The Effect of Providing g Peer Information on Retirement S i Savings Decisions D i i JJohn h Beshears, B h James J J.J Choi, Ch i David Laibson, Brigitte C. Madrian, Katherine L. Milkman Why might people imitate peers? Peers know something that I don’tt Peers know something that I don (Banjeree, 1992; Bikhchandi, Hirshleifer, and Welch, 1992; Ellison and Fudenberg, 1993) Peer behavior reveals social norms (Akerlof 1980 Bernheim 1994 Akerlof and Kranton, 2000) (Akerlof, 1980; Bernheim, 1994; Akerlof and Kranton 2000) Empirical p evidence of p peer effects Retirement savings Stock market participation Stock market participation Technology adoption Criminal activity Education Welfare usage Car purchases Car purchases Contraception Littering Social norms marketing T ll Tell people what their peers do l h h i d Field evidence of social norms marketing efficacy Restaurant entrée selection Music downloads Small charitable donations Towel re‐use in hotels Theft of petrified wood Stated intentions to vote Household energy consumption 48% of U.S. 4‐year colleges and colleges and universities have conducted a conducted a social norms marketing k ti campaign Today: Field experiment at a large firm to test the effect of social marketing on savings the effect of social marketing on savings Target employees: 401(k) non participants Target employees: 401(k) non‐participants Treatment: Randomized receipt of information Randomized receipt of information about peer 401(k) participation rates Measured outcome: Recipients’ subsequent 401(k) enrollments ( ) Social market to increase retirement saving? Fortune 500 manufacturing firm 20% of workers unionized 401(k) plan Immediate eligibility • Immediate eligibility • 21 mutual funds (11 target date) Match thresholds Match structure 1 Match threshold Match threshold 6% M t h t t Match structure 2 2 2% 2% up to $325 t $325 Match structure 3 8% Match structure 4 6% Match structure 5 Match structure 5 4% Match structure 6 6% Match structure 7 0% Automatic 401(k) ( ) enrollment Non‐union No Non‐participants automatically enrolled on a ti i a t auto ati ally e olled o January 1, 2008 New hires after January 1, 2008 automatically enrolled after 60 days Union Opt‐in enrollment Sample p selection Non‐union Never enrolled Enrolled, contributing < 6% Quick E Enrollment ll t Easy E l ti Escalation Union Enrolled, contributing ≥ 6% Never enrolled Enrolled, contributing < 6% and < threshold Quick E Enrollment ll t Easy E l ti Escalation Enrolled, contributing ≥ 6% or ≥ threshold Passivity y taxonomy y of sample p Quick Enrollment Non‐union Easy Escalation Union Non‐union Union Never enrolled Never enrolled Enrolled, contributing < 6% Enrolled, contributing < 6% and < 6% and < threshold Active d i i decision Passive Active d i i decision Active d i i decision Effect of Q QE / EE at this firm Enrollment / Cont. escalation Before Intervention During Intervention Difference QE, union 2.0% (0.4) 7.6% (0.8) 5.6%*** (0.9) QE, non-union 2.9% (0.8) 2.3% (0.7) -0.7% (1.1) EE union EE, 88.0% 0% (0.9) 10.6% 10 6% (1.1) 22.6% 6%* (1.5) EE, non-union 9.4% (0 5) (0.5) 8.2% (0 5) (0.5) -1.1% (0 8) (0.8) Quick Q Enrollment letter without peer information Quick Q Enrollment letter with peer information Join the 87% of 25‐29 y year old employees at p y <company> who are already in the <plan>. Easy y Escalation letter without peer information Easy y Escalation letter with peer information Join the 76% of 20‐29 y year old <plan> p participants at <company> who are already contributing already contributing at least 6% to the <plan>. Three randomly assigned letters (sent to non‐ participants): • No peer information • Information about other employees in the same 5‐year age bucket as the letter recipient • Information about other employees in the same 5‐year age bucket as the letter recipient Random assignment of an illustrative 24 year old employee: 24-year-old A No peer info A. No peer info B. Peer info from employees aged 20‐29 p y g (participation rate of 82%) C P C. Peer info from employees aged 20‐24 i f f l d 20 24 (participation rate of 77%) D. Also compare to 25‐year‐old employee receiving peer info from employees 25‐29 receiving peer info from employees 25‐29 (participation rate of 87%) Peer information – Participation rates 100 90 80 70 60 50 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 Peer information – Fraction of participants contributing >=6% 100 90 80 70 60 50 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 Peer information presence effect, Union employees Enrolled in next 2 months 9.9% 6.5% No peer info Peer info Peer info effect: ‐3.4%* (1.9) Peer information presence effect, Union non-participants Contribution rate change Peer info effect: 0.58% 0.39% No peer info Peer info ‐0.19%* (0.10) Peer information presence effect, Non-union employees Enrolled in next 2 months Peer info effect: 2.3% (1.3) 3.0% 0.7% No peer info Peer info Peer information presence effect, Non-union non-participants Contribution rate change Peer info effect: 0.13% (0.09) 0.17% 0.04% No peer info Peer info Peer information presence effect, Union participants Increased contrib. in next 2 months 10.6% 10.6% Peer info effect: ‐0.1% (2.4) No peer info Peer info Peer information presence effect, Non-union participants Increased contrib. in next 2 months 8.2% 8.3% Peer info effect: 0.1% (1.1) No peer info Peer info Table 6: Peer info presence effect Enrollment / Contribution escalation Peer info received Male dummy log(Tenure) log(Salary) Cont. rate dummies log(Balance) Sample size Nonparticipants participants, union -0.04** (0.02) -0.02 (0.02) -0.03*** ((0.01)) 0.04** (0.02) No Nonparticipants participants, non-union 0.02 (0.01) -0.03 (0.02) -0.01 ((0.01)) 0.05** (0.03) No N = 1,024 N = 399 Participants with low cont rates, cont. rates union -0.01 (0.03) -0.05** (0.03) -0.02 ((0.01)) 0.04 (0.04) Yes 0.02** (0.01) N = 746 Participants with low cont rates, cont. rates non-union 0.00 (0.01) 0.00 (0.01) -0.02*** ((0.01)) 0.04*** (0.01) Yes 0.01*** (0.00) N = 2,753 Table 6: Peer info presence effect Contribution rate change Peer info received M l dummy Male d log(Tenure) log(Salary) Cont. rate dummies log(Balance) Sample size Nonparticipants participants, union -0.21* (0.11) -0.08 0 08 (0.11) -0.19*** ((0.05)) 0.23** (0.11) No Nonparticipants participants, non-union 0.13 (0.08) -0.14 0 14 (0.11) -0.08* ((0.04)) 0.33** (0.15) No N = 1,024 N = 399 Participants with low cont rates, cont. rates union -0.02 (0.10) -0.15 0 15 (0.10) -0.10 ((0.06)) 0.16 (0.14) Yes 0.07** (0.03) N = 746 Participants with low cont rates, cont. rates non-union 0.07 (0.06) 0 02 0.02 (0.05) -0.09*** ((0.03)) 0.30*** (0.10) Yes 0.08*** (0.01) N = 2,753 Estimating peer information number magnitude effect Problem Peer number is correlated with age Peer number is correlated with age Age is correlated with propensity to respond Solution Include linear age spline with knot points every 5 years starting at age 22.5 every 5 years starting at age 22.5 Table 7: Peer info number effect Enrollment / Contribution escalation Peer info number M l dummy Male d log(Tenure) log(Salary) Cont. rate dummies log(Balance) Sample size Nonparticipants participants, union -1.74** (0.73) 0 00 0.00 (0.02) -0.02* ((0.01)) 0.02 (0.02) No Nonparticipants participants, non-union 1.15** (0.57) -0.06* 0 06* (0.03) -0.02** ((0.01)) 0.09** (0.04) No N = 687 N = 264 Participants with low cont rates, cont. rates union 2.28 (1.90) -0.03 0 03 (0.03) -0.01 ((0.02)) 0.04 (0.06) Yes 0.01 (0.01) N = 511 Participants with low cont rates, cont. rates non-union 0.60 (0.81) -0.00 0 00 (0.01) -0.02*** ((0.01)) 0.04*** (0.02) Yes 0.01*** (0.00) N = 1,822 Effects of variation in peer behavior For union workers, a ten percentage point p g p increase in the participation rate of the comparison group reduces enrollment likelihood by 17.4% (**). For non‐union workers, a ten percentage point increase in the participation rate of the comparison group increases enrollment likelihood by 11.5% (**). Table 7: Peer info number effect Contribution rate change Peer info number M l dummy Male d log(Tenure) log(Salary) Cont. rate dummies l (B l log(Balance) ) Sample size Nonparticipants participants, union -10.54** (4.59) 0 04 0.04 (0.13) -0.10* (0.06) 0.11 (0.13) No Nonparticipants participants, non-union 5.92** (3.01) -0.28* 0 28* (0.17) -0.13** (0.05) 0.54** (0.23) No N = 687 N = 264 Participants with low cont rates, cont. rates union 10.94 (7.09) -0.04 0 04 (0.12) -0.11 (0.08) 0.18 (0.21) Yes 0 07* 0.07* (0.04) N = 511 Participants with low cont rates, cont. rates non-union 8.23* (4.24) 0 00 0.00 (0.06) -0.12*** (0.05) 0.36** (0.15) Yes 0 09*** 0.09*** (0.02) N = 1,822 Table 8: Sources of identification Enrollment Peer info number Nonparticipants, i union -1.93** (0.81) Nonparticipants, non-union i 1.50* (0.87) Knocking out age bucket discontinuities Contribution rate change Peer info number Nonparticipants, union -11.56** (5.05) Nonparticipants, non-union 9.08* (5.17) Nonparticipants, union i -0.77 (1.22) Nonparticipants, non-union i 1.07 (1.05) Knocking out 5‐year vs. 10‐year peer number y p difference Nonparticipants, union -5.44 (7.57) Nonparticipants, non-union 3.63 (4.02) Table 9: Robustness Enrollment Peer info number Nonparticipants, i union -1.72** (0.73) Nonparticipants, non-union i 1.44** (0.68) Age spline with knot points every 2.5 years p y y Contribution rate change Peer info number Nonparticipants, union -10.43** (4.61) Nonparticipants, non-union 7.32** (3.36) Nonparticipants, union i -1.91*** (0.72) Nonparticipants, non-union i 1.06* (0.58) Controlling for target retirement date fund offered Nonparticipants, union -11.55** (4.53) Nonparticipants, non-union 6.21* (3.45) Table 9: Robustness Enrollment Peer info number Nonparticipants, i union -2.08*** (0.80) Nonparticipants, non-union i 1.05* (0.62) Controls for years from age group mean g g p (linear, squared) Contribution rate change Peer info number Nonparticipants, union -12.46** (4.98) Nonparticipants, non-union 6.06 (3.69) Nonparticipants, union i -1.76** (0.75) Nonparticipants, non-union i 1.29* (0.68) Controls for percentile within age group g g p (linear, squared) Nonparticipants, union -10.43** (4.65) Nonparticipants, non-union 7.16* (3.91) Possible mechanisms Union non‐participants were surprised by how few coworkers saved Problem: Union enrollment is decreasing in peer info number peer info number Possible mechanisms Union non‐participants believe their optimal g g y savings choice is negatively correlated with coworkers’ savings choices Problem: Why don’t union participants believe the same thing? g Problem: Zero correlation seems plausible, but Problem: Zero correlation seems plausible, but negative correlation may not be Potential mechanisms Union members were discouraged and de‐ yp motivated by peer information Union members distrust messages sent to Union members distrust messages sent to them by the company Possible mechanisms Union non‐participants distrust messages y p y sent to them by the company Problem: Why don’t union participants react the same way? the same way? Main results: Non‐union Enrollment increases when peer info given Enrollment increases with the participation rate in the comparison participation rate in the comparison group Main results: Union Enrollment decreases when peer info given info given Enrollment decreases with the participation rate in the comparison participation rate in the comparison group Conclusion Social norms marketing can backfire Social norms marketing can backfire for certain subgroups. Main results Union and non‐union participants No significant peer info effect Quick Enrollment Q Simplified enrollment mechanism 1. Check box on form 2. Return in pre‐addressed, postage‐ paid envelope paid envelope 3. You will be enrolled at – 6% before‐tax contribution rate – Age Age‐appropriate appropriate target retirement fund target retirement fund Q Quick Enrollment effect at another firm Fraction E Ever Participating in Sa avings Plan n 50% 40% 30% 20% 10% 0% 0 3 6 9 12 15 18 21 24 27 Time Since Baseline (months) Std. enrollment (2000) Std. enrollment (2002) Std. enrollment (2001) 30 33 Q Quick Enrollment effect at another firm Fraction E Ever Participating in Sa avings Plan n 50% 40% 30% 20% 10% 0% 0 3 6 9 12 15 18 21 24 27 30 Time Since Baseline (months) Std. enrollment (2000) Std. enrollment (2002) Std. enrollment (2001) Quick enrollment (2003) 33 Easy y Escalation Simplified way to raise contribution rate 1. Check box on form 2 Return in pre‐addressed, postage‐ 2. Return in pre addressed postage paid envelope 3. Your contribution rate will be raised to 6% raised to 6% New rate <6% New rate 6% Date New rate >6% Dec-05 Sep-05 Jun-05 Mar-05 Dec-04 Sep-04 Jun-04 Mar-04 Dec-03 Sep-03 Jun-03 Mar-03 F Fraction of participa ants witth contribution rate e < 6% Easy y Escalation effect at another firm 14% 12% 10% 8% 6% 4% 2% 0% Sample p characteristics No peer information 69. Percent male Panel A: Non-participants U i Union Information Information about 5-yr. about 10-yr. age group age group 66.0 61.6 No peer information 77.2 Non-union N i Information about 5-yr. age group 75.4 Information about 10-yr. age group 75.4 Age Mean Std. dev. 41.0 13.4 40.7 13.4 41.0 13.4 40.4 11.4 41.2 11.6 41.3 12.2 Tenure at company (years) Mean Std. dev. 9.4 12.0 9.5 12.1 9.0 12.1 7.2 9.3 7.5 9.5 7.8 8.4 Annual salary Mean Std. dev. 36 6 36.6 17.7 36 2 36.2 16.2 36.9 36 9 19.9 44.6 44 6 21.6 45.2 45 2 22.8 43.0 43 0 20.8 N = 343 N = 347 N = 349 N = 136 N = 130 N = 134 Sample size Sample p characteristics Panel B: Participants with low contribution rates Union Information Information No peer about 5-yr. about 10-yr. No peer information age group age group information 61.3 51.8 52.0 67.7 Percent male Non union Non-union Information about 5-yr. age group 67.6 Information about 10-yr. age group 69.5 Age Mean Std. dev. 39.9 39 9 11.7 40.8 40 8 11.8 41.1 41 1 11.9 41.8 41 8 10.6 42.0 42 0 10.7 42.0 42 0 10.5 Tenure at company (years) Mean Std. dev. 11.4 10.1 10.6 9.8 10.5 10.6 10.7 10.2 10.5 9.6 11.1 9.9 Annual salary (thousands of $) Mean Std. dev. 43.4 16.4 41.5 13.7 40.6 14.6 56.3 30.6 54.9 25.1 57.0 28.4 Before-tax cont. rate Mean Std. dev. 2.5 1.8 2.5 1.8 2.6 1.8 1.9 1.7 1.8 1.8 1.8 1.8 Before-tax balance Mean Std dev. Std. dev 18.0 29.6 29 6 16.8 26.4 26 4 17.4 32 8 32.8 33.2 61 8 61.8 29.1 53.9 53 9 31.5 55 1 55.1 N = 235 N = 255 N = 256 N = 931 N = 916 N = 911 Sample size