Small Cues Change Savings Choices James J. Choi Emily Haisley - - PowerPoint PPT Presentation

Small Cues Change Savings Choices

James J. Choi Emily Haisley Jennifer Kurkoski Cade Massey

Enormous cross-sectional variation in wealth accumulation

In 5th earnings decile: 90th percentile of wealth is 35 times 10th 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%

• 1.4%

0.3% 0.4% High cue Low cue Raise match rate by 125% (Kusko, Poterba, and Wilcox, 1998) Raise match threshold by 2% (Choi et al., 2002)

Contribution rate effects

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

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000 $9,000 $0 $2,500 $5,000 $7,500 $10,000 $12,500 $15,000 Company match 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:

1.

Log in to Vanguard, our 401(k) vendor. (If you’ve never logged in before, you will need the [Firm] plan number, XXX.)

2.

Click on “Change paycheck deductions” under the “I want to…” menu

3.

Adjust your percentages in the boxes

4.

Click continue and follow directions until you see the confirmation page. A confirmation will also be emailed

• r mailed to you.

Happy saving!

• X

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

• n what the right contribution increase is for you.)

60% threshold treatment You can contribute up to 60% of your income in any pay period.

2009 treatments

Company match in 2009

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000 $9,000 $0 $2,500 $5,000 $7,500 $10,000 $12,500 $15,000 Company match 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 968 60% threshold 252 651 971

Data

Administrative data from retirement plan administrator

0% 2% 4% 6% 8% 10% 12% 14% 3-Jan-08 3-Mar-08 3-May-08 3-Jul-08 3-Sep-08 3-Nov-08 3-Jan-09 3-Mar-09 3-May-09 3-Jul-09 3-Sep-09 3-Nov-09 3-Jan-10 3-Mar-10 3-May-10 3-Jul-10 3-Sep-10

• Avg. before-tax + after-tax + Roth
• contrib. rate

Last payday before email

• Avg. total contribution rate,

all control groups

Auto-escalation Bonus paid Auto-escalation Bonus paid

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

2009 anchor effect

• 4%
• 3%
• 2%
• 1%

0% 1% 2% 3% 4% 5% 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 19-Mar-10 2-Apr-10 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 Control 1% anchor 1.4%** 1.1%*

2009 anchoring effect averaged across time

November 27 – February 19 (7 paydays) March 5 bonus payday March 19 – October 15 (16 paydays) 1% anchor

• 0.8%*

(p = 0.047) 0.05% (p = 0.933)

• 0.8%

(p = 0.076)

1% anchor effect on prob. of contribution rate different than 11/13/09

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

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 19-Mar-10 2-Apr-10 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

Control 1% anchor

0% 1% 2% 3% 4% 5% 6% 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 19-Mar-10 2-Apr-10 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

$0 - $2,500 projected contributions group

Control 60% threshold

60% threshold effect

2.9%*

60% threshold effect averaged across time

November 27 – February 19 (7 paydays) March 5 bonus payday March 19 – October 15 (16 paydays) 60% threshold 1.8%* (p = 0.012) 1.4% (p = 0.153) 0.7% (p = 0.366)

60% threshold effect on prob. of contribution rate different than 11/13/09

0% 10% 20% 30% 40% 50% 60% 70% 80%

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 19-Mar-10 2-Apr-10 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

$0 - $2,500 projected contributions group

Control 60% threshold

0% 1% 2% 3% 4% 5% 6% 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 19-Mar-10 2-Apr-10 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

$2,501 - $5,000 projected contributions group

Control 60% threshold

60% threshold effect

• 4%
• 3%
• 2%
• 1%

0% 1% 2% 3% 4% 5% 6% 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 19-Mar-10 2-Apr-10 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

$5,001 - $16,500 projected contributions group

Control 60% threshold

60% threshold effect

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

Company match in 2009

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000 $9,000 $0 $2,500 $5,000 $7,500 $10,000 $12,500 $15,000 Company match Employee contribution

Why the different treatment effects?

• Salary
• Match rate
• Baseline savings desire
• Gap between status quo and cue

Treatment interactions

Dependent variable Contribution rate difference relative to 11/13/09 Explanatory variables 60% threshold dummy Pre-email contribution rate = 0-1% dummy Interactions of the above

60% threshold interactions, $0- $2,499 projected contributions

Panel A: $0 - $2,499 projected 2009 contributions 11/27/09 12/11/09 12/24/09 1/8/10 1/22/10 2/5/10 2/19/10 3/5/10 60% threshold 0.956 1.400 1.715 1.523 0.127 0.420 0.312 0.821 (1.172) (1.344) (1.541) (1.408) (1.080) (0.972) (0.975) (1.267) 60% threshold 3.929* 3.748+ 3.460 3.099 1.577 1.390 1.300 1.843 × 0-1% rate (1.747) (2.003) (2.297) (2.106) (1.615) (1.451) (1.455) (1.888) 0-1% rate 0.911 2.135 3.275* 2.599+ 2.321* 1.977+ 2.193* 2.431+ (1.219) (1.397) (1.603) (1.464) (1.124) (1.010) (1.012) (1.315) Constant 0.554 0.708 0.900 1.140 0.295

• 0.297
• 0.305
• 0.575

(0.857) (0.982) (1.125) (1.029) (0.788) (0.710) (0.712) (0.926)

60% threshold interactions, $2,500-$4,499 projected contributions

Panel B: $2,500 - $4,999 projected 2009 contributions 11/27/09 12/11/09 12/24/09 1/8/10 1/22/10 2/5/10 2/19/10 3/5/10 60% threshold

• 0.275
• 0.568
• 0.371
• 0.178

0.151

• 0.163
• 0.046

0.354 (0.415) (0.520) (0.533) (0.450) (0.384) (0.357) (0.348) (0.587) 60% threshold 3.519+ 5.656* 5.615* 9.083** 8.320** 7.032** 5.665** 6.840* × 0-1% rate (2.099) (2.630) (2.694) (2.263) (1.931) (1.819) (1.770) (2.988) 0-1% rate 1.052

• 0.568

0.089

• 1.234

0.005

• 0.087

0.037 0.057 (1.626) (0.520) (2.086) (1.752) (1.495) (1.422) (1.384) (2.335) Constant 1.329** 2.275** 2.292** 2.330** 1.805** 1.937** 1.813** 2.343** (0.292) (0.366) (0.376) (0.317) (0.271) (0.252) (0.245) (0.414)

60% threshold interactions, $5,000-$16,499 projected contributions

Panel B: $5,000 - $16,499 projected 2009 contributions 11/27/09 12/11/09 12/24/09 1/8/10 1/22/10 2/5/10 2/19/10 3/5/10 60% threshold

• 0.167

0.340 0.204

• 0.304
• 0.177
• 0.304
• 0.103
• 0.494

(0.476) (0.574) (0.614) (0.568) (0.489) (0.465) (0.453) (0.587) 60% threshold 6.305* 3.386 4.984 2.024 6.624* 5.959* 2.724

• 2.283

× 0-1% rate (2.968) (3.575) (3.816) (3.526) (3.029) (2.917) (2.841) (3.684) 0-1% rate 7.077** 5.338* 4.267+ 8.528** 9.352** 11.484** 13.995** 18.200** (1.953) (2.353) (2.512) (2.321) (1.994) (1.896) (1.846) (2.394) Constant 2.785** 3.800** 3.182** 4.162** 2.337** 1.861** 1.385** 1.386** (0.337) (0.407) (0.435) (0.403) (0.347) (0.330) (0.321) (0.417)

Our leading hypothesis

Larger gap between current and cued savings level is more motivating

Questions for 2010 experiments

• Are high savings level cues that are exogenously

further from status quo more motivating?

• What is the effect of high anchors?
• Can any cue significantly increase average

contributions of those with high YTD contributions?

Company match in 2010

$0 $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000 $9,000 $0 $2,500 $5,000 $7,500 $10,000 $12,500 $15,000 Company match Employee contribution

$3,000 threshold treatment

The next $X of contributions you make between now and December 31 will be matched at a 100% rate.

$16,500 threshold treatment

Contributing $Y more between now and December 31 would earn you the maximum possible match.

2010 treatments

3% anchor treatment

For example, you could increase your contribution rate by 3% of your income …

10% anchor treatment

… by 10% of your income …

20% anchor treatment

… by 20% of your income …

2010 treatments

$7,000 goal treatment

For example, suppose you set a goal to contribute $7,000 for the year and you attained it. You would earn $X more in matching money this year than you’re currently on pace for.

$11,000 goal treatment

For example, suppose you set a goal to contribute $11,000 for the year and you attained it. You would earn $Y more in matching money this year than you’re currently on pace for.

2010 treatments

People per cell in 2010

Projected 2010 contributions < $3,000 $3,000 - $5,999 $6,000 - $16,499 Control 263 560 3% anchor 561 10% anchor 562 20% anchor 565 $7,000 goal 263 $11,000 goal 262 $3,000 threshold 226 $16,500 threshold 225

2010 contributions

0% 5% 10% 15% 20% 25% $0-$299 $300-$599 $600-$899 $900-$1,199 $1,200-$1,499 $1,500-$1,799 $1,800-$2,099 $2,100-$2,399 $2,400-$2,699 $2,700-$2,999 $3,000-$3,299 $3,300-$3,599 $3,600-$3,899 $3,900-$4,199 $4,200-$,499 ≥ $4,500

< $3,000 projected contributions group

$3,000 threshold $16,500 threshold $3,000 threshold

Dollar threshold effect

0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 4.5% 5.0% 15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11 Total contribution rate change relative to 11/13/09 $3,000 threshold $16,500 threshold 1.5%*

Dollar threshold effect averaged across time

January 7 – July 8 non-bonus paydays March 11 bonus payday $16,500 threshold 0.9%* (p = 0.043) 0.7% (p = 0.359)

Goal effect

0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 4.5% 15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11 Total contribution rate change relative to 11/13/09 Control $7,000 goal $11,000 goal 2.2%*

Goal effect averaged across time

October 29 – March 4 (10 paydays) March 11 bonus payday March 18 – September 2 (13 paydays) $7,000 goal 0.0% (p = 0.989)

• 1.5%

(p = 0.145) 0.0% (p = 0.912) $11,000 goal 1.1%* (p = 0.044) 0.1% (p = 0.893)

• 0.3%

(p = 0.369)

• 4%
• 2%

0% 2% 4%

15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11

3% anchor

• 4%
• 2%

0% 2% 4%

15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11

10% anchor

• 4%
• 2%

0% 2% 4%

15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11

20% anchor

• 3%
• 2%
• 1%

0% 1% 2% 3% 4% 15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11 Total contribution rate change relative to 11/13/09 Control 3% anchor 10% anchor 20% anchor

2010 anchor effect

1.9%**

2010 anchoring effect averaged across time

October 29 – March 4 (10 paydays) March 11 bonus payday March 18 – September 2 (13 paydays) 3% anchor

• 0.2%

(p = 0.451) 0.5% (p = 0.588) 1.1%* (p = 0.028) 10% anchor

• 0.2%

(p = 0.458) 0.7% (p = 0.421) 1.1%* (p = 0.031) 20% anchor

• 0.1%

(p = 0.836) 0.5% (p = 0.521) 1.0%* (p = 0.019)

0% 10% 20% 30% 40% 50% 60% 70% 15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11

3% anchor

0% 10% 20% 30% 40% 50% 60% 70% 15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11

10% anchor

0% 10% 20% 30% 40% 50% 60% 70% 15-Oct-10 29-Oct-10 12-Nov-10 26-Nov-10 10-Dec-10 24-Dec-10 7-Jan-11 21-Jan-11 4-Feb-11 18-Feb-11 4-Mar-11 18-Mar-11 1-Apr-11 15-Apr-11 29-Apr-11 13-May-11 27-May-11 10-Jun-11 24-Jun-11 8-Jul-11 22-Jul-11 5-Aug-11 19-Aug-11 2-Sep-11

20% anchor

Anchoring results summary

Pre-bonus Post-bonus 1% anchor – – 3% anchor + 10% anchor + 20% anchor +

Joint significance test

Null hypothesis: {Pre-bonus, bonus, post- bonus} × {cue} × {marginal match category} treatment effects jointly equal 0 p-value on 30 treatment coefficients: 0.021

Summary

Anchors raise savings by up to 1.9% of income or lower savings by up to 1.4% of income High thresholds raise savings by up to 2.9%

• f income

High goals raise savings by up to 2.2% of income

Summary

We have intent-to-treat estimates; true effects are bigger Effects are long-lasting Bad news for econometricians wanting to explain variation Good news for policymakers wanting to influence savings behavior

External validity

Goda, Manchester, and Sojourner (2012)

• University of Minnesota employees
• Sent brochure on retirement savings plan
• Treatments: Info on increased

contributions on

– Expected balances at retirement – Expected annual income in retirement

Axis variation

$50 $100 $250 + $19,000 + $38,000 + $94,000 $0

Contribution rate changes

0.09% 0.27% 0.11% 0.30%

Balance (low axis) Balance (high axis) Income (low axis) Income (high axis)

0.19% of income effect from high vs. low axis