Small Cues Change Savings Choices James J. Choi Emily Haisley Jennifer Kurkoski Cade Massey
Enormous cross-sectional variation in wealth accumulation In 5 th earnings decile: 90 th percentile of wealth is 35 times 10 th percentile of wealth (Households aged 51-61; Venti and Wise, 1998)
Why does wealth vary so much? “Thus the primary determinant of the dispersion of wealth at retirement is evidently the choice to save or spend while young. We find no evidence that chance plays a large role .” ― Venti and Wise (1998)
Why do savings choices vary so much? Not explained by stable differences in • Time discount rates • Risk tolerance • Exposure to uncertainty • Perceived life expectancy • Tastes for goods complementary to leisure in retirement • Work-related expenses • Expected defined benefit pension and Social Security income in retirement (Bernheim, Skinner, and Weinberg, 2001)
Behavioral factors Financial literacy (Lusardi and Mitchell, 2007a,b) Propensity to plan (Ameriks, Caplin, and Leahy, 2003)
Social psychology Mischel (1968) : Transitory situational factors are much more powerful determinants of choice than fixed personal traits
This paper Field experiment : Randomized savings cues in retirement savings plan e-mails Design of cues inspired by past psychology and behavioral economics literature that predicts choices will assimilate to the cue High savings cue → Higher savings choice Low savings cue → Lower savings choice Nontrivial portion of wealth variation may be due to variation in cue exposure
Cue effects are large 2.9% Contribution rate effects 0.4% 0.3% -1.4% High cue Low cue Raise match rate Raise match by 125% (Kusko, threshold by 2% Poterba, and (Choi et al., 2002) Wilcox, 1998)
Not just about overcoming inertia Cues affect choices even when probability of action is unchanged → Ss models of infrequent attention and adjustment in savings are useful but incomplete (Duffie and Sun, 1990; Grossman and Laroque, 1990; Lynch, 1996; Gabaix and Laibson, 2002; Sims, 2003; Reis, 2006; Carroll et al., 2009; Abel, Eberly, and Panageas, 2007, 2012; Alvarez, Guiso, and Lippi, 2012)
Two rounds of emails November 17, 2009 First round of emails sent October 19, 2010 Second round of emails sent
Benefits director’s e -mail Dear James, We want to remind you that [Firm] matches your qualified contributions (pre-tax and Roth) to the [Firm] 401(k) Plan. In other words, [Firm] will give you free money for saving in your 401(k). What is the [Firm] match? [Firm] matching contribution is the greater of: (a) 100% of your qualified 2009 401(k) contributions up to $2,500; or (b) 50% of your qualified 2009 contributions up to $16,500 for a total possible match of $8,250.
Company match in 2009 $9,000 $8,000 $7,000 Company match $6,000 $5,000 $4,000 $3,000 $2,000 $1,000 $0 $0 $2,500 $5,000 $7,500 $10,000 $12,500 $15,000 Employee contribution
Benefits director’s e -mail Where am I at right now? You’ve contributed $X in pre -tax plus Roth contributions to the [Firm] 401(k) Plan so far this year. To take greater advantage of [Firm’s] 2009 match, increase your contribution rate for the remaining six weeks of 2009. [TREATMENT TEXT, IF ANY, INSERTED HERE]
How do I increase my contribution? To change your contribution rate at any time, follow these steps: Log in to Vanguard, our 401(k) vendor. (If you’ve never 1. logged in before, you will need the [Firm] plan number, XXX.) Click on “Change paycheck deductions” under the “I 2. want to…” menu Adjust your percentages in the boxes 3. Click continue and follow directions until you see the 4. confirmation page. A confirmation will also be emailed or mailed to you. Happy saving! - X
2009 treatments 1% anchor treatment For example, you could increase your contribution rate by 1% of your income and get more of the match money for which you’re eligible. ( 1% is just an example, and shouldn’t be interpreted as advice on what the right contribution increase is for you.) 60% threshold treatment You can contribute up to 60% of your income in any pay period.
Company match in 2009 $9,000 $8,000 $7,000 Company match $6,000 $5,000 $4,000 $3,000 $2,000 $1,000 $0 $0 $2,500 $5,000 $7,500 $10,000 $12,500 $15,000 Employee contribution
People per cell in 2009 Projected 2009 contributions < $2,500 $2,500 - $4,999 $5,000 - $16,499 Control 257 651 973 1% anchor 0 0 968 60% threshold 252 651 971
Data Administrative data from retirement plan administrator
Avg. before-tax + after-tax + Roth Avg. total contribution rate, contrib. rate 10% 12% 14% 0% 2% 4% 6% 8% 3-Jan-08 all control groups 3-Mar-08 3-May-08 3-Jul-08 3-Sep-08 Auto-escalation 3-Nov-08 Bonus paid 3-Jan-09 3-Mar-09 3-May-09 3-Jul-09 3-Sep-09 before email Last payday Auto-escalation 3-Nov-09 3-Jan-10 3-Mar-10 Bonus paid 3-May-10 3-Jul-10 3-Sep-10
Econometric approaches Event study • Separate regression for each post-email pay period • Dependent variable: Difference between contribution rate in pay period and last pre- email contribution rate Averaged effect across pay periods • Dependent variable: Difference between average contribution rate over multiple periods and last pre-email contribution rate
Statistical significance symbols Significant at 10% level Significant at 5% level Significant at 1% level
Total contribution rate change relative to 11/13/09 -4% -3% -2% -1% 0% 1% 2% 3% 4% 5% 13-Nov-09 2009 anchor effect 27-Nov-09 11-Dec-09 25-Dec-09 8-Jan-10 22-Jan-10 5-Feb-10 1.4%** Control 19-Feb-10 5-Mar-10 19-Mar-10 2-Apr-10 1% anchor 16-Apr-10 30-Apr-10 14-May-10 28-May-10 11-Jun-10 25-Jun-10 9-Jul-10 23-Jul-10 6-Aug-10 20-Aug-10 3-Sep-10 17-Sep-10 1-Oct-10 1.1%* 15-Oct-10
2009 anchoring effect averaged across time November 27 – March 5 March 19 – February 19 bonus October 15 (7 paydays) payday (16 paydays) 1% anchor -0.8%* 0.05% -0.8% ( p = 0.047) ( p = 0.933) ( p = 0.076)
100% 10% 20% 30% 40% 50% 60% 70% 80% 90% rate different than 11/13/09 1% anchor effect on prob. of contribution 0% 13-Nov-09 27-Nov-09 11-Dec-09 25-Dec-09 8-Jan-10 22-Jan-10 5-Feb-10 19-Feb-10 5-Mar-10 Control 19-Mar-10 2-Apr-10 16-Apr-10 1% anchor 30-Apr-10 14-May-10 28-May-10 11-Jun-10 25-Jun-10 9-Jul-10 23-Jul-10 6-Aug-10 20-Aug-10 3-Sep-10 17-Sep-10 1-Oct-10 15-Oct-10
Total contribution rate change relative to 11/13/09 0% 1% 2% 3% 4% 5% 6% 13-Nov-09 60% threshold effect 27-Nov-09 $0 - $2,500 projected contributions group 11-Dec-09 25-Dec-09 2.9%* 8-Jan-10 22-Jan-10 Control 5-Feb-10 19-Feb-10 5-Mar-10 19-Mar-10 2-Apr-10 60% threshold 16-Apr-10 30-Apr-10 14-May-10 28-May-10 11-Jun-10 25-Jun-10 9-Jul-10 23-Jul-10 6-Aug-10 20-Aug-10 3-Sep-10 17-Sep-10 1-Oct-10 15-Oct-10
60% threshold effect averaged across time November 27 – March 5 March 19 – February 19 bonus October 15 (7 paydays) payday (16 paydays) 60% 1.8%* 1.4% 0.7% threshold ( p = 0.012) ( p = 0.153) ( p = 0.366)
10% 20% 30% 40% 50% 60% 70% 80% contribution rate different than 11/13/09 60% threshold effect on prob. of 0% 13-Nov-09 27-Nov-09 11-Dec-09 25-Dec-09 $0 - $2,500 projected contributions group 8-Jan-10 22-Jan-10 5-Feb-10 19-Feb-10 Control 5-Mar-10 19-Mar-10 2-Apr-10 16-Apr-10 60% threshold 30-Apr-10 14-May-10 28-May-10 11-Jun-10 25-Jun-10 9-Jul-10 23-Jul-10 6-Aug-10 20-Aug-10 3-Sep-10 17-Sep-10 1-Oct-10 15-Oct-10
Total contribution rate change relative to 11/13/09 0% 1% 2% 3% 4% 5% 6% 13-Nov-09 60% threshold effect 27-Nov-09 $2,501 - $5,000 projected contributions group 11-Dec-09 25-Dec-09 8-Jan-10 22-Jan-10 Control 5-Feb-10 19-Feb-10 5-Mar-10 19-Mar-10 2-Apr-10 60% threshold 16-Apr-10 30-Apr-10 14-May-10 28-May-10 11-Jun-10 25-Jun-10 9-Jul-10 23-Jul-10 6-Aug-10 20-Aug-10 3-Sep-10 17-Sep-10 1-Oct-10 15-Oct-10
Total contribution rate change relative to 11/13/09 -4% -3% -2% -1% 0% 1% 2% 3% 4% 5% 6% 13-Nov-09 60% threshold effect 27-Nov-09 $5,001 - $16,500 projected contributions group 11-Dec-09 25-Dec-09 8-Jan-10 22-Jan-10 Control 5-Feb-10 19-Feb-10 5-Mar-10 19-Mar-10 2-Apr-10 60% threshold 16-Apr-10 30-Apr-10 14-May-10 28-May-10 11-Jun-10 25-Jun-10 9-Jul-10 23-Jul-10 6-Aug-10 20-Aug-10 3-Sep-10 17-Sep-10 1-Oct-10 15-Oct-10
Why the different treatment effects? • Salary • Match rate • Baseline savings desire • Gap between status quo and cue
Salary × treatment interactions are insignificant Entire 2009 sample excluding 1% anchor recipients 11/27/09 12/11/09 12/24/09 1/8/10 60% threshold 0.269 0.471 0.474 0.250 (0.310) (0.372) (0.398) (0.362) 60% threshold -0.031 -0.059 -0.707 -0.348 × log(salary) (0.785) (0.943) (1.008) (0.917) Log(salary) 1.782** 2.338** 2.202** 2.436** (0.560) (0.674) (0.720) (0.656) Constant 2.229** 3.170** 2.938** 3.506** (0.219) (0.263) (0.281) (0.256) 10% salary increase decreases treatment effect by at most 0.07% of income
Why the different treatment effects? • Salary • Match rate • Baseline savings desire • Gap between status quo and cue
Recommend
More recommend
