Citation
Thaler, R., & Benartzi, S. (2004). Save More Tomorrow™: Using behavioral economics to increase employee saving. Journal of Political Economy, 112(S1), S164-S187.
Highlights
- The study aimed to determine whether a retirement savings plan that automatically increased 401(k) contributions whenever an enrolled employee’s salary increased improved savings rates among employees.
- The authors compared savings rates among employees who opted to participate in the plan with savings rates among employees who declined to join the plan.
- Across the three experimental sites, average savings rates tended to increase, with higher increases for program participants than for nonparticipants.
- The quality of causal evidence presented in this study is low. This low rating means that we are not confident that the difference in savings rates between program participants and nonparticipants results from participation in the savings plan.
Intervention Examined
Save More Tomorrow™
Findings
- After implementing the SMarT program, average savings rates at the unnamed U.S. manufacturer increased from 4.4 to 10.6 percent of income.
- Among the employees at the anonymous company, those who joined the SMarT plan observed the greatest increase in savings rates, from 3.5 percent of income before the program to 13.6 percent of income by the fourth pay raise after joining the program.
- Average savings levels across all employees at Ispat Inland changed only slightly (from 5.5 to 5.8 percent of income) by the first pay raise after SMarT program implementation. However, savings rates among those who joined SMarT increased more over this period than did savings rates among those who did not enroll. Among employees who were already 401(k) plan participants, the saving rates among SMarT participants increased by 1.76 percentage points but decreased slightly for non-participants. Among employees who were not already 401(k) plan participants, increases were 2.28 and 0.26 percentage points, respectively.
- Savings rates among SMarT enrollees at Philips Electronics were 1.3 percentage points higher than among employees in comparison divisions after program implementation; at each of the two divisions for which SMarT was available, the savings rate increased more among employees who had not previously been saving through the 401(k) plan.
Considerations for Interpreting the Findings
In this study, employees at each of the three sites were offered the opportunity to enroll in the savings program. Thus, those who chose to enroll might differ systematically from those who chose not to enroll, whether in their predisposition or motivation to save for retirement or in other areas. Because the authors did not apply any statistical controls that would account for these differences or demonstrate that the groups were equivalent before enrolling or not enrolling in SMarT, the observed differences in savings rates might reflect factors other than savings program participation. It is also important to note that different sites applied different eligibility criteria that could also affect retirement savings outcomes. At Philips Electronics, for example, only less highly compensated employees were eligible to enroll in SMarT.
Causal Evidence Rating
The quality of causal evidence presented in this study is low. A low causal evidence rating means that we cannot be confident that the differences in 401(k) contribution rates between program participants and nonparticipants resulted from program participation alone. To provide more convincing evidence of the program’s effect on savings rates that satisfies CLEAR criteria, the authors could have used statistical models to control for pre-intervention differences in the participant and nonparticipant populations, thereby determining how much of the difference in savings reflected population characteristics and how much resulted from the savings program itself. Even such controls, however, do not account for intrinsic traits such as interest in saving that could account for some or all of the observed differences; only a randomized controlled trial can separate the effects of an intervention from the effects of both observable and unobservable characteristics.