Development of an Individual Measure of Loss Aversion

Development of an Individual
Measure of Loss Aversion
John W. Payne (Duke)
Suzanne B. Shu (UCLA and NBER)
Elizabeth C. Webb (Columbia)*
Namika Sagara (Duke)
Overview
Development of a model-free
individual-level measure of loss
aversion
3 Oct 2015
ACR Conference 2015
2
Overview
How loss aversion measure
correlates with other individuallevel measures
Development of a model-free
individual-level measure of loss
aversion
3 Oct 2015
ACR Conference 2015
3
Overview
How loss aversion measure
correlates with other individuallevel measures
Development of a model-free
individual-level measure of loss
aversion
Whether loss aversion measure is
predictive of consumer behavior
3 Oct 2015
ACR Conference 2015
4
Overview
How loss aversion measure
correlates with other individuallevel measures
Development of a model-free
individual-level measure of loss
aversion
Whether loss aversion measure is
predictive of consumer behavior
Tested on over 6,600
respondents across six studies
3 Oct 2015
ACR Conference 2015
5
Development of the Loss Aversion Measure
Existing Approaches
3 Oct 2015
ACR Conference 2015
6
Development of the Loss Aversion Measure
Existing Approaches
Assume an underlying model
of decisions over risk
Kahneman & Tversky, 1979; Tversky &
Kahneman, 1992; Toubia et al., 2013
3 Oct 2015
ACR Conference 2015
7
Development of the Loss Aversion Measure
Existing Approaches
Assume an underlying model
of decisions over risk
Use 50:50 two-outcome
gambles
Kahneman & Tversky, 1979; Tversky &
Kahneman, 1992; Toubia et al., 2013
3 Oct 2015
ACR Conference 2015
8
Development of the Loss Aversion Measure
Existing Approaches
Our Approach
Assume an underlying model
of decisions over risk
Use 50:50 two-outcome
gambles
Kahneman & Tversky, 1979; Tversky &
Kahneman, 1992; Toubia et al., 2013
3 Oct 2015
ACR Conference 2015
9
Development of the Loss Aversion Measure
Existing Approaches
Our Approach
Assume an underlying model
of decisions over risk
Model-free
Use 50:50 two-outcome
gambles
Kahneman & Tversky, 1979; Tversky &
Kahneman, 1992; Toubia et al., 2013
3 Oct 2015
ACR Conference 2015
10
Development of the Loss Aversion Measure
Existing Approaches
Our Approach
Assume an underlying model
of decisions over risk
Model-free
Use 50:50 two-outcome
gambles
Slightly more complex, mixed
three-outcome gambles
Kahneman & Tversky, 1979; Tversky &
Kahneman, 1992; Toubia et al., 2013
3 Oct 2015
Brooks & Zank, 2005
ACR Conference 2015
11
Development of the Loss Aversion Measure
1
2
3
3 Oct 2015
ACR Conference 2015
12
Development of the Loss Aversion Measure
Participants are asked to choose
between two gambles in each
step
1
2
3
3 Oct 2015
ACR Conference 2015
13
Development of the Loss Aversion Measure
Participants are asked to choose
between two gambles in each
step
1
2
One gamble in the pair is always
the more loss averse choice
3
3 Oct 2015
ACR Conference 2015
14
Development of the Loss Aversion Measure
1
69%
2
58%
3
44%
3 Oct 2015
ACR Conference 2015
Most respondents express
some degree of loss aversion
by preferring a loss-averse
gamble to a matched gainseeking gamble
15
Development of the Loss Aversion Measure
Participants are asked to choose
between two gambles in each
step
1
2
One gamble in the pair is always
the more loss averse choice
3
Yields an overall measure of loss
aversion per participant
3 Oct 2015
ACR Conference 2015
16
The Studies
Study 1
N = 832
48.7% female, Mage = 50.5
Study 4
N = 1,452
69.7% female, Mage = 53.3
Study 2
N = 1,432
49.8% female, Mage = 44.3
Study 5
N = 1,010
71.2% female, Mage = 57.4
Study 3
N = 1,113
49.7% female, Mage = 53.1
Study 6
N = 831
49.1% female, Mage = 35.1
3 Oct 2015
ACR Conference 2015
17
The Studies
Study 1
N = 832
48.7% female, Mage = 50.5
Study 4
N = 1,452
69.7% female, Mage = 53.3
Study 2
N = 1,432
49.8% female, Mage = 44.3
Study 5
N = 1,010
71.2% female, Mage = 57.4
Study 3
N = 1,113
49.7% female, Mage = 53.1
Study 6
N = 831
49.1% female, Mage = 35.1
3 Oct 2015
ACR Conference 2015
Across Studies
N = 6,670
18
The Studies
Study 1
N = 832
48.7% female, Mage = 50.5
Across Studies
N = 6,670
Study 4
N = 1,452
69.7% female, Mage = 53.3
Demographic Variables
Study 2
N = 1,432
49.8% female, Mage = 44.3
Study 5
N = 1,010
71.2% female, Mage = 57.4
Study 3
N = 1,113
49.7% female, Mage = 53.1
Study 6
N = 831
49.1% female, Mage = 35.1
3 Oct 2015
ACR Conference 2015
-
Age
Expected Age (Life Expectancy)
Gender
Subjective Health
19
The Studies
Study 1
N = 832
48.7% female, Mage = 50.5
Across Studies
N = 6,670
Study 4
N = 1,452
69.7% female, Mage = 53.3
Demographic Variables
Study 2
N = 1,432
49.8% female, Mage = 44.3
Study 5
N = 1,010
71.2% female, Mage = 57.4
-
Study 3
N = 1,113
49.7% female, Mage = 53.1
Study 6
N = 831
49.1% female, Mage = 35.1
Other Variables
- Social Security Solvency
- Intertemporal Patience
- Retirement Savings (Behaviors)
3 Oct 2015
ACR Conference 2015
Age
Expected Age (Life Expectancy)
Gender
Subjective Health
20
Results: Loss Aversion Measure, By Study (Raw Score)
3 Oct 2015
ACR Conference 2015
21
Results: Loss Aversion Measure, By Study (Raw Score)
Study 1
N = 832
Study 2
N = 1,474
Study 3
N = 1,113
Pct of Participants
30%
25%
20%
15%
10%
5%
0%
0
1
2
3
4
5
6
7
Loss Aversion Score
8
9
0
1
2
Study 4
N = 1,452
3
4 5 6 7 8
Loss Aversion Score
9
10 11
0
1
2
Study 5
N = 1,010
3
4
5
6
7
8
Loss Aversion Score
9
10
11
9
10
Study 6
N = 831
30%
Pct of Participants
25%
20%
15%
10%
5%
0%
0
3 Oct 2015
1
2
3
4
5
6
7
Loss Aversion Score
8
9
10
0
1
2
3
4
5
6
7
Loss Aversion Score
ACR Conference 2015
8
9
10
0
1
2
3
4
5
6
7
Loss Aversion Score
8
22
Results: Loss Aversion Measure, Across Studies 2 – 6
20%
N = 5,880
18%
16%
Pct of Participants
16%
14%
12%
12%
12%
12%
10%
9%
8%
6%
7%
7%
1
2
10%
9%
5%
4%
2%
0%
0
3
4
5
6
7
8
9
Loss Aversion Score
3 Oct 2015
ACR Conference 2015
23
Results: Loss Aversion Measure, Across Studies 2 – 6
20%
N = 5,880
18%
Pct of Participants
16%
14%
Loss aversion measure is
relatively normally distributed
– no clustering at either end of
the scale
16%
12%
12%
12%
12%
10%
9%
8%
6%
7%
7%
1
2
10%
9%
5%
4%
2%
0%
0
3
4
5
6
7
8
9
Loss Aversion Score
3 Oct 2015
ACR Conference 2015
24
Results: Lambda Across Studies 2 - 6
20%
N = 5,880
18%
16%
Pct of Participants
16%
16%
14%
12%
12%
12%
12%
8%
6%
10%
9%
10%
7%
5%
4%
2%
0%
0.33
0.83
1.00
1.25
1.75
2.25
2.75
4.00
5.00
Lambda Coefficient
3 Oct 2015
ACR Conference 2015
25
Results: Lambda Across Studies 2 - 6
20%
18%
16%
Pct of Participants
16%
16%
14%
Average lambda across
studies is 2.16
12%
12%
12%
12%
10%
9%
10%
8%
6%
N = 5,880
7%
5%
4%
2%
0%
0.33
0.83
1.00
1.25
1.75
2.25
2.75
4.00
5.00
Lambda Coefficient
3 Oct 2015
ACR Conference 2015
26
Results: Lambda Across Studies 2 - 6
20%
18%
16%
Pct of Participants
16%
16%
14%
The lambda coefficient shows a
jump from lambdas below one
(loss-seeking) to lambdas above
one (loss aversion)
12%
12%
12%
12%
10%
9%
10%
8%
6%
N = 5,880
7%
5%
4%
2%
0%
0.33
0.83
1.00
1.25
1.75
2.25
2.75
4.00
5.00
Lambda Coefficient
3 Oct 2015
ACR Conference 2015
27
Correlates of Loss Aversion
Std Age
Std Expected Age
Gender
(0 = Female, 1 = Male)
Savings
SSA Exist
Loss Aversion
(1)
0.04
(0.02)
-0.14***
(0.02)
-0.43***
(0.05)
-0.01
(0.01)
-0.0004
(0.001)
Loss Aversion
(2)
0.003
(0.03)
-0.10***
(0.03)
-0.38***
(0.06)
-0.01
(0.01)
-0.001
(0.001)
-0.11***
(0.03)
5,863
4,305
Intertemporal Choice
Health
N
3 Oct 2015
ACR Conference 2015
Loss Aversion
(3)
-0.05
(0.10)
-0.21**
(0.07)
-0.53***
(0.11)
-0.01
(0.01)
-0.001
(0.003)
-0.08
(0.07)
-0.04
(0.05)
1,016
* p < 0.05, ** p < 0.01, *** p < 0.001
28
Correlates of Loss Aversion
Std Age
Std Expected Age
Gender
(0 = Female, 1 = Male)
Savings
SSA Exist
Loss Aversion
(1)
0.04
(0.02)
-0.14***
(0.02)
-0.43***
(0.05)
-0.01
(0.01)
-0.0004
(0.001)
Loss Aversion
(2)
0.003
(0.03)
-0.10***
(0.03)
-0.38***
(0.06)
-0.01
(0.01)
-0.001
(0.001)
-0.11***
(0.03)
5,863
4,305
Intertemporal Choice
Health
N
3 Oct 2015
ACR Conference 2015
Loss Aversion
(3)
-0.05
(0.10)
-0.21**
(0.07)
-0.53***
(0.11)
-0.01
(0.01)
-0.001
(0.003)
-0.08
(0.07)
-0.04
(0.05)
1,016
* p < 0.05, ** p < 0.01, *** p < 0.001
29
Loss Aversion Measure, by Gender
20%
18%
N = 5,876
Pct of Participants
16%
14%
12%13%
12%
10%
8%
7%
5%
6%
4%
12%
12%
12%
10%
10%
9%
9%
Male
15%
Males are less loss averse
than females (p < .001)
14%
Female
18%
9%
8%
8%
8%
6%
4%
2%
0%
0
1
2
3
4
5
6
7
8
9
Loss Aversion Measure
3 Oct 2015
ACR Conference 2015
30
Correlates of Loss Aversion
Std Age
Std Expected Age
Gender
(0 = Female, 1 = Male)
Savings
SSA Exist
Loss Aversion
(1)
0.04
(0.02)
-0.14***
(0.02)
-0.43***
(0.05)
-0.01
(0.01)
-0.0004
(0.001)
Loss Aversion
(2)
0.003
(0.03)
-0.10***
(0.03)
-0.38***
(0.06)
-0.01
(0.01)
-0.001
(0.001)
-0.11***
(0.03)
5,863
4,305
Intertemporal Choice
Health
N
3 Oct 2015
ACR Conference 2015
Loss Aversion
(3)
-0.05
(0.10)
-0.21**
(0.07)
-0.53***
(0.11)
-0.01
(0.01)
-0.001
(0.003)
-0.08
(0.07)
-0.04
(0.05)
1,016
Average life expectancy is
significantly correlated with the
loss aversion measure –
individuals who expect to live
longer have lower loss aversion
measures on average
* p < 0.05, ** p < 0.01, *** p < 0.001
31
Correlates of Loss Aversion
Std Age
Std Expected Age
Gender
(0 = Female, 1 = Male)
Savings
SSA Exist
Loss Aversion
(1)
0.04
(0.02)
-0.14***
(0.02)
-0.43***
(0.05)
-0.01
(0.01)
-0.0004
(0.001)
Loss Aversion
(2)
0.003
(0.03)
-0.10***
(0.03)
-0.38***
(0.06)
-0.01
(0.01)
-0.001
(0.001)
-0.11***
(0.03)
5,863
4,305
Intertemporal Choice
Health
N
3 Oct 2015
ACR Conference 2015
Loss Aversion
(3)
-0.05
(0.10)
-0.21**
(0.07)
-0.53***
(0.11)
-0.01
(0.01)
-0.001
(0.003)
-0.08
(0.07)
-0.04
(0.05)
1,016
Intertemporal patience is
significantly correlated with the
loss aversion measure –
participants who are more
patient are also less loss averse
on average
* p < 0.05, ** p < 0.01, *** p < 0.001
32
Correlates of Loss Aversion
Std Age
Std Expected Age
Gender
(0 = Female, 1 = Male)
Savings
SSA Exist
Loss Aversion
(1)
0.04
(0.02)
-0.14***
(0.02)
-0.43***
(0.05)
-0.01
(0.01)
-0.0004
(0.001)
Loss Aversion
(2)
0.003
(0.03)
-0.10***
(0.03)
-0.38***
(0.06)
-0.01
(0.01)
-0.001
(0.001)
-0.11***
(0.03)
5,863
4,305
Intertemporal Choice
Health
N
3 Oct 2015
ACR Conference 2015
Loss Aversion
(3)
-0.05
(0.10)
-0.21**
(0.07)
-0.53***
(0.11)
-0.01
(0.01)
-0.001
(0.003)
-0.08
(0.07)
-0.04
(0.05)
1,016
Self-reported level of savings
and health, as well as
subjective beliefs about Social
Security solvency are not
significantly correlated with the
loss aversion measure
* p < 0.05, ** p < 0.01, *** p < 0.001
33
Implications: Retirement Behaviors
Claiming
Annuities
Bonds
Savings
6.45***
Combined
64.19***
30.55***
41.64***
-0.52***
(1.0)
(10.5)
(9.0)
(1.9)
(0.2)
Yes
Yes
Yes
Yes
Yes
-0.19**
-0.03***
Constant
Demographics
-0.06
-0.68
0.85**
Loss aversion
Manipulation
Controls
3 Oct 2015
(0.04)
(0.46)
(0.39)
(0.08)
(0.01)
Yes
Yes
Yes
Yes
Yes
ACR Conference 2015
** p < 0.05, *** p < 0.01
34
Implications: Retirement Behaviors
Claiming
Annuities
Bonds
Savings
6.45***
Combined
64.19***
30.55***
41.64***
-0.52***
(1.0)
(10.5)
(9.0)
(1.9)
(0.2)
Yes
Yes
Yes
Yes
Yes
-0.19**
-0.03***
Constant
Demographics
-0.06
-0.68
0.85**
The loss aversion
measure is a significant
predictor of several
consumer financial
decision-making
behaviors
Loss aversion
Manipulation
Controls
3 Oct 2015
(0.04)
(0.46)
(0.39)
(0.08)
(0.01)
Yes
Yes
Yes
Yes
Yes
ACR Conference 2015
** p < 0.05, *** p < 0.01
35
Implications: Retirement Behaviors
Claiming
Annuities
Bonds
Savings
6.45***
Combined
64.19***
30.55***
41.64***
-0.52***
(1.0)
(10.5)
(9.0)
(1.9)
(0.2)
Yes
Yes
Yes
Yes
Yes
-0.19**
-0.03***
Constant
Demographics
-0.06
-0.68
0.85**
Loss aversion
Manipulation
Controls
3 Oct 2015
(0.04)
(0.46)
(0.39)
(0.08)
(0.01)
Yes
Yes
Yes
Yes
Yes
ACR Conference 2015
The loss aversion
measure is a significant
predictor of several
consumer financial
decision-making
behaviors
More loss averse
individuals prefer bonds
(over equities), save
less, and show overall
myopia in retirementrelated decisions
** p < 0.05, *** p < 0.01
36
Implications: Retirement Behaviors
Claiming
Annuities
Bonds
Savings
6.45***
Combined
64.19***
30.55***
41.64***
-0.52***
(1.0)
(10.5)
(9.0)
(1.9)
(0.2)
Yes
Yes
Yes
Yes
Yes
-0.19**
-0.03***
Constant
Demographics
-0.06
-0.68
0.85**
Loss aversion
Manipulation
Controls
3 Oct 2015
(0.04)
(0.46)
(0.39)
(0.08)
(0.01)
Yes
Yes
Yes
Yes
Yes
ACR Conference 2015
The loss aversion
measure is a significant
predictor of several
consumer financial
decision-making
behaviors
More loss averse
individuals prefer bonds
(over equities), save
less, and show overall
myopia in retirementrelated decisions
** p < 0.05, *** p < 0.01
37
Implications: Retirement Behaviors
Claiming
Annuities
Bonds
Savings
6.45***
Combined
64.19***
30.55***
41.64***
-0.52***
(1.0)
(10.5)
(9.0)
(1.9)
(0.2)
Yes
Yes
Yes
Yes
Yes
-0.19**
-0.03***
Constant
Demographics
-0.06
-0.68
0.85**
Loss aversion
Manipulation
Controls
3 Oct 2015
(0.04)
(0.46)
(0.39)
(0.08)
(0.01)
Yes
Yes
Yes
Yes
Yes
ACR Conference 2015
The loss aversion
measure is a significant
predictor of several
consumer financial
decision-making
behaviors
More loss averse
individuals prefer bonds
(over equities), save
less, and show overall
myopia in retirementrelated decisions
** p < 0.05, *** p < 0.01
38
Takeaways
Formulated an easy-to-implement,
model-free loss aversion measure
3 Oct 2015
ACR Conference 2015
39
Takeaways
Formulated an easy-to-implement,
model-free loss aversion measure
Demonstrated the relationship
between demographic/psychographic
variables and our loss aversion
measure
3 Oct 2015
ACR Conference 2015
40
Takeaways
Formulated an easy-to-implement,
model-free loss aversion measure
Demonstrated the relationship
between demographic/psychographic
variables and our loss aversion
measure
Our loss aversion measure is a
significant predictor of important
financial decision-making behaviors
3 Oct 2015
ACR Conference 2015
41
THANK YOU!!
Elizabeth C. Webb
[email protected]