Skip to main content

Saving for retirement on the path of least resistance (Choi et al. 2004)

Citation

Choi, J., Laibson, D., Madrian, B., and Metrick, A. (2004). Saving for retirement on the path of least resistance. Working paper. Cambridge, MA.

Highlights

  • The study examined the impact of different 401(k) plan designs on employees’ saving behavior at several U.S.-based firms.
  • The analysis was based on administrative data provided by each company. The study used five analyses—one based on an interrupted time series (ITS) design, one based on a randomized controlled trial (RCT), and three quasi-experimental regression analyses—to estimate impacts on 401(k) participation and contribution rates.
  • The study found that employees’ decision making was influenced by defaults, such as automatic enrollment and automatic disbursements from 401(k) plans. It also found that the gap in 401(k) participation rates between new employees facing a mandatory waiting period for eligibility and those with immediate eligibility disappeared shortly after the former group gained eligibility.
  • The quality of causal evidence presented in this report is low. This means we are not confident that the estimated effects are attributable to the various 401(k) program designs evaluated in this study. Other factors are likely to have contributed.

Intervention Examined

401(k) Plan Design Changes

Study Sites

  • Company A: a large U.S. food company employing about 10,000 people
  • Company B: an office equipment firm employing about 30,000 people
  • Company C: an insurance firm employing about 30,000 people
  • Company D: a food industry firm employing about 20,000 people
  • Company F: a consumer packaged goods firm employing about 40,000 people
  • Company G: an insurance firm employing about 50,000 people
  • Company H: a manufacturing firm of unknown size
  • Company I: a retail firm employing about 130,000 people
  • Company J: a financial services firm employing about 50,000 people
  • Company K: a pharmaceutical firm employing about 10,000 people

Study 1 used data from companies B, C, D, and H; study 2 from companies B, D, I, and J; study 3 from company A; study 4 from companies F, G, and K; and study 5 from company C.

Findings

  • Study 1. Automatic enrollment was associated with increases in 401(k) participation of 50 to 67 percentage points. Employees were more likely to enroll at the default rate and contribute to the specified automatic enrollment fund when automatic enrollment was in place.
  • Study 2. Sixty-five to 90 percent of terminated employees with 401(k) balances of less than $1,000 took a cash disbursement compared with only 20 to 33 percent of employees with balances above $5,000.
  • Study 4. The gap in participation rates between new employees with a mandatory waiting period for 401(k) eligibility and those with immediate eligibility was large at the beginning of employment but became statistically insignificant after 9 to 22 months of tenure, depending on the company.

Considerations for Interpreting the Findings

  • Study 1 used an ITS design and considered four different implementations of automatic enrollment; however, only one pre- and one post-intervention observation was analyzed for three of the four implementations. Thus, it is unclear whether changes in employees’ behavior before the intervention precipitated the changes in policy.
  • Studies 2, 3, and 5 used quasi-experimental designs and analyzed impacts using regression analysis or comparisons of means. In all cases, the authors did not adequately control for differences between the intervention and comparison groups in key characteristics, such as income and gender.
  • Study 4 used an RCT with high attrition and did not control for differences between the treatment and comparison groups or demonstrate that these differences were small.

Causal Evidence Rating

The quality of causal evidence presented in this report is low; none of the five studies conducted by the authors meet the Clearinghouse for Labor Evaluation and Research criteria for the given design. This low causal evidence rating means we are not confident that the estimated effects are attributable to the various 401(k) program designs evaluated in this study. Other factors are likely to have contributed.

Additional Sources

Published as: Choi, James, Laibson, David, Madrian, Brigette, and Metrick, Andrew. (2006). Saving for retirement on the path of least resistance. In Edward McCaffrey and Joel Slemrod (Eds.), Behavioral Public Finance. New York: Russell Sage.

Reviewed by CLEAR

April 2015