Choi, J., Haisley, E., Kurkoski, J., & Massey, C. (2012). Small cues change savings choices. National Bureau of Economic Research working paper 17843. Cambridge, MA: NBER.
- The study’s objective was to gauge the effect of savings cues on 401(k) contribution rates by exposing randomly selected groups of employees to different hypothetical savings rates or amounts.
- The authors analyzed the relationship between exposure to savings rate cues and savings choices among employees at an unnamed U.S. technology company who contributed less than the maximum yearly amount to their retirement savings plans prior to the intervention.
- Overall, the authors found that participants adjusted their 401(k) contribution rates in response to cues. Those who were cued with the maximum contribution amount were 5.7 percent more likely than those who received no cue to contribute at maximal levels.
- The quality of causal evidence provided in this study is low because it is a randomized controlled trial with unknown attrition and unknown equivalence of the analytic groups. This means we are not confident the estimated effects are attributable to savings cues; other factors are likely to have contributed.
- Receiving the 2009 maximum contribution threshold treatment increased 401(k) contribution rates only among low savers. Differences also decreased throughout the year of the evaluation. Pooling across pay periods, participants in this treatment group were 5.7 percent more likely than their control counterparts to contribute at the maximum level in the first year post-intervention.
- In the 2010 experiment, high savers cued with any additional contribution treatment made significantly higher contributions than did those who received no savings cue. The version of the cue did not affect savings rates.
- In the 2010 experiment, the high savings threshold treatment led to significantly higher contributions among both low and middle savers.
Considerations for Interpreting the Findings
Study participants were all employees at the same technology company and thus cannot be considered a representative sample of the general population.
Causal Evidence Rating
The level of causal evidence provided in this study is low. This means we are not confident the estimated effects are attributable to the various savings cues; other factors are likely to have contributed. Although the authors conducted a randomized controlled trial, no information was provided on sample attrition, making it impossible to assess the validity of the experiment. In cases of high or unknown attrition, a study can receive a moderate causal evidence rating if the analysis controls for possible differences in background characteristics of the analytic treatment and control groups. However, the authors did not use such controls, and therefore differences in the outcomes of interest may reflect factors other than the experimental treatments.