Citation
Benartzi, S., Peleg, E., & Thaler, R. (2007). Choice architecture and retirement savings plans. Los Angeles, Ca. SSRN working paper.
Highlights
- The report’s objective was to examine the impact of choice architecture—that is, aspects of a savings plan’s structure and enrollment process—on 401(k) savings decisions.
- The authors discussed five original, observational studies. Two studies conducted regression analyses using survey data. One used an interrupted time series design and data from Vanguard on employees at 13 companies. Two additional studies conducted regression analyses using Vanguard data on 1.5 million participants in 1,830 of Vangaurd’s defined contribution retirement plans.
- The study receiving a moderate rating (study 5) found that members of retirement plans that offer retirement date funds, relative to members of retirement plans that offer risk-based funds or neither, held comparatively more equity when they were younger and comparatively less equity when they were older.
- The quality of causal evidence presented in this report is moderate because study 5 was a quasi-experimental design with adequate controls. We must note, however, that the remaining four studies provide causal evidence rated as low. The overall moderate evidence rating means we are somewhat confident that the estimated effects are attributable to the employees’ access to asset allocation funds, but other factors might also have contributed. In the case of the remaining four studies, however, we are not confident that the estimated effects are attributable to the interventions analyzed. Other factors are likely to have contributed.
Intervention Examined
The Choice Architecture of Retirement Savings Plans
Findings
- Study 1. Significantly more 401(k) plan members expressed interest in joining a program that automatically increased contribution rates after 12 months than in programs that automatically increased rates after 0 through 11 months.
- Study 2. Survey respondents expressed significantly greater interest in savings programs that linked contribution rate increases to pay rates than in programs that increased contribution rates after a fixed interval, with 38 percent describing themselves as very or extremely interested in the synchronized version of the program compared with only 32 percent for the nonsynchronized version.
- Study 3. Changing to automatic enrollment in a contribution escalator program led participation in the program to increase from 25.1 to 83.5 percent.
- Study 4. The availability of asset allocation funds mitigated the relationship between contribution amounts and the probability of equity holdings, although both monthly contributions and the availability of asset allocation funds remained statistically significant predictors of equity holdings.
- Study 5. Members of retirement plans that offered retirement date funds, relative to members of retirement plans that offered risk-based funds or neither, held comparatively more equity when they were younger and comparatively less equity when they were older.
Considerations for Interpreting the Findings
The first four studies in this report did not demonstrate adequate statistical controls for factors other than the intervention that might have contributed to the outcome of interest. For example, in studies 1 and 2, which described survey responses, the authors did not describe how respondents were assigned to survey versions or whether respondents’ background characteristics were comparable across versions. Assignment to treatment conditions in these studies might have been random, which would lead us to expect comparable distributions of background characteristics in all conditions, but the Clearinghouse for Labor Evaluation and Research (CLEAR) was unable to confirm the assignment mechanism. If assignment to treatment was not random, variation in these background characteristics could also explain variation in responses, weakening the causal relationship the authors sought to establish.
The fifth study employed a quasi-experimental design that comprehensively controlled for observable characteristics that could influence the outcome of interest. Controls included plan-level measures of the proportion of participants’ demographics; average contributions, account balances, and membership durations; proportion of web-registered users; log number of participants; and indicators for whether the plan offered loans, company stock, or a brokerage account. The authors also provided a detailed discussion explaining that the availability of certain kinds of funds was not related to unobservable employee characteristics. However, quasi-experimental designs cannot control for these unobservable characteristics, which can also influence the outcome of interest.
Causal Evidence Rating
The quality of causal evidence presented in this report is moderate because the fifth study is a quasi-experimental design with adequate controls. However, the remaining four studies provide causal evidence rated as low. The overall moderate evidence rating means we are somewhat confident that the estimated effects are attributable to the employees’ access to asset allocation funds, but other factors might also have contributed. In the case of the remaining four studies, we are not confident that the estimated effects are attributable to the pertinent interventions. Other factors are likely to have contributed.