Jonathan Reuter is an Associate Professor of Finance at Boston College’s Carroll School of Management. He is also a Research Associate at the National Bureau of Economic Research and a TIAA Institute Fellow.
In my recent study prepared for the Pension Research Council of the Wharton School, I surveyed what is by now a large literature on defined contribution (DC) retirement plan design and participant behavior. My specific question was to evaluate whether behavioral nudges such as automatic enrollment actually increase plan participation rates and retirement account balances.
In the Beginning
In the early 2000s and before, the DC retirement plans that US firms offered their employees relied on voluntary enrollment (VE). In other words, employees who chose to participate in the employer-offered plan would choose their own contribution rates and decide how to allocate these contributions across the available investment options. Workers also bore the burden of having to decide when and whether to rebalance their holdings in response to market returns or changes in financial circumstances. Employees who did not voluntarily enroll contributed nothing into the plan, forgoing any employer match.
Despite the likelihood that financially literate employees were more likely to voluntarily enroll, early studies nevertheless found substantial evidence of sub-optimal participant behavior. For example, some analysts noted that workers’ asset allocation decisions often responded to the composition of investment menus but not to how employer matching contributions were invested. Other research estimated that inefficient portfolio construction reduced participants’ expected retirement wealth by as much as one-fifth, over a 35-year investment horizon.
Today and tomorrow
Following the Pension Protection Act of 2006 and subsequent legislation, many firms switched from voluntary enrollment to automatic enrollment (AE), whereby they enrolled new hires in the retirement plan unless the workers actively chose to opt out. The earliest studies of AE found large increases in participation rates under AE, particularly for younger workers and those with lower incomes, as well as those with lower levels of financial literacy. They also found a strong tendency, however, for participants to anchor on the employer’s default savings rate and the default investment option chosen by the employer, despite the fact that the default investment option was often a (low risk) money market fund. In other words, AE simultaneously increased the fraction of workers saving for retirement, and reduced active choice with respect to how much to save or how to best invest, highlighting the challenge of implementing and benchmarking behavioral nudges.
More recently, researchers have found that the long-run effects of AE on retirement account balances are smaller than the short-run effects. One reason is that those being nudged to save are more likely to take withdrawals when they change jobs, resulting in leakage from the retirement system. Another reason is that those not subject to AE actively choose to save more later, on average, which narrows the gap. Furthermore, when researchers have been able to study household assets and liabilities, they have found that some incremental retirement savings have been financed by increased household debt, especially among lower-income workers. These findings raise concerns about how plan sponsors should think about optimal defaults given their heterogeneous workforces.
The challenges associated with benchmarking nudges are perhaps best illustrated when studying state level attempts to extend automatic enrollment to firms that do not offer their own retirement plans, beginning with OregonSaves in 2018. Participant rates in OregonSaves are much lower than in other settings with AE but baseline savings rates are essentially zero, and the fact that many employees stop contributing following job loss speaks more to job market challenges than to plan design. These challenges likely contributed to the absence of employer-sponsored retirement plans in the first place. Whether participants use state-sponsored retirement plan assets to smooth consumption following periods of reduced income is an interesting open question.
The rise of target date funds as the dominant default investment option is likely to have enhanced many participants’ asset allocations, particularly those with lower levels of financial literacy. A downside, however, is that improving the default investment option appears to reduce active choice with respect to savings rates, and with advice-seeking more generally. One promising response to the evidence of treatment effects that vary across dimensions and participants would be to condition default options on employee characteristics. More immediately, managed accounts have the potential both to increase the level of portfolio customization and to trigger conversations about optimal savings rates.
Future regulation
Recent US legislation has increased the fraction of US employers offering retirement plans, the fraction of employees participating in these plans, and contribution rates. For example, all new retirement plans are required to feature both AE and automatic escalation, and the maximum rate allowed under automatic escalation has increased from 10 to 15 percent. (Automatic escalation typically increases the employee contribution rate by 1 percentage point per year, until the maximum-allowable contribution rate is reached or the employee opts out of automatic escalation.)
One of the more interesting changes allows employers to treat student loan payments as employee retirement contributions for the purposes of matching. While the intent is clearly to increase retirement savings by extending the match to employees unable to both make student loan payments and retirement contributions, a recent study finds that allowing employees to collect the employer match without making employee contributions is likely to reduce retirement contributions more generally.
Another change that will be of particular interest to economists is the ability to offer small financial incentives to participate in retirement plans. It is an open question whether such incentives can meaningfully increase participation in plans that feature voluntary enrollment.
Finally, there are also efforts to increase emergency savings through the use of defaults, something that has been found effective in the United Kingdom. For lower-income workers in particular, it may be useful to establish emergency savings accounts before saving for retirement.
Conclusions
DC retirement plans have come a long way since the US Employee Retirement Income Security Act of 1974, and they continue to evolve. Widespread acceptance of automatic enrollment, automatic escalation, and sensible default investment options allow workers to outsource retirement plan participation decisions, savings rates, and asset allocation decisions to their employers, simultaneously increasing savings and (likely) decreasing the extent to which participants make common investment mistakes. The potential downside is a reliance on one-size-fits-all solutions to complex optimization problems that depend on a wide range of household characteristics and preferences. The optimal plan design is one that can encourage active choice along those dimensions about which employees are best able to make decisions, but which otherwise relies on defaults.
Views of our Guest Bloggers are theirs alone, and not of the Pension Research Council, the Wharton School, or the University of Pennsylvania.
