In much of what we do and think, the concept of averages is so deeply embedded that we barely notice it is there. We look at averages in sports, medicine, financial markets, economic forecasting and political polling. But averages can be misleading—and in ways that are not always well-understood.

Look at the exhibit below. From two very different individuals, 24-year-old Mary and 60-year-old Phil, an average can be deduced— a 42-year-old we’ll call Average Joe. And yet there is not the slightest resemblance to either Mary or Phil.

The flaw of averages: Average Joe bears not the slightest resemblance to either Mary or Phil

CHARACTERISTICS OF 24-YEAR-OLD MARY, 60-YEAR-OLD PHIL AND A COMPOSITE AVERAGE AMERICAN, AVERAGE JOE

Source: J.P. Morgan Asset Management. For illustrative purposes only. For the full set of assumptions and model methodology, please see the appendix in the full whitepaper.

We all talk about averages, but how many people actually meet the definition? For example, we know that the average life expectancy for a 65-year old American woman is 86. How many American women die at age 86? Only 8.5%.

Average Joe and retirement stress tests

We explore this issue by looking at a composite average American. Average Joe is a 42-year-old college-educated man who earns $57,000 a year and has accumulated retirement savings of $43,000. We assume he saves 5.1% a year of his gross income, including a consistent 1.7% employer 401(k) match, from age 42 until he retires at 65. We also assume investment returns of 5%-6% a year. That’s our base case scenario, and it shows that Joe’s assets are wiped out by the time he turns 72, well before he dies at 90. To stave off that scenario, Joe would have to either more than triple his total savings rate (his own savings plus his employer match) to 16% or reduce his spending in retirement by 31%.

Then we suppose that Joe does triple his savings rate to 16% a year, but other variables change (exhibit below).

A series of stress tests reveals how changing circumstances can dramatically affect the odds of a successful retirement outcome.

BASE BREAK-EVEN SAVINGS, WITH EARLY RETIREMENT, WORSE INVESTMENT RETURN, WORSE HEALTH CARE SCENARIO

Source: J.P. Morgan Asset Management. For illustrative purposes only. For the full set of assumptions and model methodology, please see the appendix in the full whitepaper.

If he retires at 62 instead of 65, his money runs out at 74. If he faces increased health care expenses and a two-year stay in a nursing home, his assets are depleted at 79, and if his retirement portfolio generates average annual returns of 4%-5% instead of 5%-6%, his assets are gone at 83.

Planning for the edges: An individual perspective

As Joe’s story makes clear, changing life choices and circumstances can dramatically affect the odds of a successful retirement outcome. It’s not always easy to see. This is one of many reasons individuals planning for retirement may want to work with financial advisors who can help their clients understand the consequences of their choices and calibrate their tolerance for the spectrum of risks faced in retirement.

Stress testing an individual’s plan can be especially useful. Although financial advisors often use simulations to stress test market returns, simulation testing of individual behavior and circumstances is much less common, despite the fact that those factors can have a greater impact on experienced returns.

Planning for the edges: A defined contribution perspective

Employers look at the same terrain—a wide range of individual employees who often fail to understand how the choices they make long before they stop working can affect the lifestyle they experience in retirement.

Employers, working with their plan providers, can consider employees’ saving and investing behavior most broadly, not simply based on averages, when they refine their retirement plan designs and investment options. For example, when observing an increase in the plan’s average contribution rate, employers can determine if relatively few participants are driving that increase by looking at the median as well as the mean.

Recognizing the inadequacy of average assumptions in DC plan design, employers can help their participants make, in effect, “better than average” choices. Among the tools at their disposal: automatic enrollment, automatic contribution escalation, and target date funds (TDFs). In designing and refining TDFs, it can be particularly useful to move beyond average inputs by, for example, using Monte Carlo simulations to incorporate the variability of participant behavior. If the pace of salary growth slows by one percentage point, what impact would it have on the participant’s portfolio? Monte Carlo simulations can help answer that question.

Conclusion

Yes, we look at averages, but we look beyond them, too. In our view, an effective retirement plan, for a single individual or thousands, should incorporate the sometimes messy continuum of human choice and experience. That is why we firmly believe in designing for the edges as well as the average.