Prospect Theory and Choice Behaviour Strategies

1
Prospect Theory and Choice Behaviour Strategies
Review and Synthesis of Findings from
Social sciences
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
2
Content

• The Choice Behaviour Process
• The Function of Choice Behaviour
• A functional comparison of Utility and Prospect
  Theory with choice observations in Social
  Sciences
• Conclusions

2009 October 8 Evert Jan van de Kaa
TU Delft TLO
3
The Choice BehaviourProcess
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
4
Definition of choice behaviour
Human choice behaviour is conceived here as a
mental process that transforms mental
perceptions of several feasible courses of
action into a choice
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
5
Choice behaviour process
The mental process that accounts for human choice
behaviour consists of reasoning There exists a
collective view among social scientists that
reasoning is a dual-mode process
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
6
Choice behaviour process
Cognitive psychology (e.g. Stanovich
and West 2000, Kahneman 2002) System1 or
intuition runs unconscious, parallel, fast,
automatic, effortless and System 2 or analytic
intelligence runs
conscious, sequential, slow, effortful
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
7
Choice behaviour process
Social psychology (e.g. Zajonc 1980, Lieberman
et al. 2002, Dijksterhuis et al. 2008)
Feelings, hot/reflexive cognition, unconscious
thought and Analysis, cold/reflective
cognition, conscious thought
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
8
Choice behaviour process
Neuroscience (e.g. Bechara et al. 1997, Damasio)
Feelings and Conscious elaboration of
pertinent facts
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
9
Comparison of System 1 and 2
Efficiency Data processing capacity
(Dijksterhuis et al. 2004) Unconscious thought
app. 11.2 Mbps against Conscious
thought app. 60 bps
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
10
Comparison of System 1 and 2
Effectivity Correctness in a judgment task
(Mintzberg 1989) Intuitive cognition 35
correct, range of wrong assessments limited
against Analytical cognition 80
correct, range in wrong assessments twice as
large See also e.g. Hammond et al 1987,
Dijksterhuis et al. 2008
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
11
Comparison of System 1 and 2
Spread in actual choice behaviour (e.g. Zajonc
1980, Lieberman et al. 2002) Feelings,
intuition the prime in any choice process May
be followed by Analysis, conscious deliberation
as second thought
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
12
Comparison of System 1 and 2
Impact on final choice (Festinger 1975, Beach
1990, Montgomery 1989, Kahneman 2002,
Dijksterhuis 2007) Feelings, intuition,
unconscious thought default choice,
mostly decisive Against Analysis, conscious
thought occasionally overruling,
often affirmative
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
13
Process models of choice behaviour
Human choice behaviour is predominantly
covert Choice models that claim to offer a
factual picture of such a process cannot be
falsified
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
14
The Function of Choice Behaviour
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
15
The overall function of choice behaviour
The overall-function of choice behaviour is
defined as to choose one possible course of
action from a set of alternatives that, in that
particular context, meets certain of the
individuals concurrent needs
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
16
A functional view of choice behaviour
Functional decomposition yields four functions
and their interrelationships
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
17
A functional model of choice behaviour
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
18
A generic model of choice behaviour?
This model offers a generic framework to which
any set of assumptions that is able to put the
overall choice behaviour function into operation
can be attributed
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
19
Framing
Definition of relevant contextual factors Main
sub-functions Definition of The reference
state Considered alternatives with their
expected outcomes Preferences related to
subjects concurrent needs Aspiration level for
needs gratification
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
20
Framing assumptions of Utility Theory
(e.g. Katz and Rosen 1999) Reference state
irrelevant Alternatives complete range of
expected states of assets Subjects
needs non-satiable temporally stable,
context-independent, complete preference
order of all goods Aspiration level maximal
ordinal utility for gratification within
budget constraints
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
21
Framing assumptions of Prospect Theory
(Kahneman and Tversky 1979, Tversky and
Kahneman 1991, 1992) Reference state expected
no change state of assets Alternatives
context-dependent range of expected changes
(gains and losses) in assets relative to
reference state Subjects needs non-satiable
preference order dependent on the context and
the way in which alternatives are
perceived Aspiration level maximal ordinal
utility for gratification within budget
constraints
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
22
A Functional comparison ofUtility and Prospect
Theory with choice observations from Social
Sciences
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
23
Comparison of framing assumptions 1
Reference state alternatives Utility Theory
reference irrelevant, alternatives complete
range of expected states of assets Prospect
Theory reference expected no change state of
assets, alternatives context-dependent range of
expected changes (gains and losses) in assets
relative to reference Test 1 Ask people to make
two successive choices from identical choice sets
and alternatives with the same outcomes, in which
only the reference state differs (Tversky
Kahneman 1991, travel-related job choice setting)

2009 October 8 Evert Jan van de Kaa
TU Delft TLO
24
Comparison of framing assumptions 1
Reference state alternatives Utility Theory
reference irrelevant, alternatives complete
range of expected states of assets Prospect
Theory reference expected no change state of
assets, alternatives context-dependent range of
expected changes (gains and losses) in assets
relative to reference Result about 1/3 of the
panel shifted their preference Conclusion in
paper UT is violated, PT confirmed Should be
significant part of population violated UT and
confirmed PT majority might not have violated UT
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
25
Comparison of framing assumptions 1
Reference state alternatives Utility Theory
reference irrelevant, alternatives complete
range of expected states of assets Prospect
Theory reference expected no change state of
assets, alternatives context-dependent range of
expected changes (gains and losses) in assets
relative to reference Test 2 Allow people to
swap an object that is at their disposal for a
similar object of about the same value (thus only
reference state makes a difference) (Knetsch
1989, and over a hundred more endowment studies)

2009 October 8 Evert Jan van de Kaa
TU Delft TLO
26
Comparison of framing assumptions 1
Reference state alternatives Utility Theory
reference irrelevant, alternatives complete
range of expected states of assets Prospect
Theory reference expected no change state of
assets, alternatives context-dependent range of
expected changes (gains and losses) in assets
relative to reference Result resistance to swap
an object once possessed Conclusion a
significant part of the survey populations
violated UT and confirmed PT many might not have
violated UT
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
27
Comparison of framing assumptions 2
Subjects needs Utility Theory temporally
stable, context-independent, complete preference
order of all goods Prospect Theory preference
order dependent on the context and the way in
which alternatives are perceived Test 3 let
people make two successive choices, from
identical choice sets containing alternatives
with the same outcomes and with the same
reference state, only the wording of the
alternatives is changed from positive to negative
(e.g. Levin et al. 1998 for review of 85 framing
experiments)
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
28
Comparison of framing assumptions 2
Subjects needs Utility Theory temporally
stable, context-independent, complete preference
order of all goods Prospect Theory preference
order dependent on the context and the way in
which alternatives are perceived Result by far
most studies revealed a statistical significant
preference shift Conclusion a significant part
of the study population violated this UT
assumption and confirmed that of PT many might
have adhered to the UT assumption
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
29
Comparison of judgment assumptions 1
Valuation of expected probabilities Utility
Theory unweighed, but people may attach a
positive or negative value to probabilistic
alternatives Prospect Theory Variable weight
factor, dependent on probability Test 4 Let
people make some choices from choice sets in
which all alternatives have the same expected
utility but where probability distributions of
gains and losses differ between choices (e.g.
Kahneman Tversky 1979) Result Many studies
show that a significant part of the survey
population is risk averse in a gain situation and
risk prone to avoid an almost certain loss
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
30
Comparison of judgment assumptions 2
Valuation of attribute characteristics Utility
Theory independent of sign of changes in state
of assets Prospect Theory dependent on sign of
changes loss aversion Test 5 selling of an
object that is at subjects disposal or buying
one that is not (thus only reference state makes
a difference) (Knetsch 1989, and over a hundred
more endowment studies) Results Selling
prices were on average twice the buying prices,
but large interpersonal differences
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
31
Comparison of judgment assumptions 2
Valuation of attribute characteristics Utility
Theory independent of sign of changes in state
of assets Prospect Theory dependent on sign of
changes loss aversion Conclusion presumably a
majority of the study populations were loss
aversive and violated this UT assumption others
might have adhered to the UT assumption
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
32
Comparison of judgment assumptions 3
Size-dependent valuation of attribute
characteristics Utility Theory diminishing
marginal utility for expected increases in
assets Prospect Theory diminishing sensitivity
for increases in gains and increases in
losses Test 6 for attribute levels that are
less attractive than the reference level an
econometric model estimate yields either a
significant concave (UT) or convex (PT) shape of
the utility/value function (Herne 1998 and just
a few more studies)
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
33
Comparison of judgment assumptions 3
Valuation of attribute characteristics Utility
Theory diminishing marginal utility for expected
increases in assets Prospect Theory diminishing
sensitivity for increases in gains and increases
in losses Result These few studies found a
convex value function in the loss domain and thus
confirm this PT assumption
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
34
Comparison of evaluation-and-choice assumptions
Comparison of multi-attribute alternatives
choice criterion Utility Theory
Alternative-wise, based on a compensatory
compounded ordinal utility measure highest
ordinal utility Prospect Theory
Alternative-wise, based on a single compensatory
compounded decision value highest overall
decision value Test the assumptions do not
differ significantly. Many studies in decision
theory show convincingly that minorities of the
population use non-compensatory rules and/or
satisficing choice criteria (e.g. Payne et al.
1993) Conclusion Both the UT and PT assumption
are too restrictive
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
35
Choice behaviour strategy
A choice behaviour strategy is a personal set
of procedures for framing (e.g. of reference
state and alternatives), judgment, (e.g. loss
aversive or loss neutral attribute valuation) and
evaluation-and-choice (e.g. compensatory or
non-compensatory evaluation, satisficing or
maximizing choice criterion)
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
36
Choice behaviour strategy
The comparisons of the assumptions about
framing, judgment and evaluation-and-choice with
observations in social sciences revealed for each
function non-random interpersonal heterogeneity
in the application of these rules
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
37
Choice behaviour strategy
To avoid redundancy the heterogeneity in
function completion might better be attributed to
the execution of different consistent
idiosyncratic choice behaviour strategies than to
the individual functional assumptions
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
38
Choice behaviour strategy
Heterogeneity in choice behaviour strategies
Utility Theory Each individual chooses the
alternative with the highest (compensatory
compounded) overall ordinal utility Extended
Prospect Theory People exhibit interpersonal
heterogeneity in intrapersonal consistent choice
behaviour
2009 October 8 Evert Jan van de Kaa
TU Delft TLO
39
Conclusion
Comparison of the assumptions of Utility and
Prospect Theory against choice observations from
social sciences shows that the latter assumptions
offer a better description of many if not most
observations It also shows that extension of
Prospect Theory with the assumption that people
exhibit interpersonal heterogeneity in
intrapersonal consistent choice behaviour
strategies may offer a further improvement of its
descriptive ability
2009 October 8 Evert Jan van de Kaa
TU Delft TLO