Dynamic Choice, Independence, and Emotions

1
Dynamic Choice, Independence, and Emotions

• Astrid Hopfensitz and Frans van Winden
• CISA, University of Geneva, Switzerland
• CREED, University of Amsterdam, The Netherlands
• FUR XII 2006 at LUISS in Rome

2
Motivation

• Dynamic choice
• little attention in experimental economics of
  individual choice
• Important issue independence of global risk
  (GR)
• chance to lose everything whatever one
  chooses
• theory (key axiom EUT)
• practice (cf. fear of terror)

3
Global Risk example

z endowment 1 - p decision risk x
investment x ? 0, z q global risk
4

• Findings so far
• common ratio effect gt more risk seeking if
  reduction
• isolation effect gt people ignore common
  components
• pro Kahneman Tversky (1979) Cubitt et al.
  (1998)
• con Bosman van Winden (2005)
• timing independence gt no effect of resolution
  stage
• pro ??? Cubitt Sugden 2001 accumulator
  gambles
• con Cubitt et al. (1998)
• This experimental study
• (1) different timing of GR vs. no GR, (2) role
  of affect

5
Cognition, Emotion, and Behavior
Anticipated outcomes (anticipated emotions)
Cognitive evaluation
Subjective probabilities
Behavior
Feelings (experienced emotions)
Other factors e.g. vividness, immediacy,
background moods
Sensory Cortex
high road
low road
AMYGDALA
(quick and dirty)
Emotional Responses
Emotional Stimulus
based on Loewenstein et. al (2001) and LeDoux
(1996)
6
Behavior and Emotions

• anticipated emotions economists
• regret theory Loomes Sugden (EJ 1982) Bell
  (OR 1982)
• disapp theory Bell (OR 1985)Loomes Sugden
  (RES 1986)
• anxiety theory Wu (TD 1999) Caplin Leahy (QJE
  2001)
• experienced emotions psychologists
• mood and risk taking Isen (JCP 2001)
• anxiety and risk taking Raghunathan Pham
  (OBHDP 1999)
• anger and risk taking Leith Baumeister (JPSP
  1996) Lerner Keltner (JPSP 2001)

7
Experimental Design Task
Allocation of 15 Euro to two projects
project A
project B

1 - p 1/2
p 1/2
0
Safe no gain, no loss
Risky with p 1/2 2.5 x investment with 1- p
1/2 nothing
8
Experimental Design Treatments
z endowment 1 - p decision risk x
investment x ? 0, z q global risk
9
Experimental Design Emotion Measures

• self reports
• emotion traits
• subjects indicate feelings in general (anxiety)
• experienced emotions
• subjects indicate current feelings
• anticipated emotions
• subjects indicate expected feelings
• why self reports ?
• easy to use
• very well validated
• distinguish emotions (e.g. anxiety vs. anger)
• Robinson Clore (2002) most common and
  potentially the best way to measure

10
Experimental Design Timing Emotion Measures
outcome is announced resolution of any risk
taking of decision
EMOTION
anticipated emotions
REGRET-A (B) REJOICE-A (B)
REGRET-E DISSAP-E
time
ANXIETY-1
REGRET-X DISSAP-X
ANXIETY-2
experienced emotions
HOPE-1 IRRITATE-1
ANXIETY-trait
11
Experiment

• treatments
• Baseline
• GR-Pre
• GR-Inter
• GR-Post
• CREED laboratory at the University of Amsterdam
• 192 Dutch participants from various fields
• (average age 22 years)
• experiment lasted less than 60 minutes
• average earnings 17 Euro

12
Isolation Effect Experiment Isolation effect
assumes existence of an editing phase in decision
making where (transparent) common components of
alternatives are cancelled.
z endowment 1 - p decision risk x
investment x ? 0 z q global risk
13
Isolation Effect - Behavioral Results
fraction of working money invested
Mann-Whitney, p 0.011 Kolm.-Smirnov,
p0.012 gt isolation effect rejected
14
Isolation Effect - Emotional Results
Censored tobit regression of investment on
emotions

• Hope positive effect on investment
• Irritation negative effect on investment
• Rel. rejoicing positive effect on investment
• Anxiety no effect on investment

15
Timing Independence Experiment

• Timing independence - an agent, if required to
  precommit to an action to be taken conditional on
  a prior act of nature, precommits to the action
  which would be chosen if the moment of choice was
  delayed until after that act of nature.

z endowment 1 - p decision risk x
investment x ? 0 z q global risk
16
Timing Independence - Behavioral Results
fraction of working money invested

• Mann-Whitney, p 0.247 Kolm.-Smirnov, p
  0.312
• timing independence not rejected
• also no difference investment GR-Pre and
  Baseline (separability)
• Mann-Whitney, p 0.117 Kolm.-Smirnov, p
  0.454

17
Timing Independence - Emotional Results

• Mood maintenance hypothesis not supported
• happy survivers' of global risk do not invest
  more
• Tobit regression model not significant for GR-Pre
• prob gt ?2 0.152 coeff. rel. rejoice
  0.08, p 0.024
• gt emotionally, treatments differ
• but in some respects GR-Pre GR-Inter more alike
• greater arousal (anxiety, hope)
• may have positively influenced investment
• Leith Baumeister 1996

18
Adding GR-Post

• timing of resolution decision risk and global
  risk reversed
• Timing independence principle extended
• Furthermore, it should not matter whether the
  act of nature happens just before (GR-Inter) or
  just after (GR-Post) the resolution of the
  decision risk.

19
Adding GR-Post - Behavioral Results
Mann-Whitney, p 0.117 Kolm.-Smirnov, p
0.454
Mann-Whitney, p 0.495 Kolm.-Smirnov, p
0.676
Mann-Whitney, p 0.011 Kolm.-Smirnov, p
0.012
Mann-Whitney, p 0.653 Kolm.-Smirnov, p
0.949
20
Adding GR-Post - Results

• Isolation effect now supported
• in contrast with GR-Inter and against Bosman
  van Winden (2005)
• GR-Post vs. Baseline Mann-Whitney, p 0.062
  Kolm.-Smirnov, p 0.103
• Timing independence supported against Cubitt et
  al. (1998)
• GR-Pre/Inter/Post Mann-Whitney, p gt 0.247
  Kolm.-Smirnov, p gt 0.312
• Affect model not significant for GR-Post (like
  GR-Pre)
• prob gt ?2 0.269 coeff. Irritate-1 0.130, p
  0.078
• positive correlation irritation/investment
  (Spearman 0.259, p0.099)
• arousal (anxiety, hope) more like Baseline
• gt emotionally, treatments differ

21
Summary

• Isolation Effect violated in GR-Inter but not in
  GR-Post
• Timing Independence not rejected
• More full investment in GR
• Linear affect model explains investment in
  Baseline and GR-Inter
• with similar coefficients for anticipated
  rejoicing (),
• experienced hope (), and experienced
  irritation (-).
• Affect functions differently in GR-Pre and
  GR-Post
• only anticipated rejoicing (pos.) correlated
  with investment in both
• in GR-Post also experienced irritation, but
  with a positive sign.

22
Summary Regret/Rejoicing

• Is anticipated regret/rejoicing equally
  important in treatments?

Baseline GR-Pre GR-Inter GR-Post
Regret motive for project A -0.554 (p0.021) -0.432 (p0.008) -0.528 (p0.017) -0.249 (p0.241)
Rejoicing motive for project B 0.134 (p0.609) 0.440 (p0.006) 0.383 (p0.096) 0.530 (p0.008)
Spearman correlation coefficients of investment
with emotions.
Estimated and experienced regret correlated in
all treatm. Spearman, 0.5 p lt 0.07