Statement of Jonathan Skinner, Ph.D., 
John French Professor of Economics,
Dartmouth College, and Professor of Community and Family Medicine, 
Dartmouth Medical School, Hanover, New Hampshire

Testimony Before the Subcommittee on Oversight
of the House Committee on Ways and Means

Hearing on Retirement Security and Defined Benefit Pension Plans

June 20, 2002

My name is Jonathan Skinner, and I am the John French Professor of Economics at Dartmouth College in Hanover, NH.  As is well understood, there has been a dramatic shift during the last several decades away from defined benefit (DB) towards defined contribution (DC) pension plan, most notably 401(k) plans.  Has the shift toward 401(k) plans enhanced or detracted from American workers’ financial security at retirement? In my comments below, I will focus on four major criticisms of DC and 401(k) plans that seem to suggest that future retirement security of current American workers could be jeopardized by this trend:

• Compared to DB plans, 401(k) plans cannot provide a comfortable pension at retirement, even when workers are contributing the recommended fraction of their salary.
• Many workers do not contribute even the recommended amounts, or eschew contributing altogether.  When they switch jobs, they spend their 401(k) assets on houses, boats, or travel.
• Defined contribution plans paying out lump sums at retirement can be spent down quickly by unsophisticated retirees, rather than providing a fixed annual income (or annuity), as in defined benefit plans. 
• The Enron debacle has focused attention on what is perhaps the most serious charge against 401(k) plans – that they are just too risky for use in retirement planning.  The risks come not simply from Employee Stock Ownership Plans (ESOPs), but also from uniformed or overly aggressive investment choices, or just weak overall stock market and bond returns. 

In a research project, funded in part by the National Institute on Aging, Andrew Samwick of Dartmouth College and I have considered each of these four concerns.^[1]  A systematic analysis of DB and DC plans has proven difficult in the past because of the varying characteristics of DB plans across firms, and varying contribution rates, matching plans, and investment decisions for workers in DC plans, again across the universe of firms offering such plans.  This variability has made it difficult to generalize about “typical” DC or DB plans.  Comparisons are further complicated by the continued evolution of both DB and DC plans during the past several decades.

We addressed this shortcoming by using data from the Federal Reserve Board of Governors’ Surveys of Consumer Finances (SCF) from 1983 to 1998, and, in particular, the Pension Provider Surveys (PPS) that accompanied the 1983 and 1989 SCFs.  The Surveys asked detailed information about family assets, income, 401(k) contributions, and demographic information, while the pension supplements provided information about nearly one thousand pension plans in 1983, and about half that number in 1989.  For each survey respondent who was covered by a pension, an attempt was made to obtain the summary plan descriptions from the employer or union. These descriptions were then coded into computer software by the Survey Research Center at the University of Michigan.  We used a substantially enhanced version of this software to compute pension entitlements.   The detailed information has been merged with more recent data from the SCF on characteristics of 401(k) through 1998.

Briefly, Samwick and I found that 401(k) plans are not the ticking time bomb that many fear.  If some 401(k) plans do not provide adequately for retirement, then neither did many DB plans. These findings do not depend on the extraordinary gains in equity markets during the 1990s because in our analysis, we excluded all equity returns after 1990, a year in which the Dow Jones Industrial Average never closed above 3000.  The result is robust to lower simulated equity rates of return, to job mobility where workers “spend down” some part of their balances, and to the presence of workers contributing little or nothing to their accounts. 

To provide a flavor for the findings of our research, I consider each of the four criticisms of DC plans in more detail.

1.  Compared to DB plans, DC plans cannot provide for a comfortable retirement pension  

There are a number of reasons why similar workers may receive different pension benefits at retirement beyond the simple reason that one worker has a DB plan and the other a DC plan.  Individual workers may get different pension benefits from the same pension plan because of different earnings, asset allocation, or investment returns over their lifetime.  Within DB plans, there is remarkable variability in the nature of plans across firms, in particular with regard to how benefits are “backloaded” with respect to earnings in the final 3 or 5 years of service.  To capture this variability, we simulated a wide variety of different worker earning “histories” through age 62.  In other words, we created a group of nearly 40,000 synthetic “workers” with complete earnings histories, and asked how this group of workers would have fared over their lifetime with the set of defined benefit plans available in 1983.  Then we took the same synthetic “workers” with the same earning histories and provided for them randomly chosen 401(k) plans from the universe of plans offered in 1995 as well as in other years.  These 401(k) plans are also subject to risk from stock and bond returns; this was done by randomly assigning historical rates of return from stocks and bonds during the 1900-1990 period.  Thus it was possible for some 401(k) enrollees to experience the dismal stock rates of return from 1932 and 1933 multiple times during their employment.  Summary information about our results is presented in Table 1, with all values expressed in $1995 dollars, and time trends are shown graphically for 401(k) plans in other years (1992, 1998) for the median worker in Figure 1.  

In the first column of Table 1, mean 1983 DB pension benefits are $13,917, in contrast to the mean expected annuitized annual benefits of $30,880 for typical 1995 401(k) plans in Column 2.  Recall that these expected DB and 401(k) pension benefits are for workers with identical earnings histories, and do not reflect the fact that earnings were higher in 1995 than in 1983.  Median pension benefits for DB plans are $9,227 while median pension benefits for 401(k) plans are $12,694.  Even among those with the very worst pension benefits -- at the 10^th percentile --  401(k) benefits, $1,890, are higher than those for the DB plans, $1,638.  In short, while 401(k) plans do not provide large pension benefits for the bottom 10^th percentile of the working population because of low earnings, poor stock returns, or low contribution rates, so also does the bottom 10^th percentile of DB pensions provide minimal benefits.^[2]  By contrast, 401(k) plans provided far more generous retirement benefits for the vast majority of workers.  The 1989 DB pension benefits are on average more generous, perhaps because of the attrition of weaker DB plans during this period, but even with these more generous benefits, the universe of 1995 401(k) plans still dominates (see Samwick and Skinner, 2001, op. cit., Table 4B).

In many respects, the most generous plans are those that combine both DB and DC plans, as shown in Columns 3 and 4.  Replacing combined DB and DC plans with just 401(k) plans results (not surprisingly) in a small average decline, and a larger median decline, in overall benefits. For these workers, 401(k) plans were used to supplement existing DB plans.   

Figure 1 shows the secular change in the generosity for the median worker in both DB, DC, and 401(k) plans.  (Non-401(k) DC plans tended to be more generous than just 401(k) plans.)  Median expected payments from DB plans rose between 1983 and 1989, as noted above.  Other researchers have not found increased generosity in DB plans since 1989, however.^[3]  At the same time, the average expected generosity of 401(k) plans has been increasing through the 1990s, because of both higher contribution rates by participants and a larger share of investments in equity rather than in bonds.

To return to the first criticism of DC plans: we cannot say whether they will provide an “adequate” level of retirement security for workers in the future.  But as Table 1 shows, the typical DC plan is expected to yield a higher return than the typical DB plan for a broad range of earnings and investment experiences.

2.  Workers are not even contributing the recommended amounts to their DC plans, and when they change jobs, they spend down the accumulated 401(k) assets on houses, boats, or vacations. 

This combination of low contribution rates and spent pension balances when workers change jobs would appear to predict serious problems for the future of American Workers. In 1993, Myron Mintz, chair of the Pension Benefit Guaranty Corporation (PBGC) stated  “A whole generation of people are going to wake up years from now and say, ‘God, I wish I had known when I was 32 that I should have been putting this money in.’"^[4]

It true that some workers do not roll over their DC plan balances when they move, but instead spend them on houses, cars, or vacation travel, particularly when the balances are small.   However, rollover rates are considerably higher when the amount of the balance is large. Furthermore, Andrew Samwick and I show that even when workers spend a large fraction of their DC plans when they change jobs, they still do as well or better than DB plans.  The reason is that DB plans often have vesting provisions, are not adjusted for inflation, or are tied to the last few years of work, meaning that they are worth little or nothing to workers who change jobs.  At least under DC plans, workers get to spend their money on things they want to, rather than having it revert back to the employer entirely.

What about workers who neglect to contribute to 401(k)s altogether? It is true that a large number of workers fail to contribute to their 401(k) plans.  However, the ones who fail to contribute typically have an alternative plan, such as a DB pension; the 401(k) is simply a supplemental plan.  In a related study, we estimated that between 2-4% of all workers are offered 401(k) plans to which they fail to contribute, and have no alternative pension plan.  By contrast, nearly half of all workers during the period of analysis did not even have the option of a pension, since their employees offer no pension coverage.^[5]

To sum up, it may be true that some employees with the option to save through defined contribution plans may not be saving enough for retirement; pension benefits in the bottom two deciles are very low.  But as we have shown, defined benefit plans also fell short at providing enough for retirement; on net, it appears that DC plans may do a better job particularly with regard to workers who switch from job to job.

3.  Defined Contribution plans typically pay benefits as lump sum disbursements at retirement rather than providing annuitization.

One concern with lump sum disbursements is that they may be invested poorly by the recipient, or spent too quickly particularly if the recipient (or spouse) lives for an unusually long time.  By contrast, DB plans typically pay benefits as an annuity that insures the recipient against variation in longevity and prevent the recipients from spending their wealth “too” quickly.  One option for retirees with large lump-sum payments is to roll them into Individual Retirement Accounts (IRAs).  This approach continues to defer tax payments and allows the individual to continue to accumulate (if they wish) until age 70 ½, at which point beneficiaries must withdraw according to an actuarial schedule.  While such an approach does not provide an annuity in the way that DB pensions do (it is still possible to run down one’s IRA account), it allows for smoothing out the 401(k) assets over time.   Alternatively, retirees could put the (after-tax) 401(k) dollars directly into a private annuity.  I am somewhat concerned about the lack of annuitization for households with large DC pension balances, particularly with regard to benefits for widows who may outlive the household’s assets.  Annuitization could therefore be encouraged through the use of default provisions to roll a fraction of DC balances into annuities. 

4. Defined contribution and 401(k) plans force workers to face too much uncertainty regarding their pension benefits

There are a variety of risks facing workers with 401(k) pension plans.  In Samwick and Skinner (2001, op. cit.) we considered several sources of risk: low overall stock and bond returns, inadequate contribution rates, and portfolios that were 100% in stocks or 100% in bonds.  That study did not consider the problem of employee stock option plans, or ESOPs, an issue to which we return below.   As shown in the simulation model reported in Table 1, we found that even with this investment risk, 401(k) plans in 1995 were no more risky than DB plans in 1983, and in fact provided greater pension security for nearly every retiree. 

The result may appear surprising, but the intuition is straightforward.  There are two sources of uncertainty; stock market returns, and the worker’s future earnings.  It turns out there is considerable variation in earnings, even among mature men.  Promotions, bonuses, or ill health can have profound effects on earnings in the last 5 or 3 years of service, the years upon which DB plan benefits are often based.  As well, for workers in their 50s, particularly for women, the burden of caring for aging parents may cause withdrawal from the labor market at a time when there is the greatest pension return to continued work experience.  By contrast, the typical 401(k) plan entails annual contributions over one’s entire work history, so the resulting balance reflects an average of earnings instead of the final few years.  And while some years in the stock market may cause the DC balances to jump or fall by 20 percent or more, what is important is the (geometric) average of all the stock market returns over a lengthy period.  Thus the risk faced by 401(k) enrollees appears to be no greater than that faced by DB enrollees.

Of course, recent history tells us that there are many other reasons why 401(k) pension plans can yield very poor returns.  Even before Enron, 10,000 employees of Carter-Hawley-Hale were required to put their 401(k) money into company stock; the company later declared bankruptcy.^[6]  Similarly, when Color Tile declared bankruptcy in the mid-1990s, workers found themselves out of a job and without pension benefits; one disgruntled Color Tile worker commented “I would never join a 401(k) plan again. ”^[7]  As well, unethical or uninformed employers may create “malformed” 401(k) plans with  “too few choices, arbitrarily set contribution limits, hidden fees, and other traits that can, ... at worst, seriously hobble workers’ efforts to prepare for retirement.”^[8]      Finally, there is a vast reservoir of employee ignorance about how to direct their self-directed pension funds.  Based on a survey conducted by Towers Perrin, one third of respondents thought there was no risk in investing in bonds, (despite the sensitivity of bond prices to nominal interest rate changes), while 40 percent of those in a self-directed saving plan did not know how their pension assets were invested.  These shortcomings are not intrinsic to 401(k) plans themselves, however.  They are the consequence of imprudently administered 401(k) plans that leave the worker uneducated and with poor options.

Policy Implications and Conclusions

I have drawn on recent research with Andrew Samwick to address criticisms of emerging defined contribution plans such as 401(k)s.^[9]   We found that DC plans certainly have their faults, but their faults are probably less severe  than those found in existing DB plans.  For example, we found that DC plans are expected to entail less risk and provide generally better pension benefits compared to DB plans.  And while workers may spend some of their lump-sum 401(k) disbursements when they leave their jobs, at least they get to spend it on something they like.  By contrast, when workers leave firms with DB plans, they either get nothing, or their pension payments beginning at age 65 will be seriously eroded by inflation. 

Nonetheless, it is clear that there is a great deal of room for improvement in the design of DC pension plans, for example by improving the rollover rates for DC plans when workers switch jobs.  The Economic Growth and Tax Relief Reconciliation Act of 2001 (EGTRRA) mandated that rolling the 401(k) to an IRA should be the default for balances of between $1,000 and $5,000, for example.  While changing defaults have been shown to have important effects on individual behavior, one could strengthen this policy further by requiring workers to roll over at least 50 percent of their 401(k) balances in excess of $1,000 into a qualified IRA account.  This compromise allows access to funds during a potentially difficult time such as unemployment, but still preserves at least half of the retirement nest egg.  Another option is to encourage rolling some fraction of the 401(k) balance at retirement into a qualified annuity fund.  And it goes without saying that the lack of legal enforcement leading to “malformed” pension plans with inappropriate investment choices and restrictions, and the lack of worker financial education, are the real time bombs threatening public (and legislative) perception of DC plans.

Finally, I return to a major shortcoming of the pension system more generally; namely, that even by the time workers reach their 50s, one-third of households do not have any pension wealth at all (Gustman et al, 1999, op. cit.).    Thus, a real concern with pension plans is how to make it easier for firms to offer pensions to workers currently not covered by any plan.  In this arena, 401(k) plans offer clear advantages over DB plans with their greater portability and fewer administrative burdens.  Defined contribution pension plans may not provide the solution to the problem of low retirement saving among many workers.  But we have argued they have the potential to play an important role to improve workers’ financial security at retirement.

Source: Samwick and Skinner, 2001 (op. cit.)

Figure 1:  Median Pension Benefits, by Type of Plan and Year

[1] “How Will Defined Contribution Pension Plans Affect Retirement Income?”, Dartmouth College (September 2001).  http://www.dartmouth.edu/~samwick/dbdc200110.pdf.  Also see our paper  “Abandoning the Nest Egg? 401(k) Plans and Inadequate Pension Saving.” in Sylvester J. Schieber and John B. Shoven (eds.) Public Policy Toward Pensions. Cambridge: MIT Press, 1997, 197-217.

[2]These very low DB benefits may reflect Social Security “offsets” in which the DB payment is reduced dollar-for-dollar as Social Security monthly benefits rise. 

[3]Gustman, Alan L., Isha Archer, Mariam Malik, and Toinu Reeves. “Pension Changes from 1990 to 1995 Based on Data from Watson Wyatt Reports on Pensions for the Largest Fifty Firms,” mimeo, Dartmouth College (1998).
[4]Vise, David A. "A Pensionless Future? Workers at Risk as Firms Abandon Plans." The Washington Post. May 13, 1993, Section A, p. 1.

[5]Samwick and Skinner, 1997, op. cit.  Of course, this does not mean that only half of workers will ever have pensions; indeed by the time they are close to retiring, nearly two-thirds of households have some accumulated pension wealth (Gustman, Alan L., Olivia S. Mitchell, Andrew A. Samwick, and Thomas L. Steinmeier, “Pension and Social Security Wealth in the Health and Retirement Study,” in J.P. Smith and R.J. Willis, eds., Wealth, Work, and Health: Innovations in Measurement in the Social Sciences. University of Michigan Press, 1999.)

[6]Kahn, Virginia Munger, “The Perils of Company Stock for Retirement,” The New York Times (March 16, 1997): Section 3, page 6.

[7]Schultz, Ellen E., “Color Tile’s 401(k) Plan Runs Aground,” The Wall Street Journal June 5, 1996, Section C, p. 1.

[8]Johnston, David Cay, “Investing it: Building a Better 401(k),” The New York Times (October 22, 1995): Section 3, page 1.

[9]Samwick and Skinner, 2001, op. cit.