Our new measure emphasizes criteria that plan officials can control.
Plan sponsors and providers understandably want to measure the effectiveness of the defined-contribution plans they offer; plan participants should want to monitor their plans all the more. But not all evaluation methodologies measure up. A retirement plan should be evaluated on criteria over which those responsible have some authority or opportunity for influence. For example, if a team of 55-year-olds enters a plan with zero retirement savings, that shouldn’t say anything about how well the plan is run or has prepared its participants for retirement.
But the most common defined-contribution-plan measures in use do just that, allowing two factors in particular—participant age and plan tenure—to have a significant impact on analyses of plan effectiveness.
In today’s dynamic workplace, the average employee tenure is approximately five years. Tenure, which plan officials have little if any control over, has a significant impact on the perceived success of a plan and its participants. For example, a plan with a high savings rate but low tenure may appear to be ineffective if participant balances were especially low when employees joined the company. Judging the plan on what its administrators can influence (the high saving rate) instead of what they cannot control (turnover) would yield a much more meaningful rating.
To address these shortcomings, we created a methodology for estimating the effectiveness of a plan at the individual participant level, which in turn can be rolled up into what we call a Retirement Plan Effectiveness Score. We use robust estimation techniques to determine the appropriate retirement plan “target balance” based not only on age and plan tenure but also gender, compensation, and other variables. Comparing the participant’s current balance to a customized target balance yields the relative funded status, and combining these individual participant metrics produces the Retirement Plan Effectiveness Score.
Our approach overcomes the weakness of the most common approach to evaluating retirement plan effectiveness—estimating the probability of a participant successfully accomplishing his or her retirement goal—by focusing on factors that plan administrators can influence and controlling for factors such as age and plan tenure. The new framework allows for meaningful comparisons across plans and through time.
Tenure, Age, and Out-of-Plan Assets
Using widely available tools, plan sponsors cannot accurately answer the basic question, “Will our participants have a successful retirement?” Common “success” metrics—such as retirement replacement rates, which is the percentage of a participant’s income he or she will be able to replace at retirement, and a success analysis, which estimates the probability the participant will be able to maintain the same standard living for retirement—require a significant number of assumptions that are beyond the influence, and knowledge, of the plan sponsor. A variety of inputs are required to estimate retirement readiness, some of which the plan sponsor has information on (such as current plan balance, compensation, plan savings, etc.); however, most plan sponsors have no information about a participant’s savings or assets outside the plan.
Tenure, which varies significantly across employers, plays a key role in measuring plans. According to the Bureau of Labor Statistics, the median tenure for wage and salary workers in the United States was 4.6 years in January 2014. Although not perfectly equal, employment tenure and plan tenure tend to be similar.
To study the effects of tenure on retirement plans, we obtained data for 186 401(k) plans with at least 50 active participants, representing 135,322 participants. This dataset came from one record-keeper, but there is nothing to suggest any unique biases or that it is not a representative sample of typical plans and participants. Analyzing this data, we found the median and average years of participation are eight and six years, respectively. The average is greater than the median due to the positive skewness of the distribution. (When people stay, they stay longer than would be expected in a normal distribution.) There is clearly a significant deviation in employee tenure across plans, with a standard deviation of approximately three years.
An important driver of average plan tenure is age; older employees tend to have longer tenures on average and can skew the tenure distribution. However, the relationship is far from perfect. There are some plans with older employees and low tenure and some plans with younger employees and longer tenure.
Logically, age and tenure both play a role in how significant out-of-plan assets are for a given participant. Older employees and lower tenure employees are more likely to have assets outside the defined-contribution plan. In terms of how significant out-of-plan assets are, according to the Investment Company Institute’s “U.S. Retirement Market Report” for the fourth quarter of 2014, IRAs held more assets ($7.4 trillion) than did defined-contribution plans ($6.8 trillion). Therefore, any kind of retirement readiness analysis completed by a plan sponsor (or by a consultant for the plan sponsor) is likely to be based on incomplete information, which makes it difficult to reach robust conclusions.
Plan Effectiveness Measurement Approach
We sought to determine an appropriate retirement plan balance—a target balance— for each individual participant of a defined-contribution plan, based on that participant’s age, compensation, plan tenure, as well as things like employer match and profit-sharing contributions. The target balance is the amount we calculate the participant should have accrued in the retirement plan to be on track to maintain the same pre-retirement lifestyle during retirement. The target balance is estimated on the unique attributes of the participant in question and is less affected by information unknown to the plan sponsor (such as assets outside the plan).
EXHIBIT 1 illustrates what target balances would be for participants of different ages, with varying years of plan participation and annual compensation. To determine whether a participant— or the entire plan—is on track, the participant’s balance is divided by the estimated participant target balance, resulting in a metric we call the “target balance ratio.” Any rollover contributions should be excluded from the comparison, because these are monies that were accrued outside the plan. The target balance ratio is similar to a funded status of a defined-benefit pension plan. A target balance ratio of 100% suggests that the participant has accumulated enough funds while participating in the plan to be on track to meet his or her retirement goals. A target balance ratio greater than 100% suggests that the participant is likely ahead on saving for retirement, while a ratio less than 100% indicates the participant is lagging behind.
Once the target balance ratio has been estimated for each participant, each ratio can be aggregated to determine the overall Retirement Plan Effectiveness Score. Plan effectiveness scores can be compared across plans and over time. Any participant-specific errors should cancel each other out when averaged across a reasonably large plan.
In addition to looking at a single ratio, plan officials can look at ratios for different cohorts, such as gender, age, compensation, and years of plan participation to identify areas of relative strength and weakness. This type of analysis allows a plan sponsor to target marketing messages at certain participants who may not be saving enough.
Sample Plan Results Comparison
We tested our methodology on a single plan and contrasted the results of a typical replacement- rate analysis versus results from our proposed target balance analysis. Our sample plan had nearly 4,000 participants and $200 million in assets. The median participant age was 47, and the median tenure was nine years.
Using the available plan data, we calculated the estimated replacement rate and target balance ratios for the participants. PANEL A and PANEL B of EXHIBIT 2 display the median replacement rates and target balance ratios by age. The absolute values on the vertical axis are not directly comparable with one another, because we are looking at two unique estimates of median success. Here, one should compare the overall picture of success created by each panel, noticing that they lead to different conclusions.
The replacement-rate results in PANEL A are typical across retirement plans. Younger participants generally have higher forecast replacement rates because it’s assumed they will stick with the plan until retirement; therefore, they are able to accumulate substantial savings over their working careers. In contrast, older participants (over the age of 55) tend to have low replacement rates, on average, unless the older participants in the plan have significant tenure or tend to roll outside accounts into the plan. The obvious key piece of missing information for older participants is assets saved for retirement outside the plan (as well as a shorter assumed period to save and prepare for retirement). Broadly speaking, the results of the replacement-rate analysis in PANEL A are driven by age.
The target balance ratio analysis in PANEL B of EXHIBIT 2 tells a very different story. Our younger participants in PANEL A, who appear to be the most on track for a successful retirement, are the employees with the lowest target balance ratios in PANEL B. Here, they appear to be the group that needs to increase savings the most. Also, while there is a slight drop in the target balance ratio for older participants, they are still doing well, with target balance ratios around 70%. This drop is also not directly related to years of plan participation because, unlike replacement rates, the target balance analysis incorporates how long the participant has been active in the plan.
EXHIBIT 3 breaks down participants into the 10th, 25th, 50th, 75th, and 90th percentiles. As before, PANEL A shows replacement rates and PANEL B shows target balance ratios.
EXHIBIT 3 demonstrates a key shortfall in the replacement-rate metrics (in PANEL A). Note how replacement rates converge at older ages, yet target balance ratios (in PANEL B) don’t converge. This is because replacement-rate analysis assumes asset accumulation, while the target balance approach does not. The replacement- rate analysis assumes the participant’s current plan balance (plus Social Security benefits) is the only source of retirement income, ignoring the fact that it does not account for outside assets. Meanwhile, the target balance approach considers the balance only in the context of the current plan’s influence—saving and asset growth during the participant’s tenure.
The replacement rate and target balance ratio metrics are somewhat related. For example, participants who are saving more for retirement will have both higher target balance ratios and higher replacement rates. The overall relation between the two metrics is relatively strong for the sample plan in our study, with an R-squared of 29.73. The positive coefficient suggests that both tend to increase at the same time for a given participant, which again should not be surprising given the fact that higher levels of savings would positively affect both metrics. The dispersion of target balance ratios is wider than that of replacement rates. This can largely be attributed to the “buffering” effect that Social Security retirement benefits have on replacement-rate calculations.
Next, we looked at how target balance ratios vary across plans and participants. We once again focused on our data set of 186 defined-contribution plans representing a combined 135,322 participants. We calculated target balance ratios and replacement rates for individuals and plans.
We found that the average and median target balance ratios by plan were 0.89 and 0.84, respectively. Sixty-six percent of plans had target balance ratios less than 0.95, 12% had values between 0.95 and 1.05, and 22% had values greater than 1.05. The average and median target balance ratios by participant were 0.91 and 0.8, respectively.
Similar to plans, most participants (60%) had target balance ratios less than 0.95, while 6% had values between 0.95 and 1.05, and 34% had values greater than 1.05.
In aggregate, these findings suggest that while some plans (and some participants) are saving enough for retirement, most are not.
The distribution of replacement rates by plan median and participant were more consistent than the target balance ratio distributions. The replacement rates suggest a less favorable level of funding than the target balance ratios. For example, 83% of plan medians and 66% of participants had a replacement rate less than 0.95, 11% of both plans and participants had a replacement rate between 0.95 and 1.05, and only 6% of plans and 23% of participants had a replacement rate greater than 1.05.
A New Retirement Measure
Common retirement readiness metrics are not useful at measuring retirement plan effectiveness because they are driven by plan demographics rather than by things plan officials can influence. We introduce a new framework that attempts to control for key factors such as average tenure and average age but uses other metrics such as compensation and gender.
Our system allows for meaningful comparisons across plans and through time.
More specifically, using robust estimation techniques we estimate the appropriate retirement plan target balance to be on track for a successful retirement. The participant balance can be compared to the target balance to determine the relative funded status, and these participant metrics can be combined across the plan to yield the Retirement Plan Effectiveness Score.
Plans should calculate their Retirement Plan Effectiveness Score regularly and enact practices that aim to improve it over time.