The Effect of Providing Peer Information on Retirement Savings Decisions

• Author: JAMES J. CHOI,BRIGITTE C. MADRIAN,KATHERINE L. MILKMAN,JOHN BESHEARS,DAVID LAIBSON
• Date: 01 June 2015
• Published date: 01 June 2015
• DOI: http://doi.org/10.1111/jofi.12258

THE JOURNAL OF FINANCE •VOL. LXX, NO. 3 •JUNE 2015

The Effect of Providing Peer Information

on Retirement Savings Decisions

JOHN BESHEARS, JAMES J. CHOI, DAVID LAIBSON, BRIGITTE C. MADRIAN,

and KATHERINE L. MILKMAN∗

ABSTRACT

Using a field experiment in a 401(k) plan, we measure the effect of disseminating in-

formation about peer behavior on savings. Low-saving employees received simplified

plan enrollment or contribution increase forms. A randomized subset of forms stated

the fraction of age-matched coworkers participating in the plan or age-matched par-

ticipants contributing at least 6% of pay to the plan. We document an oppositional

reaction: the presence of peer information decreased the savings of nonparticipants

who were ineligible for 401(k) automatic enrollment, and higher observed peer sav-

ings rates also decreased savings. Discouragement from upward social comparisons

seems to drive this reaction.

IN1980, 30 MILLION U.S . WORKERS actively participated in employer-sponsored

defined benefit (DB) retirement savings plans, and 19 million actively partic-

ipated in employer-sponsored defined contribution (DC) retirement savings

plans. By 2011, participation in DB plans had nearly halved to 17 million

workers, whereas DC plan participation had skyrocketed to 74 million

∗John Beshears, David Laibson, and Brigitte C. Madrian are with Harvard University and

NBER, James J. Choi is with Yale University and NBER, and Katherine L. Milkman is with Uni-

versity of Pennsylvania. We thank Aon Hewitt and our corporate partner for conducting the field

experiment and providing the data. We are particularly grateful to Mary Ann Armatys, Diane

Dove, Pam Hess, Barb Hogg, Diana Jacobson, Larry King, Bill Lawless, Shane Nickerson, and Yan

Xu, some of our many contacts at Aon Hewitt. We thank Campbell Harvey (the Editor), an Asso-

ciate Editor, an anonymous referee, Hunt Allcott, Sherry Li, and seminar participants at Brigham

Young University, Case Western Reserve University, Cornell University, New York University,

Norwegian School of Economics, Stanford University, University of California Berkeley, Univer-

sity of Maryland, University of Pennsylvania, the NBER Summer Institute, the Harvard Business

School / Federal Reserve Bank of Boston Consumer Finance Workshop,and the Behavioral Decision

Research in Management Conference for their insightful feedback. Michael Buckley,Yeguang Chi,

Christina Jenq, John Klopfer, Henning Krohnstad, Michael Puempel, Alexandra Steiny, and Eric

Zwick provided excellent research assistance. Beshears acknowledges financial support from a Na-

tional Science Foundation Graduate Research Fellowship. Beshears, Choi, Laibson, and Madrian

acknowledge individual and collective financial support from the National Institutes of Health

(grants P01-AG-005842, R01-AG-021650, and T32-AG-000186). This research was also supported

by the U.S. Social Security Administration through grant #19-F-10002-9-01 to RAND as part of

the SSA Financial Literacy Research Consortium. The findings and conclusions expressed are

solely those of the authors and do not represent the views of the SSA, any agency of the Federal

Government, or RAND. See the authors’ websites for lists of their outside activities.

DOI: 10.1111/jofi.12258

1161

1162 The Journal of Finance R

workers.1The shift from DB plans, which set contribution levels and invest-

ment allocations on behalf of employees, to DC plans, which allow employees to

choose from a complex array of possible contribution levels and investment al-

locations, has occurred amidst concerns that workers are not equipped to make

well-informed savings choices (Mitchell and Lusardi (2011)). Employers have

thus become increasingly interested in programs designed to help employees

make good choices in DC plans. This paper studies one such program.

Specifically, we use a field experiment to investigate the effect of a peer

information intervention on retirement savings choices. Peer information in-

terventions involve disseminating information about what a target population’s

peers typically do. By sharing this information, it may be possible to teach peo-

ple that a certain behavior is more common than they had previously believed

and thereby motivate those people to engage in the behavior more themselves.

This approach has been dubbed “social norms marketing” and is used at approx-

imately half of U.S. colleges in an effort to reduce student alcohol consumption

(Wechsler et al. (2003)).

Peer information interventions may move behavior towards the peer-group

average for several reasons. First, an individual may mimic peers because

their behavior reflects private information relevant to the individual’s payoffs

(Banerjee (1992), Bikhchandani, Hirshleifer, and Welch (1992), Ellison and

Fudenberg (1993)). Another possibility is that the intervention provides in-

formation about social norms from which deviations are costly due to a taste

for conformity, the risk of social sanctions, identity considerations, or strate-

gic complementarities (Asch (1951), Festinger (1954), Akerlof (1980), Bernheim

(1994), Akerlof and Kranton (2000), Glaeser and Scheinkman (2003), Benjamin,

Choi, and Strickland (2010), Benjamin, Choi, and Fisher (2010)). Finally, indi-

viduals may directly derive utility from relative consumption (Abel (1990)).

A growing empirical literature documents that peer effects indeed play a role

in financial decisions when peers interact with each other organically. Peers

affect retirement saving outcomes (Duflo and Saez (2002,2003)), stock market

participation (Hong, Kubik, and Stein (2004), Brown et al. (2008)), corporate

compensation and merger practices (Bizjak, Lemmon, and Whitby (2009), Shue

(2013)), entrepreneurial risk-taking (Lerner and Malmendier (2013)), and gen-

eral economic attitudes such as risk aversion (Ahern, Duchin, and Shumway

(2014)).2Peer information interventions such as the one we study are designed

to harness the power of these peer effects to influence behavior.

1Source: U.S. Department of Labor Employee Benefits Security Administration, Private Pension

Plan Bulletin Historical Tables and Graphs, Table E8, June 2013.

2Hirshleifer and Teoh (2003) review the literature on herding and related phenomena in fi-

nancial markets. For evidence of peer effects in other domains, see Cialdini, Reno, and Kall-

gren (1990), Case and Katz (1991), Besley and Case (1994),Hersheyetal.(1994), Foster and

Rosenzweig (1995), Glaeser, Sacerdote, and Scheinkman (1996), Bertrand, Luttmer, and Mul-

lainathan (2000), Kallgren, Reno, and Cialdini (2000), Sacerdote (2001), Munshi (2004), Munshi

and Myaux (2006), Sorensen (2006), Gerber, Green, and Larimer (2008), Grinblatt, Keloharju,

and Ik¨

aheimo (2008), Kuhn et al. (2011), Narayanan and Nair (2013), and Chalmers, Johnson,

and Reuter (2014). Manski (2000) provides an overview of related work in the social interaction

literature.

Peer Information and Retirement Savings Decisions 1163

Many studies find that peer information interventions cause behavior to more

closely conform to the disseminated peer norm.3Our field experiment, however,

yields the surprising result that peer information interventions can generate

an oppositional reaction: information about the high savings rates of peers can

lead low-saving individuals to shift away from the peer norm and decrease their

savings relative to a control group that did not receive peer information. Our

evidence suggests that this effect is driven in part by peer information causing

some individuals to become discouraged, making them less likely to increase

their savings rates.

We conducted our experiment in partnership with a large manufacturing

firm and its retirement savings plan administrator. Employees received differ-

ent letters depending on their 401(k) enrollment status. Employees who had

never participated in the firm’s 401(k) plan were mailed Quick Enrollment (QE)

letters, which allowed them to start contributing 6% of their pay to the plan

with a preselected asset allocation by returning a simple reply form. Employees

who had previously enrolled but were contributing less than 6% of their pay

received Easy Escalation (EE) letters, which included a nearly identical reply

form that could be returned to increase their contribution rate to 6%. Previ-

ous work shows that these simplified enrollment and contribution escalation

mechanisms significantly increase savings plan contributions (Choi, Laibson,

and Madrian (2009), Beshears et al. (2013a)).

We assigned the QE and EE mailing recipients to one of three randomly

selected treatments. The mailing for the first randomly selected treatment in-

cluded information about the savings behavior of coworkers in the recipient’s

five-year age bracket (e.g., employees at the firm between the ages of 20 and

24). The mailing for the second randomly selected treatment contained sim-

ilar information about coworkers in the recipient’s 10-year age bracket (e.g.,

employees at the firm between the ages of 20 and 29). The mailing for the

third randomly selected treatment contained no peer information and therefore

served as a control condition. For the QE recipients, the two peer information

mailings stated the fraction of employees in the relevant age bracket who were

already enrolled in the savings plan. For the EE recipients, the two peer infor-

mation mailings stated the fraction of savings plan participants in the relevant

age bracket contributing at least 6% of their pay on a before-tax basis to the

plan.

Employees in our study naturally fall into four subpopulations distinguished

along two dimensions: QE recipients versus EE recipients, and employees who

were automatically enrolled at a 6% contribution rate unless they opted out

3For example, providing information about peers moves behavior towards the peer norm in

domains such as entr´

ee selections in a restaurant, contributions of movie ratings to an online

community, small charitable donations, music downloads, towel re-use in hotels, taking petrified

wood from a national park, and stated intentions to vote (Cai, Chen, and Fang (2009), Chen

et al. (2010), Frey and Meier (2004), Salganik, Dodds, and Watts (2006), Goldstein, Cialdini, and

Griskevicius (2008), Cialdini et al. (2006), Gerber and Rogers (2009)). However, Beshears et al.

(2013b) find that disseminating short printed testimonials from peers is not effective at increasing

conversion from brand-name prescription drugs to lower-cost therapeutic equivalents.