted by a joint Federal Reserve-IRS study published in 2005 that notes that while rollovers aren’t taxable, SOI measures of pension and annuity income include rollovers even if households don’t declare them on their tax forms. In other words, the IRS data apparently treat funds taken from one account and rolled over to another as “income.” This issue was earlier flagged by John Woods in his 1996 critique of Schieber’s research, who mostly focused on the problem in the context of data from the Bureau of Economic Analysis’s National Income and Product Accounts (NIPA), which Schieber also cites.

This begs the question of why Schieber relies on SOI and NIPA data rather than the Federal Reserve Survey of Consumer Finances (SCF), the standard source of information on savings in retirement accounts and other forms of wealth. You’d think Schieber would be eager to cite SCF data, since it shows not just what seniors actually withdraw from these accounts, but the total amounts households have saved in this and other forms. SCF data (not CPS data) underlie most serious attempts to gauge retirement preparedness, such as the Boston College Center for Retirement Research’s National Retirement Risk Index. The Center also used SCF data to estimate the $6.6 trillion Retirement Income Deficit cited by Senator Warren.

The most likely explanation is that while SCF data show that average account balances are substantial and growing, they are also highly unequally distributed, lending support to those who, like Michael Lind and President Obama, want to restructure upside-down tax subsidies to better target them to low- and middle-income households. For households in the peak saving years (55-64), the median account balance in 2010 was a measly $12,000, too small to make a meaningful difference in retirement. Though the median balance for households with savings in retirement accounts was $100,000, this not only represents the 70th percentile for this age group as a whole, it also translates into an annuity worth less than $5,000 a year.1

The long and short of this story is that one measure of senior incomes understates the resources available to some, mostly upper-income, households. However, the lump sum withdrawals from retirement accounts excluded from this measure don’t make a meaningful difference to most retiree households and have little bearing on the debate over expanding Social Security. If anything, more complete data reinforce the need to better target subsidies for retirement savings. Last but not least, those accusing others of “making a molehill out of a mountain” appear to be using highly questionable data to do the reverse.

1. Annuity calculated using the Thrift Savings Plan Retirement Income Calculator based on the following assumptions: the annuity starts when a couple are both aged 65, the interest rate is the current TSP annuity rate of 3%, and the annuity provides a 50% survivor benefit with rising payments to offset inflation. A 20-year annuity at the 3% interest rate translates into a $6,700 annual payment but loses value with inflation and offers no longevity protection.