GAO Report on Retirement Savings: Overall Gaps Identified, but the Focus of Retirement Security Reform Should be on the Uncovered Population



By Jack VanDerhei, EBRI

The Government Accountability Office’s new retirement analysis reviewed nine studies conducted between 2006 and 2015 by a variety of organizations and concluded that generally one-third to two-thirds of workers are at risk of falling short of their retirement savings targets.

However, many of these studies use a “replacement rate” standard: Most commonly, this analysis concludes that you need to replace 70–80 percent of your preretirement income to be assured of a successful retirement. This is a convenient metric to use to convey retirement targets to individuals—and no doubt provides useful information to many workers who are attempting to determine whether they are “on-track” with respect to their retirement savings and/or what their future savings rates should be. However, replacement rates are NOT appropriate in large-scale policy models for determining whether an individual will run short of money in retirement. Why?

Because simply setting a target replacement rate at retirement age and suggesting that anyone above that threshold will have a “successful” retirement completely ignores:

  1. Longevity risk.
  2. Post-retirement investment risk.
  3. Long-term care risk.

In fact, looking at just the first two risks above, if you use a replacement rate threshold based on average longevity and average rate of return, you will, in essence, have a savings target that will prove to be insufficient about 50 percent of the time. Of course, this would not be a problem if retirees annuitized all or a large percentage of their defined contribution and IRA balances at retirement age; but the data suggest that only a small percentage of retirees do this.

In contrast, EBRI has been working for the last 14 years to develop a far more inclusive, sophisticated, realistic—and, yes, complex—model that deals with all these risks. It’s our Retirement Security Projection Model® (RSPM), and produces a Retirement Readiness Rating (basically, the probability that a household will NOT run short of money in retirement).

Blog.JV.GAO-rpt.June15.Fig1Our most recent Retirement Readiness Ratings by age are shown in Figure 1 (left). Our baseline results do include long term care costs (the red bars), but we also run the numbers assuming that these costs are NEVER paid by the retirees (the green bars). This latter assumption is not likely to be realistic for many retirees, but we include it to show how important it is to include these costs (unlike many other models).

Even more important is Fig. 2 (right), which shows Retirement Readiness Ratings as a function of preretirement income AND the number of future years of eligibility for a defined contribution plan for Gen Xers.

Blog.JV.GAO-rpt.June15.Fig2Even controlling for the impact of income on the probability of a successful retirement, the number of future years that a Gen Xer works for an employer that sponsors a defined contribution plan will make a tremendous difference in their Retirement Readiness Ratings (even with long-term care costs included).

The evidence from EBRI’s simulation modeling certainly agrees with the GAO that a significant percentage of households will likely run short of money in retirement if coverage is not increased. However this is because we model all the major risks in retirement and do not simply assume some ad-hoc replacement rate threshold.

Moreover, using an aggregate number to portray the percentage of workers at risk for inadequate retirement income is really missing the bigger picture. The retirement security landscape for today’s workers can be bifurcated into those fortunate enough to work for employers that sponsor retirement plans for a majority of their careers vs. those who do not. In general, those who have an employer-sponsored retirement plan for most of their working careers appear to be well on their way to a secure retirement.

Perhaps the focus of any retirement security reform going forward needs to be on those who do not work for employers offering retirement plans and those in the lowest-income quartile.


Jack VanDerhei is research director at the Employee Benefit Research Institute.

EBRI’s Jack VanDerhei Interviewed in PIMCO Dialogue

The July 2010 issue of PIMCO’s DC Dialogue features an extensive interview with Jack VanDerhei, research director of the Employee Benefit Research Institute, on worker retirement-income adequacy and confidence. The full interview is online here.

In the article, “All Shook Up,” VanDerhei shares that only a small percentage of workers enjoyed the “good-old days” of defined benefit (DB) pension plans, and that, going forward, he anticipates defined contribution (DC) plans likely will replace a higher percentage of final pay for more people than DB plans have to date.
VanDerhei notes a significant decline in workers’ confidence in their ability to retire successfully since the recent market crisis. While many younger workers regained lost ground in their account values, some longer-tenured employees have yet to recover their losses. VanDerhei believes many of these longer-tenured workers are permanently shaken.

Jack VanDerhei, EBRI research director

In this environment, VanDerhei underscores the need to increase contribution rates in DC plans, encouraging auto enrollment and more rapid contribution escalation—measures that can help workers achieve their retirement goals.

Use of Fiduciary Benchmarks’ Retirement Readiness Index (FB-RRI) Could Lead To A Fiduciary Briar Patch

By Jack VanDerhei, EBRI Research Director

A recent news article (“Perceptions of Retirement Preparation,”) focused on differences in Americans’ expected preparations for retirement as presented by a recent for-profit start-up firm, Fiduciary Benchmarks, and the nonprofit Employee Benefit Research Institute (EBRI). This blog expands on several of the excellent points raised in that article.

Jack VanDerhei, EBRI research director

EBRI has published work dealing with retirement readiness for decades. For example, the Retirement Confidence Survey (RCS) has allowed workers and retirees to provide opinion data on what they believe their status to be for 20 years. EBRI developed a Retirement Readiness Rating in 2000. The EBRI/ICI 401(k) Accumulation Projection Model was constructed in 2002 to provide an assessment of estimated retirement accumulations for 401(k) participants under a variety of scenarios, and the Ballpark E$timate® interactive tool at offers the ability to provide individual input and get a deterministic view of whether you are on track to reach your retirement goal, and, if not, how much more needs to be saved to reach that goal by retirement age. The EBRI Retirement Security Projection Model (RSPM) was presented in the November 2003 and February 2004 EBRI Issue Briefs with data based assessments of retirement readiness. The newest Retirement Security Projection Model® Retirement Readiness Rating will be published in early summer 2010.

Before I provide a critique of the Fiduciary Benchmarks’ Retirement Readiness Index (FB-RRI), let me begin by stating that any comparison of RRI with the Retirement Confidence Survey (RCS) is not a legitimate comparison for the following reasons:

1. RCS is a survey of confidence among workers and retirees with respect to their perceived ability to having enough money for a comfortable retirement. In contrast, RRI alleges to be a measure of “how well workers are preparing for a secure retirement.”

2. Although RRI uses the word “workers,” it appears that they are (at least currently) limiting their analysis to participants of defined contribution retirement plans. RCS surveys all workers whether or not they are currently participating in a retirement plan. These are two very different groups.

Upon review, those that use the tool from Fiduciary Benchmarks could be opening themselves up to a fiduciary briar patch of problems by telling employers that their participants, or the participants themselves, are on track for a “secure retirement” when they are not.

With respect to the critique of FB-RRI, I will base my comments on the information provided by Fiduciary Benchmarks on page 7 of their description of the RRI, “How does the RRI work: key inputs”:

1. Required replacement ratio. While replacement ratios and similar targets (e.g., multiples of final earnings at retirement age) can be useful metrics for basic projections, they simply cannot deal with many of the risks inherent in retirement income adequacy. For a comprehensive explanation of why this is too limited to compute “retirement readiness” see VanDerhei (2006); briefly, all that replacement ratios typically attempt to do is compute the equivalent amount individuals will need in retirement after adjusting for the differences in pre- and post-retirement taxes, savings, and age-specific expenses (such as health care expenditures). This says little, if anything, about an individual’s readiness for retirement.

2. Retirement age. If I had to pick a single age, SSNRA (Social Security normal retirement age) may be the best available. However, a near majority retire ahead of the SSNRA and a significant percentage of those for health reasons.

3. Life expectancy. Using “life expectancy” (even the “conservative” assumption of a female employee) is far too risky. Saving enough for “life expectancy” in essence means that (with the exception of the rare occurrence of an employee annuitizing ALL their savings at retirement age) approximately 50 percent of the time, the individual will outlive their savings. For a full analysis of what needs to be included in a study to provide the employee with a better than a 50–50 chance of sufficient money, see VanDerhei and Copeland (2003).

4. Starting Age
5. Starting Wage
6. Starting Account Balance
All three of these “data points” share the same problem: they are only valid for a single stylized circumstance and will not even come close to representing the situation for the vast majority of actual participants in a plan. Perhaps the most troubling is the starting account balance. Since 1996, EBRI and ICI have published annual reports of tens of millions of INDIVIDUAL account balances (not simple plan averages), and it is quite obvious the distribution of balances are very skewed (even adjusting for age and tenure).

Moreover, the implicit assumption in this methodology appears to be that the “average” employee will continue to work with the same employer for the remainder of their career or at the very least have the quite unlikely prospect of changing jobs and ending up with another plan with exactly the same distribution. Also, as is well documented, when employees change jobs they often cash out their account balances instead of saving them.

7. Average participant and employer contributions
This also suffers from several limitations:
• Assuming the plan is NOT automatic enrollment (they do not seem to bifurcate the plans as any serious analysis would), we know from several studies that there is a wide distribution of contribution rates, that these rates tend to increase with age, and that many employees will stop contributing at least temporarily during their careers (especially if the employer suspends the matching contribution).
• If the plan IS automatic enrollment, there should be some type of recognition of the trend to automatic escalation of contributions.

8. Return on investments
I have no idea why “an analysis of more than 61,000 historical holding periods” would suggest that Treasury rates plus 50 basis points would be a realistic proxy for return on investments; however, EBRI/ICI analysis from 1996–2008 documents quite extensively the distribution of asset allocations of millions of 401(k) participants at all age ranges. Unless one is essentially assuming no equity premium and very low or zero volatility going forward, this would appear to be an ad hoc proxy with little, if any, empirical justification.

9. Inflation
I am not sure why one would use a “long-term inflation rate assumption” for wage growth. Even if we are willing to ignore all the empirical evidence for age/wage profiles, I still know of no credible forecasts that assume wage growth will be equal to inflation.

For additional information on how a study of 401(k) participants could be conducted that would correct for many of the limitations enumerated above, please see:
Holden and VanDerhei (2002) for a pre-automatic enrollment analysis
Holden and VanDerhei (2005) for an expansion of the techniques to include automatic enrollment
VanDerhei (2007) for an expansion of the technique to include automatic escalation of contributions for automatic enrollment plans

As with RSPM, I assume that the FB-RRI is a work in progress and will change over time. As it does, I look forward to reviewing it.

—Jack VanDerhei, EBRI

• Buckner, Gail, “Perceptions of Retirement Preparation,” FoxBusiness, March 22, 2010.
Fiduciary Benchmarks, Retirement Readiness Index, Portland, OR.
• Holden, Sarah and Jack VanDerhei (2002). “Can 401(k) Accumulations Generate Significant Income for Future Retirees?” EBRI Issue Brief no. 251 (November 2002).
• Holden, Sarah and Jack VanDerhei (2005). “The Influence of Automatic Enrollment, Catch-Up, and IRA Contributions on 401(k) Accumulations at Retirement.” EBRI Issue Brief no. 283 (July 2005).
• VanDerhei, Jack and Craig Copeland (2003). “Can America Afford Tomorrow’s Retirees: Results From the EBRI-ERF Retirement Security Projection Model,” EBRI Issue Brief no. 263 (November 2003).
• VanDerhei, Jack.2006. “Measuring Retirement Income Adequacy: Calculating Realistic Income Replacement Rates,” EBRI Issue Brief, No. 297 (September 2006)
• VanDerhei, Jack.2007. “The Expected Impact of Automatic Escalation of 401(k) Contributions on Retirement Income.” EBRI Notes, September 2007 (pp. 1–8).