This document summarizes the results of a study on how framing effects influence retirement decisions. The study used hypothetical questions from surveys in the Netherlands and US to examine how individuals' preferences for pension contributions and investment profiles are affected by how the choices are framed or presented. The results show strong evidence that framing effects, choice set effects, and the standard option presented influence individuals' decisions. Preferences were also found to be more vulnerable to framing for more complex decisions. The findings suggest communication around retirement choices needs to properly frame options to avoid unintended influences on individuals' decisions.
Federica Teppa, Maarten van Rooij. Are Retirement Decisions Vulnerable to Framing Effects?
1. Are Retirement Decisions Vulnerable
to Framing Effects?
Empirical Evidence from NL and US
Federica Teppa
(De Nederlandsche Bank & Netspar)
Maarten van Rooij
(De Nederlandsche Bank & Netspar)
Bank of Estonia seminar
Tallinn - September, 2013
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
1 / 29
2. 1. Introduction and Motivation
The rapidly ageing trend experienced by the population in all
developed countries in the recent years and in the years to come has
put public finances into severe distress.:
1
OECD : in 2010 the largest proportion of retirement income (some
60 percent) is provided by the state, mostly via PAYG systems
where benefits are paid out of current taxes
The need for cutting the costs of public pension provisions has been at
the basis of a number of pension reforms taking place recently.
1
Governments are no longer going to act as the main (or even the
only) provider of old-age income, but more and more limiting their
goal as to ensure a minimum level of adequate income to the
elderly
2
Shifting risks to the individuals who are becoming increasingly
responsible for the fraction of their retirement income exceeding
the publicly provided component.
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
2 / 29
3. 1. Introduction and Motivation
Behavioral aspects of the individual decision making process are being
studied extensively.
1
2
Increasing availability of both administrative records and survey
data on real and hypothetical choices in several countries
Combination of several disciplines
economics (Butler and Teppa, 2007; Agnew et al., 2008; Brown et
¨
al., 2008)
psychology (Kahneman and Miller, 1986; Gonzales et al., 2005),
neurology (Rogers et al., 1999; De Martino et al., 2006)
sociology
3
Final outcomes of any decision making process are very often
contingent upon how alternatives are displayed to the agents
The significant role of this framing effect represents probably the most
striking violation of rationality in standard economic theory.
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
3 / 29
4. 1. Introduction and Motivation
Brown et al. (2011)
RAND American Life Panel (ALP)
Participants were confronted with alternative information formats
about how benefits would be adjusted if they were to claim
retirement benefits early versus later
Individual intentions with regard to Social Security claiming ages
are sensitive to how the early versus late claiming decision is
framed
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
4 / 29
5. 1. This paper
1
Methodology
Hypothetical questions from DHS and ALP in 2006
Individual preferences over pension premium contributions and
pension savings investment profiles
2
Main findings
Strong role of framing effect and choice set effect; less role of order
effect
Strong and robust role of standard option for both pension premium
contributions and pension savings investment profiles
Evidence of consistency between hypothetical and actual
preferences (e.g. portfolio holdings)
Positive correlation btw framing effects and complexity of decisions
to be made
3
Relevance and policy implications
Context of increasing responsibility of individuals decisions and
increasing complexity
Importance of communication, as well as regulation and
supervision on how alternative choices are presented to individuals
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
5 / 29
7. 2. The data - DHS
Table 1: DNB Household Survey response rates by questionnaire
Questionnaire
Questionnaire 1
Questionnaire 2
Questionnaire 3
Questionnaire 4
Contacted
2,476
2,485
2,528
2,284
Completed
1,648
1,722
1,915
1,590
Response rate (%)
66.5
69.2
75.7
70.7
When fielded
June 2-6, 2006
June 30-July 3, 2006
June 16-20, 2006
July 14-18, 2006
Source: DHS, 2006
- The second column reports the number of panel members aged 16 years and older selected
and contacted;
- The third column reports the number of selected panel members who returned the completed
questionnaire;
- The response rate is simply the ratio between column three and column two.
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
7 / 29
8. 2. DHS questionnaires
1
Object
Pension premium contributions
Pension savings asset allocation
2
Presentation
Large vs restricted choice set
High vs low numbers
Labels vs numbers
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
8 / 29
9. 2. DHS questionnaires
Pension premium contributions
Q11. “Restricted choice set” and “Low numbers”:
Imagine you have a job with a defined contribution pension plan and
you have to decide how much to contribute. If you pay a higher
contribution rate your net wage will be lower, but your final pension
income will be higher. The standard contribution rate that you pay is
12% of your gross wage. However, you can also opt for a high
contribution rate (14%) or a low contribution rate (10%).
Which plan would you choose?
− Standard contribution rate
− High contribution rate
− Low contribution rate
− I do not want to say it
− I do not know
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
9 / 29
10. 2. DHS questionnaires
Q12. “Large choice set” and “Low numbers”:
Imagine you have a job with a defined contribution pension plan and
you have to decide how much to contribute. If you pay a higher
contribution rate your net wage will be lower, but your final pension
income will be higher. The standard contribution rate that you pay is
12% of your gross wage. However, you can also opt for a very high
contribution rate (16%) or a high contribution rate (14%) or a low
contribution rate (10%) or a very low contribution rate (8%).
Which plan would you choose?
− Standard contribution rate
− Very high contribution rate
− High contribution rate
− Low contribution rate
− Very low contribution rate
− I do not want to say it
− I do not know
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
10 / 29
11. 2. DHS questionnaires
Q13. “Restricted choice set” and “High numbers”:
Imagine you have a job with a defined contribution pension plan and
you have to decide how much to contribute. If you pay a higher
contribution rate your net wage will be lower, but your final pension
income will be higher. The standard contribution rate that you pay is
20% of your gross wage. However, you can also opt for a high
contribution rate (22%) or a low contribution rate (18%).
Which plan would you choose?
− Standard contribution rate
− High contribution rate
− Low contribution rate
− I do not want to say it
− I do not know
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
11 / 29
12. 2. DHS questionnaires
Pension savings investment profiles
Q15. “Restricted choice set” and “Low numbers”:
Suppose you have a job and the pension system is reformed so that everyone
in responsible for his own retirement savings (through his company’s pension
fund) and everyone should tell how his pension fund contributions must be
invested in equities and bonds. The final pension depends on this decision
and the returns on financial markets. Now imagine that your pension will allow
you to tell how your pension contributions must be allocated into equities and
bonds. You can opt for a standard allocation (30% stocks-70% bonds).
However, you can ask your employer to change the portfolio allocation so that
you can have a risky profile (45% stocks-55% bonds) or a safe profile (15%
stocks-85% bonds). Which portfolio allocation would you choose?
− Standard investment profile
− Risky investment profile
− Safe investment profile
− I do not want to say it
− I Teppanot know
do & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
12 / 29
13. 2. DHS questionnaires
Q17. “Restricted choice set” and “High numbers”:
Suppose you have a job and the pension system is reformed so that everyone
in responsible for his own retirement savings (through his company’s pension
fund) and everyone should tell how his pension fund contributions must be
invested in equities and bonds. The final pension depends on this decision
and the returns on financial markets. Now imagine that your pension will allow
you to tell how your pension contributions must be allocated into equities and
bonds. You can opt for a standard allocation (70% stocks-30% bonds).
However, you can ask your employer to change the portfolio allocation so that
you can have a risky profile (85% stocks-15% bonds) or a safe profile (55%
stocks-45% bonds). Which portfolio allocation would you choose?
− Standard investment profile
− Risky investment profile
− Safe investment profile
− I do not want to say it
− I do not know
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
13 / 29
14. 2. DHS questionnaires
Q21. “Restricted choice set” and “Low numbers”:
Imagine you have a job with a defined contribution pension plan and
you have to decide how much to contribute. If you pay a higher
contribution rate your net wage will be lower, but your final pension
income will be higher.
Which plan would you choose?
− 10% contribution rate
− 12% contribution rate
− 14% contribution rate
− I do not want to say it
− I do not know
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
14 / 29
15. 3. Results
Table 2: Pension premium contributions and Pension savings investment profiles - labels
Restricted choice set
Standard
High
Low
I do not want to say it
I do not know
TOTAL
Large choice set
Standard
Very high
High
Low
Very low
I do not want to say it
I do not know
TOTAL
Teppa & van Rooij (DNB)
Pension premium contributions
Low numbers version
High numbers version
Freq. Percent
Cumulative
Freq. Percent
Cumulative
470
58.39
58.39
457
54.21
54.21
178
22.11
80.50
194
23.01
77.22
42
5.22
85.71
43
5.10
82.33
6
0.75
86.46
6
0.71
83.04
109
13.54
100
143
16.96
100
805
100
843
100
Freq. Percent
Cumulative
Freq. Percent
Cumulative
366
45.47
45.47
371
44.01
44.01
23
2.86
48.32
21
2.49
46.50
229
28.45
76.77
205
24.32
70.82
43
5.34
82.11
55
6.52
77.34
17
2.11
84.22
16
1.90
79.24
5
0.62
84.84
7
0.83
80.07
122
15.16
100
168
19.93
100
805
100
843
100
Retirement decisions and framing effects
September, 2013
15 / 29
16. 3. Results
Table 2: Pension premium contributions and Pension savings investment profiles - labels
Restricted choice set
Standard
Risky
Safe
I do not want to say it
I do not know
TOTAL
Large choice set
Standard
Very Risky
Risky
Safe
Very Safe
I do not want to say it
I do not know
TOTAL
Teppa & van Rooij (DNB)
Pension savings investment profiles
Low numbers version
High numbers version
Freq. Percent
Cumulative
Freq. Percent
Cumulative
289
35.50
35.50
250
29.98
29.98
47
5.77
41.28
31
3.72
33.69
307
37.71
78.99
389
46.64
80.34
8
0.98
79.98
7
0.84
81.18
163
20.02
100
157
18.82
100
814
100
834
100
Freq. Percent
Cumulative
Freq. Percent
Cumulative
231
28.38
28.38
216
25.90
25.90
6
0.74
29.12
3
0.36
26.26
46
5.65
34.77
36
4.32
30.58
283
34.77
69.53
288
34.53
65.11
72
8.85
78.38
119
14.27
79.38
9
1.11
79.48
6
0.72
80.10
167
20.52
100
166
19.90
100
814
100
834
100
Retirement decisions and framing effects
September, 2013
16 / 29
17. 3. Results
Table 3: Pension premium contributions and investment profiles - numbers
Restricted choice set
10 % (low) or 18 % (high)
12 % or 20 %
14 % or 22 %
I do not want to say it
I do not know
TOTAL
Large choice set
8 % (low) or 16 % (high)
10 % or 18 %
12 % or 20 %
14 % or 22 %
16 % or 24 %
I do not want to say it
I do not know
TOTAL
Teppa & van Rooij (DNB)
Pension premium contributions
Low numbers version
High numbers version
Freq.
Percent
Cumulative
Freq.
Percent
Cumulative
127
29.60
29.60
115
27.06
27.06
163
38.00
67.60
172
40.47
67.53
67
15.62
83.22
34
8.00
75.53
5
1.17
84.38
6
1.41
76.94
67
15.62
100
98
23.06
100
429
100
425
100
Freq.
Percent
Cumulative
Freq.
Percent
Cumulative
76
16.85
16.85
87
20.86
20.86
91
20.18
37.03
67
16.07
36.93
118
26.16
63.19
128
30.70
67.63
28
6.21
70.29
20
4.80
72.42
17
2.11
76.50
16
1.90
79.24
4
0.68
77.38
7
1.68
76.98
102
22.62
100
96
23.02
100
451
100
417
100
Retirement decisions and framing effects
September, 2013
17 / 29
18. 3. Results
Table 3: Pension premium contributions and investment profiles - numbers
Restricted choice set
15-85% (low) or 45-55% (high)
30-70% or 70-30%
45-55% or 85-15%
I do not want to say it
I do not know
TOTAL
Large choice set
0-100% (low) or 40-60% (high)
15-85% or 55-45%
30-70% or 70-30%
45-55% or 85-15%
60-40% or 100-0%
I do not want to say it
I do not know
TOTAL
Pension savings investment profiles
Low numbers version
Freq.
Percent
Cumulative
105
22.93
22.93
128
27.95
50.87
73
15.94
66.81
7
1.53
68.34
145
31.66
100
458
100
Freq.
Percent
Cumulative
29
7.06
7.06
54
13.14
20.19
106
25.79
45.99
55
13.38
59.37
32
7.79
67.15
7
1.70
68.86
128
31.14
100
411
100
Freq.
220
60
18
6
127
431
Freq.
137
82
52
17
10
4
120
422
High numbers version
Percent
Cumulative
51.04
51.04
13.92
64.97
4.18
69.14
1.39
70.53
29.47
100
100
Percent
Cumulative
32.46
32.46
19.43
51.90
12.32
64.22
4.03
68.25
2.37
70.62
0.95
71.56
28.44
100
100
Notes: For investment profiles, the percentages represent the combination stocks-bonds.
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
18 / 29
19. 3. Results
Table 4: Role of the standard option
Pension premium contributions
Freq.
Percent
Cumulative
Standard (14 %)
629
54.22
54.22
High (16 %)
304
26.21
80.43
Very high (18 %)
79
6.81
87.24
I do not want to say it
3
0.26
87.50
I do not know
145
12.50
100
TOTAL
1,160
100
Low (14 %)
66
5.77
5.77
Standard (16 %)
691
60.40
66.17
High (18 %)
253
22.12
88.29
I do not want to say it
15
1.31
89.60
I do not know
119
10.40
100
TOTAL
1,144
100
Standard (18 %)
875
73.16
73.16
Low (16 %)
126
10.54
83.70
Very low (14 %)
52
4.35
88.04
I do not want to say it
9
0.75
88.80
I do not know
134
11.20
100
TOTAL
1,196
100
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
19 / 29
20. 3. Results
Table 4: Role of the standard option
Pension savings investment profiles
Freq.
Percent
Standard (45% stocks-55% bonds)
313
26.71
Safe (30% stocks-70% bonds)
449
38.31
Very safe (15% stocks-85% bonds)
249
21.25
I do not want to say it
5
0.43
I do not know
156
13.31
TOTAL
1,172
100
Safe (15% stocks-85% bonds)
422
35.79
Standard (30% stocks-70% bonds)
520
44.11
Risky (45% stocks-55% bonds)
58
4.92
I do not want to say it
9
0.76
I do not know
170
14.42
TOTAL
1,179
100
Standard (15% stocks-85% bonds)
812
70.73
Risky (30% stocks-70% bonds)
143
12.46
Very risky (45% stocks-55% bonds)
23
2.00
I do not want to say it
13
1.13
I do not know
157
13.68
TOTAL
1,148
100
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
Cumulative
26.71
65.02
86.26
86.69
100
35.79
79.90
84.82
85.58
100
70.73
83.19
85.19
86.32
100
September, 2013
20 / 29
21. 3. Results
Table 5: Order, choice set and framing effects
Pension premium contributions
Low numbers version
p-value
N.Obs.
0.543
805
0.030
805
0.000
805
0.000
289
0.000
327
Full dataset
Order effect - 3 choices with labels
Order effect - 5 choices with labels
Choice set effect - 3 choices vs 5 choices with labels
Framing effect - labels vs numbers (3 choices)
Framing effect - labels vs numbers (5 choices)
W/o DKs
Order effect - 3 choices with labels
Order effect - 5 choices with labels
Choice set effect - 3 choices vs 5 choices with labels
Framing effect - labels vs numbers (3 choices)
Framing effect - labels vs numbers (5 choices)
Teppa & van Rooij (DNB)
0.781
0.008
0.000
0.000
0.000
690
678
655
229
235
Retirement decisions and framing effects
High numbers version
p-value
N.Obs.
0.098
843
0.423
843
0.000
843
0.000
311
0.000
295
0.020
0.250
0.000
0.000
0.000
694
668
640
215
196
September, 2013
21 / 29
22. 3. Results
Table 5: Order, choice set and framing effects
Pension savings investment profiles
Low numbers version
p-value
N.Obs.
0.061
814
0.095
814
0.000
814
0.000
313
0.000
294
Full dataset
Order effect - 3 choices with labels
Order effect - 5 choices with labels
Choice set effect - 3 choices vs 5 choices with labels
Framing effect - labels vs numbers (3 choices)
Framing effect - labels vs numbers (5 choices)
W/o DKs
Order effect - 3 choices with labels
Order effect - 5 choices with labels
Choice set effect - 3 choices vs 5 choices with labels
Framing effect - labels vs numbers (3 choices)
Framing effect - labels vs numbers (5 choices)
Teppa & van Rooij (DNB)
0.619
0.165
0.000
0.000
0.000
308
304
617
192
186
Retirement decisions and framing effects
High numbers version
p-value
N.Obs.
0.736
834
0.009
834
0.000
834
0.000
305
0.000
310
0.137
0.025
0.000
0.020
0.000
321
319
645
193
200
September, 2013
22 / 29
23. 3. Results
Table 6a: Role of standard option - Pension premium contributions - probit estimates
Variable
14 percent
16 percent
Coefficient
Coefficient
[Marg.eff.]
[Marg.eff.]
(Std. Err.)
(Std. Err.)
Labeled standard
1.027 ***
-0.177 *
[0.311]
[-0.069]
(0.125)
(0.099)
SAS financial literacy
0.114
0.025
[0.043]
[0.010]
(0.091)
(0.088)
Female
-0.153 *
0.082
[-0.056]
[0.032]
(0.080)
(0.077)
Age less than 40 years
0.346 **
-0.020
[0.132]
[-0.008]
(0.127)
(0.121)
Age between 40 and 64 years
0.224 **
-0.170 *
[0.083]
[-0.067]
(0.107)
(0.102)
Gross household income (in logs)
0.010
-0.183 **
[0.003]
[-0.072]
(0.088)
(0.086)
Total fin. assets (in logs)
-0.010
-0.016
[-0.003]
[-0.006]
(0.028)
(0.027)
Constant
-0.978
1.422 **
(0.721)
(0.705)
Log-likelihood
-749.708
-816.210
Pseudo R2
0.066
0.028
N.Obs.
1215
1215
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
18 percent
Coefficient
[Marg.eff.]
(Std. Err.)
0.208 **
[0.082]
(0.098)
-0.184 **
[-0.073]
(0.088)
-0.066
[-0.026]
(0.076)
-0.334 **
[-0.132]
(0.121)
-0.196 *
[-0.077]
(0.102)
0.006
[0.002]
(0.085)
0.045 *
[0.018]
(0.027)
-0.428
(0.701)
-817.545
0.024
1215
September, 2013
23 / 29
24. 3. Results
Table 6b: Role of standard option - Pension savings investment profile - probit estimates
Variable
45-55 percent
30-70 percent
Coefficient
Coefficient
[Marg.eff.]
[Marg.eff.]
(Std. Err.)
(Std. Err.)
Labeled standard
0.712 ***
-0.115
[0.158]
[-0.046]
(0.112)
(0.081)
SAS financial literacy
0.339 ***
-0.187 **
[0.094]
[-0.074]
(0.102)
(0.090)
Female
-0.209 **
-0.159 **
[-0.053]
[-0.063]
(0.094)
(0.078)
Age less than 40 years
0.263 *
0.020
[0.072]
[0.008]
(0.147)
(0.122)
Age between 40 and 64 years
0.191
0.024
[0.049]
[0.009]
(0.124)
(0.009)
Gross hh income (in logs)
0.032
0.063
[0.008]
[0.025]
(0.104)
(0.087)
Total fin. assets (in logs)
0.082 **
-0.038
[0.021]
[-0.015]
(0.033)
(0.028)
Constant
-2.118 **
0.205
(0.842)
(0.704)
Log-likelihood
-538.100
-801.155
Pseudo R2
0.092
0.023
N.Obs.
1184
1184
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
15-85 percent
Coefficient
[Marg.eff.]
(Std. Err.)
0.182 **
[0.069]
(0.084)
-0.163 *
[-0.061]
(0.091)
0.262 ***
[0.096]
(0.081)
-0.204
[-0.007]
(0.127)
-0.170
[-0.063]
(0.108)
-0.141
[-0.052]
(0.090)
0.012
[0.004]
(0.028)
1.087
(0.734)
-736.706
0.051
1184
September, 2013
24 / 29
28. 3. Results
Table 10: Order and framing effects - US data
Full dataset
Order effect - 3 choices with labels
Order effect - 5 choices with labels
Framing effect - labels vs numbers (3 choices)
Framing effect - labels vs numbers (5 choices)
W/o DKs
Order effect - 3 choices with labels
Order effect - 5 choices with labels
Framing effect - labels vs numbers (3 choices)
Framing effect - labels vs numbers (5 choices)
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
p-value
0.437
0.929
0.000
0.000
N.Obs.
526
506
526
506
0.158
0.814
0.000
0.000
486
470
468
454
September, 2013
28 / 29
29. 4. Concluding remarks
1
Main findings
Retirement choices are very sensitive to institutional/plan design
Strong role of framing effect and choice set effect; less role of order
effect
Strong and robust role of standard option for both pension premium
contributions and pension savings investment profiles
Consistency between hypothetical preferences and actual choices
2
Relevance and policy implications
Context of increasing responsibility of individuals decisions and
increasing complexity
Importance of communication, as well as regulation and
supervision on how alternative choices are presented to individuals
Teppa & van Rooij (DNB)
Retirement decisions and framing effects
September, 2013
29 / 29