The following guest post was contributed by Andrew Biggs. He has studied the issue of retirement savings for a couple of orders of magnitude longer than I, so I wanted to give him the opportunity to outline his perspective on the topic. He regularly blogs on his own blog and, along with about four dozen other people, over here.
After our exchange regarding Tuesday’s blog on The Retirement Problem in the Washington Post (which started over at AEI’s Enterprise blog and continued here), James generously offered to let me guest-post my thoughts on Americans’ level of preparation for retirement. Overall I’m not so pessimistic, although there are surely problems that must be addressed. But most of the detailed research out there points to problems, but not a crisis.
Both James’s analysis and my own response were built on relatively simple projections using stylized workers who pay into Social Security and participate in 401(k) plans. These illustrations are useful for fleshing out basic issues – plus, in this case, finding how the SSA’s online benefit calculator may have skewed some of the results.
But the best research on retirement preparedness is more involved than this. Most analysis of current retirees uses survey data, such as from the Health and Retirement Study (HRS), the Survey of Income and Program Participation (SIPP), the Fed’s Survey of Consumer Finances (SCF) and the Current Population Survey (CPS). Each survey has strengths and weaknesses.
In addition, broader models of the population are built using this survey data. These models allow for simulations of how policy changes affect current retirees, as well as projecting the population into the future. Such comprehensive models include the Social Security Administration/Urban Institute MINT (Modeling Income in the Near Term) model, the Congressional Budget Office’s CBOLT (CBO Long Term) and the Policy Simulation Group’s PSG suite of models, used by the Government Accountability Office and the Department of Labor for Social Security and private pension projections. While these models, like any others, rely on assumptions regarding a large number of factors, they are also the most closely scrutinized to ensure these assumptions are consistent with current trends.
I’ll first run through what some of my own work has found, and then highlight some work from elsewhere. This paper, written with Glenn Springstead when I was at SSA, used the MINT model to analyze replacement rates for current retirees and, using projected data, for individuals retiring in the 2040s. We looked at income from a wide variety of sources, including Social Security, defined benefit and defined contribution pensions, earnings, co-resident income and government programs such as SSI. We examined individuals aged 64-66 in 2005 and projections for individuals aged 64-66 in 2040. The typical couple in 2005 had total retirement income equal to 185% of final earnings, while the median projected couple in 2040 has a replacement rate of 131%. Now, these are projections, but bear in mind that the MINT model is probably the most comprehensive and best-vetted projection model of retirement income, with a large number of economists, demographers and social scientists working on it and regularly assessing its results.
As second paper used the Policy Simulation Group models, which look only at Social Security and private pensions. In this paper I used a different definition of replacement rates and also adjusted replacement rates to account for differences in household size and the number of children. (Larger households have economies of scale; children consume income during working years that doesn’t need to be replaced in retirement, so childless couples would need to save more for retirement than couples with children.) Here I looked at people born in 1940 and 1960 to track replacement rates from today through the 2020s. The median Social Security/pension replacement rate for the 1940 cohort was 92 percent while the median for the 1960 cohort was 82 percent. Working just an additional year would bring the 1960 birth cohort up to 1940 levels. While replacement rates do decline over time, due to a rising Social Security retirement age and declining DB pension coverage, most people at most income levels are doing ok.
This well-received paper by John Karl Scholz and Ananth Seshadri of the University of Wisconsin–Madison and Surachai Khitatrakun of the Urban Institute uses the HRS to examine how well the Baby Boomers are prepared for retirement. It uses a more sophisticated measure of retirement readiness than the replacement rates I used, but the basic idea is similar. They find that
over 80 percent of HRS households have accumulated more wealth than their optimal targets. These targets indicate the amounts of private saving households should have acquired at the time we observe them in the data, given their life cycle planning problem and Social Security and defined-benefit pension expectations and realizations. For those not meeting their targets, the magnitudes of the deficits are typically small.
In a draft follow-up paper Scholz and Seshadri reach similar conclusions for younger cohorts.
Now, there have been good research projects that have come to alternate conclusions, the most prominent being the National Retirement Risk Index generated by the Center for Retirement Research at Boston College. Without doing a very detailed nuts-and-bolts comparison of the CRR’s model to the others it’s hard to say exactly what’s driving the differences. While I really like the CRR’s work, I would say that models like SSA’s MINT are more comprehensive and have been around longer, so I would tend to favor those results. CRR’s model was built for the NRRI research project and I’m not sure its general results fully match those from other models.
None of this is to say retirement preparation doesn’t face problems. DC plans like 401(k)s present challenges to savers and both participation rates and asset management haven’t been as good as we’d like, although policies to address both problems are being implemented. Likewise, Social Security is almost sure to reduce benefits in the future, at least for middle and high earners, and Americans will need to save more to make up for these losses. But the key is to look closely at who’s unprepared for retirement, why, and what we should or can do about it. When we think about a “crisis” in retirement saving, hasty solutions might make things worse rather than better.
By Andrew Biggs