Serenity Now, Save Later? Evidence on Retirement Savings Puzzles from a 401(k) Field Experiment

How to cite

Bhargava, Saurabh; Conell-Price, Lynn. 2022. "Serenity Now, Save Later? Evidence on Retirement Savings Puzzles from a 401(k) Field Experiment." AEA RCT Registry. AEARCTR-0010007. https://doi.org/10.1257/rct.10007-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Saurabh Bhargava* , Lynn Conell-Price

* Primary investigator

Registry ID
AEARCTR-0010007
DOI
https://doi.org/10.1257/rct.10007-1.0
First registered
Last updated
Status
Completed
Trial dates
2016-07-19 – 2016-07-30
Countries
United States of America ()
Registry record
http://www.socialscienceregistry.org/trials/10007

Abstract

Economists have advanced several psychological frictions to explain why many 401(k)-eligible employees undersave for retirement despite generous matching incentives. We provide evidence on four of these frictions through a high-compliance field experiment randomizing undersaving employees to information- and incentive-based treatments linked to a survey assessing each friction’s baseline incidence. We report four findings: (1) We corroborate evidence of pervasive deficits in retirement literacy and their correlation with saving but precisely reject any meaningful increase in saving from personalized recommendations that demonstrably improve literacy. (2) In (unplanned) analyses of plan confusion, we estimate that 20 to 37 percent of non-participants mistakenly believed themselves to be enrolled—these employees enrolled at high rates when prompted to review their plan status. (3) We find no evidence that enrollment complexity impedes saving—few employees perceived enrollment as prohibitively time-consuming and simplifying enrollment further did not increase saving. (4) We directly implicate present focus as a cause of undersaving by showing a significant share of employees increased saving in response to a small but immediate $10 gift card but not to clarification of the dramatically larger, but delayed, plan match. A survey of policy/industry stakeholders suggests these findings challenge perceived plan engagement best-practices. Finally, calibrations indicate a beta-delta model of present bias cannot account for the observed behavior and beliefs of employees. We propose an alternative model of anxiety-based present focus and deferred optimism that does explain our findings—and possibly other retirement savings puzzles—and offers a psychological rationale for the use of microincentives to increase engagement and for more structural reforms seeking to link traditional saving accounts to more liquid accounts designed to relieve near-term anxiety.

Intervention

We administered the field study by inviting qualified employees to participate in an online survey marketed as an opportunity to provide confidential workplace feedback. Beyond capturing demographic and financial background, the first instrument module included questions diagnosing each candidate friction. Employees were then randomized to one of several experimental variants of a second module promising a personalized assessment of retirement preparedness. Across variants, the assessment truthfully conveyed that the employee was not “on track” for retirement security with an accompanying graphic, advised the employee to increase their contribution rate, provided simple instructions to do so, and, finally, asked the employee about their future saving expectations. To test each friction, the treatments varied the presence of (1) a personalized saving recommendation, (2) information clarifying the magnitude of the plan match, and (3) a small, but immediate, reward ($10 Amazon gift card) to encourage employees to visit the enrollment portal and engage the decision to save.

Experimental design

Research Design:
To investigate the causal relationship between the four candidate frictions and 401(k) plan engagement, we administered a field experiment, embedded in an online survey, to undersaving employees at a large US financial services firm. We marketed the online instrument as an employer-sponsored opportunity for employees to provide confidential feedback regarding the workplace and benefit programs. The first module of the instrument featured a survey that, beyond capturing relevant background information, assessed the employee-specific incidence of each candidate friction. The second module promised employees to assess their retirement preparedness and, if necessary, provide guidance to improve their preparedness. This module implemented the field experiment by randomizing employees to one of several information- and/or incentive-based treatments. While the survey provided evidence as to the baseline prevalence of each friction and the field experiment provided evidence as to the average effect of reducing/engaging each friction on saving, jointly, the survey and field experiment clarify the heterogeneous importance of frictions across baseline incidence.

Sample Construction:
On July 19, 2016, we invited a pre-specified sample of undersaving employees of low-to-moderate income by email to participate in a ten- to fifteen-minute survey marketed as an opportunity to provide confidential feedback on the workplace and employee benefit programs. The email explained that while the survey was part of a broader partnership with the firm to help improve employee well-being, it was independently designed and administered by academic researchers from Carnegie Mellon University. Employees were directed to participate in the survey, hosted on the Qualtrics platform, by clicking a personalized link within the ten-day survey period. To encourage a high response rate, survey respondents were entered into a raffle for an Apple iPad and were reminded to complete the survey via email.
The construction of the invitation sample was shaped by two considerations—a desire to target undersaving employees of low-to-moderate income and the firm’s request to limit invitations to 5,000 employees. To differentiate between employees who did and did not fully claim the plan match, we ultimately invited two non-overlapping samples of undersaving employees to the survey. The primary sample (the “Low Arm”) comprised the universe of 3,719 401(k) plan-eligible employees who, as of late June 2016, were 25 to 55 years of age, earned less than $100k annually and contributed less than 4 percent to their 401(k) plan (inclusive of non-participants). A second sample of 1,000 (the “Moderate Arm”) comprised a random draw of plan-eligible employees who, as of late June 2016, were 25 to 55 years of age, earned less than $100k annually and contributed 4 to 9 percent to their 401(k) plan. Twenty-eight percent of invited employees participated in the study. We attribute the relatively high response rate for an email solicitation to the lottery-based incentive, email reminder, and sponsorship/promotion by the employer. After excluding those who exited the survey prematurely or were already contributing at or above their recommended rate, we randomly assigned 1,137 employees to an experimental treatment within one of two study arms (780 in the Low Arm; 357 in the Moderate Arm).

Design details

Randomization method
Randomization by computer (Qualtrics)
Randomization unit
Individual
Clusters
1137
Observations
1137
Arms
Low Arm: 780 across 3 treatments Moderate Arm: 357 across 2 treatments
Primary outcomes
Increased 401(k) contribution rate as inferred from administrative data from the pay-cycles immediately before and after the study

Cited by

No citation data available for this registration yet.

Keywords