This is a reader request, based on an email asking how my recent Wall Street Journal op-ed – which deemed the so-called retirement crisis to be “phony” – jibes with the National Institute for Retirement Security’s (NIRS) study finding that America faces a $14 trillion “retirement savings gap” with a staggering 92 percent of working-age households falling short of their retirement savings goals. So here goes.
The answer is that NIRS dramatically overstates the amount by which households are undersaving for retirement. In fact, the better academic studies – written by well-regarded economists using more sophisticated methods and published in peer-reviewed journals – find a retirement savings gap of $1 trillion or less, by my estimates, out of the more than $93 trillion in total retirement plan assets and accrued Social Security benefits on which Americans rely in retirement.
But specifically, what drives the NIRS numbers upwards? I’ve identified four main assumptions in NIRS’s methodology that will overstate the true retirement savings gap.
NIRS sets target retirement incomes too high. According to the Social Security Administration, most financial planners recommend that a typical worker retire with an income equal to 70% of his pre-retirement earnings. Robert Myers, a former SSA chief actuary, recommended 70 to 75% for a middle income retiree, ranging from about 90% to a very low earner (making about 25% of the national average wage) to 60% for someone earning the Social Security taxable maximum wage each year. NIRS instead adapts a Fidelity Investments recommendation that workers reach retirement with savings equal to eight times their final salary. The Fidelity Rules might work for high-income workers, who make up Fidelity’s customer base, but not for middle and low-wage workers. Assuming Fidelity’s 8x final salary saving rule, for a middle income worker those savings on top of Social Security produce a retirement income equal to about 100% of the retiree’s final salary. That’s overkill.
NIRS assumes that individuals must begin saving for retirement at age 25 and counts them as undersaving if they don’t, even if they reach retirement with precisely the correct amount. The “lifecycle model” in economics predicts that most individuals won’t start saving for retirement until their incomes rise, which can mean saving doesn’t start until their mid-30s. And that seems to be what most workers do. But in the NIRS model, unless a worker saves on precisely the path dictated by NIRS, their model counts them as undersaving. Without recreating NIRS’s model it’s tough to assign a number to the effect. However, if I take identical workers where one starts saving at 25 and the other at 30 but both reach NIRS’ total target savings by retirement age, then sum the annual saving “shortfalls” for the latter worker, the total shortfall comes to about 8 percent of annual earnings. So it’s potentially a significant factor.
NIRS doesn’t properly account for Social Security benefits. NIRS’s model assumes that Social Security pays every retiree the same “replacement rate,” meaning benefits as a percentage of pre-retirement earnings. In reality, Social Security’s replacement rates are highly progressive. NIRS justifies its assumption by arguing that low earners require higher replacement rates, so as a result low earners need to save the same percentage of their income as high earners. Except they don’t. For instance, former SSA Chief Actuary Robert Myers calculated that a very low wage earner requires a total replacement rate of 90% of his pre-retirement earnings; his Social Security benefit is equal to about 83% of his final pay, getting him about 92% of the way to his goal. For a maximum wage earner Myers recommended a 60% total replacement rate, but the maximum wage earner’s Social Security replacement rate is only 31%, just half his required total. So low earners don’t in fact need to save as much as high earners. But by assuming that low earners need to save as much as the rich, this will skew the NIRS model to finding an overly-large “retirement savings gap.”
NIRS assumes pension underfunding will be solved 100% with benefit cuts. Here things get a bit strange. For households with defined benefit pensions, NIRS assigns the assets held in those pensions to the household and counts them toward the household’s retirement savings goals. But defined benefit pensions are significantly underfunded, particularly in the public sector. The NIRS methodology implicitly assumes that the entire funding shortfall will be addressed by reducing benefits to the level payable via current assets, when in reality only a very small part is likely to be. This is a particularly interesting assumption given that NIRS is the research arm of the defined benefit pensions industry and in other circumstances argues that DB plans are far superior to other forms of retirement saving. In other words, NIRS concludes there is a retirement crisis by assuming that DB pensions will enact massive benefit cuts, then argues that the solution to that retirement crisis is more DB pensions. If you say so. The correct way to do it would be to take the present value of total pension liabilities, then assign these values to households. That will give a much more realistic picture of how much households are likely to collect in retirement. Using Fed data, focusing on defined benefit pension promised benefits rather than plan assets would boost household retirement savings by almost $7 trillion.
I’m not saying this is an easy task. Judging retirement savings adequacy is really complicated and involves a ton of steps: getting good data on current savings, retirement incomes and retirement plan participation, which is hard. Projecting those values into the future, which is harder. And then deciding “how much is enough?,” which is crucially important but hasn’t been given nearly enough attention. But when an estimate of the retirement savings gap comes in 14 times higher than what the peer-reviewed studies conclude and when the factors pushing that number upward can be clearly identified, that estimate shouldn’t be used to make public policy.