1209848546 thesis howatt · alternative to the normative economic model of expected utility theory...
TRANSCRIPT
Factors of Religiosity and Loss Aversion
by
Brian Howatt
A Thesis Presented in Partial Fulfillment of the Requirements for the Degree
Master of Science
Approved April 2017 by the Graduate Supervisory Committee:
Elias Robles-Sotelo, Chair
Perla Vargas Tess Neal
ARIZONA STATE UNIVERSITY
May 2017
i
ABSTRACT
Loss aversion manifests as a decision bias in which avoiding losses is preferred
over acquiring rewards and can drastically alter an individual’s decision-making by
overweighting potential losses relative to gains of equal magnitude. Consequently,
individuals may require greater positive compensation to offset potential losses, exhibit
contradictory choice preferences, or even avoid the decision entirely; and this behavior
may be ascribed to an over-reliance on automatic, unconscious (intuitive) judgments
rather than initiating analytic reasoning more capable of objectively evaluating outcomes.
Religion (specifically Christianity) is the topic of focus, as preliminary evidence
suggests an individual’s intuitive inclinations positively correlate with and predict
religious beliefs. Moreover, self-reported religious beliefs significantly differed as a
function of inducing either intuitive or reflective mindsets. Therefore, the purpose of this
experiment was to test the hypothesis that religious participants will display significantly
greater levels of loss aversion than nonreligious participants.
This hypothesis extends from a previous study relating large-scale cultural and
religious differences with loss aversion. While their results revealed religious orthodoxy
strongly influenced loss aversion, the parameters elicited may be less stable as only two
lottery questions were asked and religion was determined by cultural demographics. This
study used the same design, but with a total of ten lotteries and a more detailed
investigation into individual religious factors.
While loss aversion coefficients replicated the overall behavioral effect (Median θ
= 2.6), independent sample, Mann-Whitney U tests did not yield any significant
ii
differences between Christian and Nonreligious participants (p > 0.05); nor did any of the
religious factors examined account for a significant amount of variability.
This study attempted to add to current knowledge by further conflating the
relationship between religiosity and adaptive decision strategies susceptible to errant and
inconsistent behavior. While the hypotheses were unsupported, a null finding is still
important, and future research re-testing this association or introducing causational
designs may prove more fruitful in understanding these complex relationships.
iii
ACKNOWLEDGMENTS
The author would like to express gratitude to his thesis committee chair and
members, Drs. Elias Robles, Perla Vargas and Tess Neal for their invaluable assistance
and guidance during the formulation, carrying-out and completion of this project, as well
as Psychology faculty and students at Arizona State University for providing helpful
comments and questions.
iv
TABLE OF CONTENTS
Page LIST OF TABLES…………………………….………………….....……………….….vii LIST OF FIGURES…………………………………………………..…...…………….viii LIST OF EQUATIONS…………………………………………………………………..ix DEFINITIONS…………………………………………………………………………….x CHAPTER 1 INTRODUCTION…………………………………………………..……………. 1 Risk Preferences………………………….…………………..…………....1 Religion………………………..………………..…………..………..........2 Present Study………………………..……………………….……………3 2 RISK PREFERENCES…………………………………………….……………... 4 Prospect Theory………………………..………………………………….4 Alternative to Expected Utility Theory………………..…………..4 Framing Effects……………………………….…..……………….7 Forecasting…………………………...………………….………...……....8 Delay Discounting………………………………………………………...9 Behavioral Parameterization……………………..…….………….……..11 Risk Aversion………………..………………………....….……..12
v
CHAPTER Page Loss Aversion………………………………..…....……………..13 Summary…………………………………………………………………14 3 DUAL-SYSTEM FRAMEWORK OF COGNITION………………...………… 15
Dual-System Models…………………...………………………………...15 Sloman (1996)................................................................................15 Kahneman (2003)...........................................................................16
Heuristics……………………………………...……...….18 Critical Review: Evans (2008)...................................................................22 Summary…………………………………………………………………23 4 CRITICISMS…………………………………………...……………………….. 24 Heuristics……………………...…….………..………………………….24 Rebuttals…………………….………………….…….………….26 Semantic Inference.....................................................................................27 Prospect Theory……………………………………...…..………..……..29 Rebuttal………………………………………………………..…30 Loss Aversion………………………………..…………….…………….31 Reconciliation…………………………..…………..……………….…...32 5 RELIGION………………………………………………………………………. 33
vi
CHAPTER Page Religiosity and Risk Preferences…………………………..…………….33
Delay Discounting……………………………………………….34
Risk Aversion……………………………………………….……35 Gambling………………………….…………………………………..….36 6 PRESENT STUDY………………………………..………………………..…… 38 Methods…………………………………………………………………..38 Participants……………...………………………………….…….38 Design/Procedure…………………...……….……………….…..39 Results………………………………………………..…………………..41 Data Analysis…………………………………………………….41 Religious Identity………………………………………………...42 Religious Statements Responses…………………………………46 7 DISCUSSION…………………………………………………………………… 47 REFERENCES…………………………………………………………………………. 50 APPENDIX
A DATA COLLECTION AND REDUCTION……………………………… 60 B SUPPLEMENTARY DATA RESULTS………………………………….. 62 C IRB APPROVAL………………………………………...………………... 66
vii
LIST OF TABLES Table Page
1. Median Gain/Loss Ratios by Religious Identity………………………………... 43 2. Correlation Table of Gain/Loss Ratio and Religious Factors…………………... 45
viii
LIST OF FIGURES Figure Page 1. Prospect Theory Utility Function………………………………………...…………… 5 2. Delay Discount Functions……………………………………………………………. 11 3. Dual-Systems Framework of Cognition…………………………...…………….…... 17 4. Lottery Table Design………………….…………...………………………………… 39 5. Loss Aversion Coefficients by Religious Identity………………….………………... 44 6. Distribution of Gain/Loss Ratios by Religiosity……………………………………... 46
ix
LIST OF EQUATIONS
Equation Page 1. Relative Risk Premium………………………………………………………………. 13
x
DEFINITIONS
Expected Value (EV) – the average payout of a gamble over repeated trials, but can be
applied to single-event likelihoods. This is calculated as summing the products of every
outcome’s probability and quantitative value. This value is then multiplied by the number
of times the gamble is offered. If a coin is flipped 10 times to win either $10 or $0, the
EV is ((.5 * $10) + (.5 * $0)) * (10) = $50. A single coin-flip would therefore yield an EV
of $5.
Certainty Equivalent (CE) – a guaranteed payout that acts as a deterrent to a
probabilistic outcome. Instead of accepting the above coin-flip gamble, an individual may
prefer a certain $6 reward instead. The expected value of a CE is the given amount
multiplied by the number of times it is offered: ((1.0 * $6) * (n)) = $6 * n. The expected
values of the gamble and CE are then compared to determine which choice is statistically
optimal.
Indifference Point – The point at which an individual switches their preference between
choices; the options are of roughly equal value. If an individual prefers the guaranteed
outcome when the CE is $6, but prefers the gamble when the CE is $4, the indifference
point is about $5.
1
CHAPTER 1
INTRODUCTION RISK PREFERENCES
Individuals often make decisions involving degrees of uncertainty in the
outcomes, the consequences of which are unbeknownst prior to their manifestation.
While much of what an individual risks in daily decisions is tolerable (not bringing an
umbrella, transportation, indulging in a scratch-off lotto ticket), empirical evidence has
revealed the aversive qualities of risk-taking in pursuit of large rewards when a
guaranteed, lesser reward may appeal as an alternative. Conversely, individuals may
avoid guaranteed outcomes when presented as losses/punishments (Kahneman &
Tversky, 1979, 1984; Tversky & Kahneman, 1981). Moreover, decision-makers may
show a general reluctance to engage in 50-50 mixed-frame gambles unless the expected
gains are roughly twice as much as potential losses (Tversky & Kahneman, 1991;
Kahneman, 2003; Tom, Fox, Trepel, & Poldrack, 2007). These behavioral phenomena
intending to reduce uncertainty and negative outcomes, referred to as ‘risk aversion’ and
‘loss aversion’ respectively (Kahneman & Tversky, 1984), are illustrated by a subjective
overweighting of the dissatisfaction derived from forgoing guaranteed gains, or
experiencing losses relative to gains of equivalent magnitudes.
An important distinction must be made here between two different types of
uncertainty in judgments and choices - risk and ambiguity. Although these terms are
often linguistically interchangeable, they diverge semantically. Risky outcomes are those
with clearly-defined, measurable probabilities (e.g., a 50% chance of success: coin-flip)
2
while ambiguity refers to outcomes with vaguely defined or even undefined likelihoods
(e.g., “some” chance of success) that requires individuals to make subjective probability
estimates (Knight, 1921). Because behavioral decision research primarily concerns the
analysis of judgments and choices under risk, one must be careful in how results are
interpreted within more ecological, ambiguous contexts. Nonetheless, models of
decision-making strive to generalize observed behavior and have effectively
demonstrated how these choice dynamics can fundamentally influence social and
economic processes, such as marketing and consumption preferences, negotiations, and
politics (Novemsky & Kahneman, 2005; Carnevale, 2008; McDermott, 2004).
Additionally, some have established religion as a framework for guiding such
dispositions, perceptions and choices in life (Hommel & Colzato, 2010; Hui, Chan, Lau,
Cheung, & Mok, 2014).
RELIGION
Religion has played an integral role in human existence and development over
thousands of years and holds an almost ubiquitously entrenched presence across human
civilizations, with approximately 80% of people worldwide (Pew Forum Research, 2012)
and 78% of U.S. residents (Pew Forum Research, 2014) identifying with a religious
group. While some researchers have argued for the emergence and evolution of religion
within a cultural context (Henrich, 2009; Gervais, Willard, Norenzayan & Henrich, 2011;
Norenzayan et al., 2016), others have asserted a purely biological origin stemming from
the evolutionary adaptations of cognitive machinery (Bloom, 2007; Boyer & Bergstrom,
3
2008), such as executive functioning (Friedman et al., 2008), that may motivate
intuitively causational ascriptions (Baumard & Boyer, 2013) to terrestrial events
(Kelemen, 2004; Kelemen & Rosset, 2009), assuaging the anxiety of environmental
uncertainty (Preston & Epley, 2005, 2009). Regardless of religion’s exact genesis, it
persists as a prevalent belief structure shaping one’s behavior through, but not limited to,
increased self-control and regulation, and prescribing intrinsically oriented goals that
temper hedonistic pursuits (McCullough & Willoughby, 2009).
PRESENT STUDY
Research aimed at studying religiosity has revealed significant relationships
between religious behavior - comfort, importance, prayer frequency and service
attendance, and increased risk aversion (Miller, 2000; Noussair, Trautmann, Kuilen, &
Vellekoop, 2013). Moreover, Rieger, Wang and Hens (2011) found preliminary evidence
suggesting Christian orthodoxy exerts a significantly stronger influence on loss aversive
behavior. However, the scope of Rieger and colleagues’ study was a large-scale, cross-
cultural and economic analysis of risk preferences, and their results were not intended to
be used on an individual level. Therefore, the purpose of this research project is a
modified replication and extension of their work, but with individual metrics of
religiosity.
4
RISK PREFERENCES
PROSPECT THEORY
Prospect Theory (PT; Kahneman & Tversky, 1979) was introduced as an
alternative to the normative economic model of Expected Utility Theory (Bernoulli,
1738/1954), which had largely dominated the study of decision-making under risk.
Expected Utility Theory (EUT) posited that risk aversion resulted from subjectivity when
assessing [dis]utility (the [dis]satisfaction derived from the acquisition or loss of valued
stimuli) of choices, and proposed a set of axioms aiding in the objective analysis of
probabilistic judgments and choices. Satisfying these axioms will then, on average, lead
to statistically optimal outcomes and utility maximization, and mathematicians John Von
Neumann and Oskar Morgenstern (1944) later rigorously proved a series of conditions
under which these axioms hold.
The three primary conditions of Von Neumann and Morgenstern’s theorem, as
well as those challenged by PT are: completeness, transitivity, and independence.
Completeness assumes economic agents have well-defined preferences; Transitivity
assumes these preferences will remain stable across a series of choices (i.e., if A > B, and
B > C, then A > C); and Independence assumes this order of preferences will remain
stable, even when the first two are presented independently of the third (i.e., A>B>C and
A>B; although this axiom is controversial because its application can actually sometimes
lead to B>A - Allais, 1953). The seminal paper of Prospect Theory, as well as their work
surrounding it (Tversky, 1969; Tversky & Kahneman, 1974, 1981), sought to
demonstrate inconsistencies in these axioms and offer explanations for them. By
5
providing participants with probabilistic scenarios and gambles, they were able to elicit
choices violating statistical intuitions and how risk-averse behavior could lead to
preference reversals.
alternative to EUT
The main conclusions of PT supplanted EUT as the predominant, descriptive
explanation for decision-making under risk, asserting the disutility of losses was a
significantly stronger force on behavior than originally accredited (Kahneman &
Tversky, 1979). Additionally, utility was no longer derived from purely an agent’s
overall status (i.e., total wealth) the outcome of a gamble contributed to; rather, it was
based on localized reference points (i.e., how much has been currently gained or lost) and
[dis]utility was judged by deviations from those baselines. This created an S-shaped
function (Figure 1) where an individual’s subjective value of a loss is greater than a gain
of equivalent magnitude.
Figure 1. - A hypothetical value function proposed by Kahneman & Tversky, 1979
6
This judgment of potential losses exerting a more potent disutility was theorized
to lead individuals to avoid engaging in statistically optimal behavior involving
uncertainty, or decision outcomes with similarly valued rewards and losses, ‘risk’ and
‘loss’ aversion, respectively. Risk aversion can be characterized by accepting a lower,
guaranteed value over the potential to gain a probabilistic choice with a greater expected
value (the average payout of a decision over repeated trials, but is also used to describe
single-event values). And contextual conditions such as the magnitude of outcomes, the
type of gamble played, and the likelihood of an outcome may all play an important role in
judgments and choice dynamics by altering one’s propensity to accept guaranteed
amounts or take risks.
First, the magnitude of an outcome may significantly affect an individual’s
preferences by decreasing risk behavior as the stakes rise or increasing risk-taking in
minute stakes (Harinck, Van Dijk, Van Beest, & Mersmann, 2007). Next, the type of
gamble, or frame in which the decision is presented, significantly impacts risk
preferences, illustrated in the now classic ‘Asian Disease Problem’ presented below
(Tversky & Kahneman, 1981; percentage of respondents favoring the choice in
parentheses):
Problem 1 (N = 152): Imagine that the U.S. is preparing for the outbreak of an unusual Asian disease, which is expected to kill 600 people. Two alternative programs to combat the disease have been proposed. Assume that the exact scientific estimates of the consequences of the programs are as follows:
If Program A is adopted, 200 people will be saved. (72%)
OR
7
If Program B is adopted, there is a one-third probability that 600 people will be saved and a two-thirds probability that no people will be saved. (28%)
Now consider another problem in which the same cover story is followed by a different
description of the prospects associated with the two programs:
Problem 2 (N = 155):
If Program C is adopted, 400 people will die. (22%)
OR
If Program D is adopted, there is a one-third probability that nobody will dies and a two-thirds probability that 600 people will die. (78%)
framing effects
This research by Tversky and Kahneman has demonstrated how different frames
can yield preference reversals - gain frames dampen risk-taking, while loss frames tend to
promote risk-taking - and they suggest that gains and losses are processed differently,
according to the aforementioned reference points. Despite the statistical equivalency of
the decision scenarios, the gain condition is focused on saving lives and therefore
processed as a reward by increasing the reference point; the loss condition is focused on
losing lives, evoking a much more emotional process that seeks to maintain the current
reference point.
An interesting wrinkle in framing effects arises though when the probabilities of
outcomes are manipulated. For gain frames, the propensity to take risks increases for
lower likelihood outcomes, while higher likelihood outcomes decrease risk. This effect
8
tends to reverse for loss frames - greater risk-taking in high likelihood outcomes and
reduced risk-taking in low likelihood outcomes. This pattern is referred to as the
“fourfold pattern of risk attitudes” (Tversky & Kahneman, 1992, p. 308) and is of special
interest to the decision research community, in that risk attitudes are flipped simply by
the types of outcomes that may occur (see Kuhberger, Schulte-Mecklenbeck, & Perner,
1999, for a review confirming the efficacy of framing effects).
This cognitive distinction between rewards and punishments often leads to “losses
looming larger than gains” (Kahneman & Tversky, 1979), where the negative feelings
anticipated from potential losses are much greater than the positive feelings aroused by a
potential reward. It is crucial though to make point of the wording and stress that “losses
looming larger than gains” does not necessarily equate to “losses are actually stronger
than gains”. Accordingly, Prospect Theory postulates asymmetric psychophysical
properties between ‘Decision’ Utility (anticipated utility - Kahneman & Tversky, 1984)
and ‘Experienced’ Utility, in that how an agent feels a decision will affect him/her may
not match the actual hedonic experience.
FORECASTING
When the precise nature of decision outcomes is unknown, an individual’s choice
is often predicated upon the outcome anticipated to yield the greatest affectual return
(Mellers, Schwartz, & Ritov, 1999; Mellers & McGraw, 2001). Loss aversion can
interfere with this process if the asymmetry between gains and losses in decision utility is
not appropriately reflected in one’s experienced utility. If, however, these utilities of
9
gains and losses are inherently different, the justification to forgo a $15 reward in lieu of
possibly incurring a $10 loss is rationally admissible (Tversky & Kahneman, 1991).
Previous research has suggested this may not be the case though, as participants tended to
overestimate the duration and intensity of their reactions to negative scenarios, such as
the dissolution of relationships, or denial of tenure and employment (Gilbert, Pinel,
Wilson, Blumberg, & Wheatley, 1998; Gilbert, Morewedge, Risen, & Wilson, 2004).
In addition, participants also displayed a significant hedonic and temporal
asymmetry of monetary losses only in the predictive stages of a decision (Kermer,
Driver-Linn, Wilson & Gilbert, 2004). Experienced losses were significantly less potent
than initially estimated, and statistically equivalent to experienced rewards. Thus, this
confluence of misjudgments may lead a person to believe that most, if not all negative
events hold a greater affectual impact relative to themselves and gains than they actually
do, though this relationship is not immutable.
DELAY DISCOUNTING
As a brief aside that will be relevant to religious behavior and the subsequent
section on cognition, it is helpful to discuss another important type of probabilistic
decision-making, notably when consequences will occur in the future. Delay discounting
is a behavioral phenomenon illustrated by an agent’s subjective value of a reward or
punishment decaying as a function of time (Chung & Herrnstein, 1967). That is, gaining
[losing] $100 today is weighted much more heavily than gaining [losing] $100 next
week, thereby promoting gains with immediate payoffs or neglecting future punishments.
10
But, because future stimuli are important to plan for, our behavior accounts for this by a
willingness to accept larger amounts of compensation as the temporal distance from the
present also increases. For example, in order to forgo a $100 reward today, the future
reward may need to be at least $150, and the larger the value of the delayed stimuli, the
greater the rate of discounting (i.e., the less future stimuli are subjectively valued). This
discount rate can then be compared within or between-subjects in myriad ways, whether
as a function of the outcome type, their value, delay, or likelihood of attainment (Rachlin,
Raineri, & Cross, 1991).
This suggests immediate behaviors and consequences are more highly prioritized,
and generally for good reason. Until recent scientific and medical advances, one’s
lifespan was often more susceptible to the volatile fluctuations of the environment, be it
resource competition, medicinal/health quality (e.g., infant mortality rate) or natural
disasters that could prematurely end a lifespan. To the extent that it afforded an
opportunity to consume resources or reproduce, it was clearly advantageous not to wait.
However, analogous to probabilistic gambles where risk aversion can lead to over-
favoring certainty, rates of future discounting can fall into a similar pattern. Instead of a
future stimuli’s value decaying equally (exponentially) across time points, there is an
extremely rapid decrease in its value over the short-run, followed by a flattening of the
value for more distant time steps (i.e., a hyperbolic function - Mazur, 1987; Figure 2 -
Green & Myerson, 1996). And behavior following this hyperbolic model often leads to
greater impulsivity in choices, spurning optimal, future outcomes in favor of immediate,
lesser ones. These behaviors can also be empirically connected to risk aversion, with
11
Frederick (2005) suggesting individuals who display steeper discount rates will also
display more pronounced risk aversion patterns in monetary choices. Ioannou and Sadeh
(2016) did not share this conclusion for either monetary or environmental risks, but
theorized this discrepancy as possibly resulting from incentivized tasks.
Figure 2 – Exponential and Hyperbolic Discount Functions
BEHAVIORAL PARAMETERIZATION
While Kahneman and Tversky were not the first researchers to recognize and
record empirical violations of economic axioms of decision-making (Markowitz, 1952;
Allais, 1953; Ellsberg, 1962; Lichtenstein & Slovic, 1971), their pioneering work on
12
formulating a rigorous descriptive model of behavior was instrumental for a deeper
comprehension of the nuances and lability of judgments and choices. Although the theory
in a general sense helped to explain misjudgments and choice inconsistencies, perhaps its
greatest contribution was laying the foundations for a formal parameterization of risky
behavior above and beyond binary choice sets. Despite Tversky and Kahneman’s
rejection of within-subjects designs, by eliciting an indifference point (or a participant’s
point of preference reversal between two choices) through repeated measures designs, an
individual’s precise specifications can be determined and compared across choices in
other tasks, participants and studies. To contextualize this, monetary gambles are a
common methodology as they offer a quantitatively standardized scale to measure
degrees of risk and loss aversion.
risk aversion
Gambles measuring risk aversion (Kahneman & Tversky, 1979) typically consist
of a choice between two alternatives: A) either a probabilistic outcome between a reward
[loss] and a null outcome ($0), or B) a guaranteed reward [loss] - also called a Certainty
Equivalent (CE), which acts as a deterrent to the gamble - illustrated below.
Which of the following would you prefer?
A: 50% chance to win $1000 and 50% chance to win nothing
OR
B: $450 for sure
13
Winning $1000 is enticing, but walking away with $450 is also quite appealing and
forgoing this reward in lieu of potentially losing the bet might sting. By having
participants clearly define their CE with an explicit value, or by scaling the amounts
through repeated choices, their ‘Relative Risk Premium’ (RRP - Eq 1; Rieger et al., 2011)
can be calculated to determine rates of risk tolerance.
Eq. 1) 𝑅𝑅𝑃 = !" ! !"!"
The lower an individual’s CE, the more positive their RRP and the more risk-averse they
are, relative to the expected value of the gamble presented (smaller and smaller amounts
of guaranteed outcomes are not enough to prefer taking the gamble). Conversely, the
higher the CE, the more negative their RRP and the more risk-taking they are, relative to
the EV of the gamble (they will continue declining higher guaranteed amounts in pursuit
of the gamble).
loss aversion
While risk and loss are similar in nature, they are also clearly different - risk does
not always entail loss, and loss can be exhibited in riskless, or non-probabilistic,
scenarios (Tversky & Kahneman, 1991). That being said, utilizing monetary gambles
similar to risk aversion is still the most effective method of parameter estimation; only in
this case by directly pitting rewards against losses without null or certain outcomes.
Suppose Kahneman and Tversky’s (1979) original gamble was presented again, but
collapsed into a single 50-50 chance of either winning or losing $100. This parameter can
14
be estimated similarly to the RRP: calculating it through a series of accept/reject choices
of gambles with varying gain and loss amounts, or specifically defining the necessary
rewards to offset potential losses. The latter method is referred to as the Gain/Loss Ratio
(G/L - Tversky & Kahneman, 1991) and derives its name from the estimation technique,
dividing the specified gain amount by the loss amount. For example, if a participant
required $150 to offset a $100 loss, their loss aversion coefficient would be 1.5
(150/100). Lower G/L ratios imply less loss aversion; larger G/L ratios imply greater loss
aversion. There are other, more intricate ways of measuring this parameter, but the G/L
ratio is simple and can reliably differentiate behavioral patterns in a between-subjects
paradigm, like in the present project.
SUMMARY
Risk preferences stemming from aversion to uncertainty and loss can lead to
interesting violations of statistically optimal decision-making in a given scenario. And
research has shown that the asymmetry between estimating the impact of an outcome and
its actual experience can be ascribed to the subjectivity of the behavioral process.
Moreover, many of these violations can be ameliorated under objective analysis,
following a set of axioms developed to maximize one’s output that are accessible but
often neglected. The question is then, why does this occur? If individuals possess the
necessary cognitive aptitudes to overcome errors in thinking and aversion to beneficial
situations, why is it so difficult to tap into?
15
CHAPTER 3
DUAL-SYSTEM FRAMEWORK OF COGNITION
DUAL-SYSTEM MODELS
Sloman (1996).
A critical element of risk preferences is the formulation of these judgments in
relation to the mechanisms of differential, yet complementarily functioning cognitive
systems. Conceptually, the roots of this notion can be traced back to Ancient Greek
philosophy (Aristotle - cited in Sloman, 1996), with Plato comparing these processes to
controlling a chariot with two horses, one well-trained and the other wild. More recently,
psychologists and philosophers such as William James (1890/1950; cited in Sloman,
1996) have also attested to a general cognitive capacity of both parallel and serial
processing mechanisms.
However, it wasn’t until Steven Sloman (1996), that a formal, empirically
oriented theory of a dual-system model of categorization and reasoning was proposed.
Although Sloman’s own work at the time was in its nascent stages, the foundation of his
arguments encompassed an extensive literature on human language processing,
reasoning, judgments and choice. Theory and data compiled from these research domains
yielded evidence distinguishing between two modes of cognition, which he labeled the
Associative System and the Rules-based System. The associative system is characterized
as judgments governed by principles of automaticity, speed, similarity and contiguity,
and Sloman illustrates these functions as: intuition, creativity, associative memory and
visual recognition. These cognitive mechanics provide an agent with the ability to
16
quickly draw rough statistical inferences from one’s environment, based on similarity and
generalization from personal experiences that serve as flexible guidelines with soft
constraints.
Conversely, the rules-based cognitive system is primarily portrayed as a symbol
manipulator, derived from one’s culture, language and formal systems. The
computational procedures in this system are slow, but engender deliberative, logical
interpretations of causality and purpose from environmental information collected. Its
processing is productive and systematic, in diametric opposition to the automaticity of the
associative system, and applies hard constraints to the reasoning strategies utilized and
judgments rendered. Sloman suggests the rules-based cognitive mechanisms must be
initiated when greater processing is required, or to monitor, and correct if necessary, the
associative system’s encoding and conclusions. However, he also notes the application of
the rules-based system to reasoning does not guarantee optimal solutions. It is merely a
process-oriented approach to behavior.
Kahneman (2003)
Daniel Kahneman (2003) built upon Sloman’s and others’ theories (Stanovich &
West, 2000; Kahneman & Frederick, 2002), primarily driven by his and Amos Tversky’s
research on judgments and heuristics investigating errors of intuitive thinking (Sloman’s
associative system was now labeled colloquially as either ‘System 1’ [Stanovich & West,
2000] or ‘intuition’ [Kahneman & Frederick, 2002]; the rules-based system was changed
to either ‘System 2’ [S&W, 2000] or ‘Reflective’/’Reasoning’ [K&F, 2002]). Figure 3
17
below shows this framework, and now includes perceptions in the model, suggesting that
while the perceptual and intuitive processes are similar, the nature of their respective
content evaluated differs. Between intuitive and reflective cognition, this relationship is
reversed, in that the content evaluated is similar but their respective processes through
which that content is evaluated are quite different.
Figure 3. - Process and Content in Two Cognitive Systems proposed by Kahneman, 2003
Kahneman then adds further dimensions to each cognitive structure, suggesting
System 1’s proficiency may be due to ‘attribute substitution’ (Kahneman & Frederick,
2002), a method of breaking down complex problems into smaller, simpler and more
manageable pieces. However, he also cautions this system can be prone to emotional
reactivity in judgments and are governed by habits, making them difficult to control or
alter. Moreover, because cognitive capacitance is limited, System 2’s effortful application
to judgments can be laborious to initiate and sustain, thus potentially causing interference
between concurrent effortful tasks, or permitting System 1 to run amok. Kahneman
18
reaffirms Sloman’s proposed cognitive hierarchy of System 2 monitoring and suppressing
System 1’s initial judgments (for neurological evidence, see De Martino, Kumaran,
Seymour, & Dolan, 2006; for integration into the dual-system framework, see Kahneman
& Frederick, 2007), as well as asserting task-interference designs can reliably demarcate
the invocation of each system between judgments, tasks and choices.
HEURISTICS
Kahneman’s (2003) dual-system model of cognition delineated between
perceptions and intuitions. This was very important, as intuition is not constrained by
immediate information as perceptions are; rather, it is a cognitive function fully capable
of integrating past information into judgments an agent will make in the future, by
forming “impressions of the attributes of objects or perception and thought” (p. 699).
However, System 1 sits (metaphorically, of course) in an interesting position then,
sharing its processing and content of evaluation with two mutually exclusive cognitive
functions. It has the capacity to rapidly filter through massive amounts of information and
come to generally accurate conclusions of the past, present and future, sans the
metabolically taxing effort and temporality of System 2. How might it accomplish such
feats?
A heuristic is a method of problem solving employed by an agent through which,
by trading off optimality for efficiency, an adequate solution can be discovered and
recalled for future utilization. Its application to cognition originated from Herbert Simon
(1955), who coined the term ‘satisficing’, judgments and choices an individual makes
19
that are “good enough” despite lacking statistical optimality. Recalling back to the
previous section on Prospect Theory, where subjects exhibited behavioral inconsistencies
that systematically violated normative models, it may come as little surprise that Tversky
and Kahneman’s empirical exploration of heuristics (Kahneman & Tversky, 1972;
Tversky & Kahneman, 1973) paved the way for Prospect Theory as a descriptive model
of choice behavior.
Their foray into pitting the statistical intuition of experts and laypeople alike
against normative calculations led to the development of a number of specific heuristics
used to expedite the cognitive process, most notably availability and representativeness.
The availability heuristic is a memory-based evaluation of content, and
representativeness is a method of assessing or organizing similar stimuli according to a
general prototype. To conceptualize these, consider the following decision vignettes and
their respective results:
Availability:
In four pages of a novel (about 2,000 words), how many words would you expect to find that have the form “ _ _ _ _ i n g” (seven-letter words ending with “ing”)? Indicate your best estimate by circling one of the values below: 0 1-2 3-4 5-7 8-10 11-15 16+
A second form of this problem asked subjects the same question, but with seven-
letter words consisting of the form “ _ _ _ _ _ n _”. Tversky and Kahneman (1983) found
that the median estimate for words ending with “ing” was 13.4, while the median
estimate for words with the form only “n” as the 6th letter was 4.7. However, because
any seven-letter word ending with the form of “ing” has ‘n’ as the 6th letter, it is
20
statistically impossible for there to be more cases of it than seven-letter words with only
‘n’ as the 6th letter. Tversky and Kahneman argued the ability to recall words ending
with “ing” is much easier than words with only ‘n’ as the 6th letter visible, leading to the
inverted results and violation.
Representativeness:
Linda is 31 years old, single, outspoken and very bright. She majored in Philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations.
○ Linda is a teacher at an elementary school ○ Linda is active in the feminist movement (F) ○ Linda is a psychiatric social worker ○ Linda is a bank teller (T) ○ Linda is an insurance saleswoman ○ Linda is a bank teller and is active in the feminist movement (T&F)
Tversky and Kahneman (1983) asked participants to rank-order which choice is
more probable, or provide probability estimates, given the personality sketch. They found
that 82-85% of subjects believed “Linda is a bank teller and is active in the feminist
movement” (T&F) as more likely than “Linda is a bank teller” (T), violating the
conjunction axiom of probability theory (they ranked Linda as a feminist (F) as more
likely than both (T) and (T&F) though). “Conjunction” in probability theory is the
likelihood that two (or more) events will occur - e.g., the chance that it is cloudy and
windy today - but without one necessarily depending on the other (which is called
conditional probability). The corollary of this axiom is that the likelihood of both events
occurring cannot exceed the likelihood of one of the single events - i.e., there cannot be a
greater chance it is both cloudy and windy today, than only cloudy or only windy.
21
Tversky and Kahneman argued these results fit representativeness, with
participants believing that because Linda resembles the prototype of a feminist, she will
be more likely to fit a conjunction comprised of the target characterization (feminist) and
a neutral target (bank teller), above and beyond an option with just the neutral target.
Base-Rate Neglect:
While base-rate neglect is not a heuristic, it serves as a salient example of how brushing
over details of single-event likelihood estimates can skew judgments. Consider the
scenario (Casscells, Schoenberger, & Graboys, 1978):
If a test to detect a disease whose prevalence is 1/1000 has a false positive rate of 5%, what is the chance that a person found to have a positive result actually has the disease, assuming no prior knowledge? Casscells and colleagues originally found that the modal response was 0.95, the
average was 0.56, and only 18% of subjects correctly responded 0.02. Cosmides and
Tooby (1990) replicated this effect, with only 12% of subjects correctly responding
(although in the next section, they show how to reduce this effect). The reason this is
called ‘base-rate neglect’ is because subjects tended to assess the likelihood of a positive
test if the patient does have the disease, and not the likelihood of the patient having the
disease (which is unknown) given a positive test, the latter of which uses the base-rate
1/1000 in its likelihood calculations.
Additionally, Slovic, Finucane, Peters and MacGregor (2007) put forth another
heuristic, affect, wherein the emotionality of judgments and outcomes may significantly
influence behavioral processes using the “pain principle” - pursue pleasure and avoid
pain. Altogether, these cognitive shortcuts aim to drastically reduce the time and effort of
22
judgments/decisions by boiling down relevant information to the essentials and
processing them in a parallel manner, through which the most satisfactory conclusion can
be made.
CRITICAL REVIEW: EVANS (2008)
To rein in the surfeit dual-system models proposed after its initial inception, each
with similar yet distinguishable qualities, Jonathan Evans (2008) collated an extensive
review and weighed in on the manner. Although generally agreeing with the overall
notion of dual-processes of cognition, Evans suggests it does not necessarily imply a
general dual-system framework, which he calls “oversimplified and misleading” (p. 270).
Because the operations and underlying architecture of each purported system are not
mutually exclusive, and both “systems” may consist of overlapping elements of
preconscious and conscious processes, it is difficult to wholly map onto independent
mechanisms. He attempts to resolve these inconsistencies by offering a key distinction -
“type 2 processes are those that require access to a single, capacity-limited central
working memory resource” (p. 270), and that the overload and interference of working
memory can disrupt slow, sequential processing.
Other researchers have taken this a step further and argued that a uni-system
model of cognition is superior (Gigerenzer & Regier, 1996; Keren & Schul, 2009;
Kruglanski & Gigerenzer, 2011), as the evidence that ostensibly supports dual-system
theories actually lacks adequate power in explaining and predicting behavior. Gigerenzer
and Regier (1996) claim both intuitive and reflective judgments are rule-based, and
23
through a “routinization” of shared rules, deliberative judgments can become intuitive
(Kruglanski & Gigerenzer, 2011). Moreover, Keren and Schul (2009) assert that System
2 relies on System 1’s input or else it would fail, clearly signaling interdependence
between the two alleged systems.
SUMMARY
Kahneman and Tversky’s (and others) ‘Heuristics and Biases Program’ (HBP;
Kahneman, Slovic, & Tversky, 1982; Tversky & Kahneman, 1974, 1983) swept through
the research world, demonstrating instances of cognitive deficiencies in rapid
probabilistic judgments and risky decisions. And this was attributed to a reliance on
intuitive cognition and failure to initiate reflective reasoning that can override automatic
judgments. Despite their acknowledgments on the efficacy of these routine mental
faculties, the evidence was as damning as it was alarming - systematic deviations from
normative probabilistic judgments were observed, even when the correct answers were
seemingly right in the decision vignettes themselves, in both laypeople and expert
statisticians alike (Tversky & Kahneman, 1972). Man as an “intuitive statistician” was in
severe jeopardy. How could one’s judgment so crassly neglect fundamental probabilistic
principles (i.e., base-rates, conjunction) built into the scenarios? More importantly
though, was what if these conclusions were wrong?
24
CHAPTER 4
CRITICISMS
While there exists literature debating Prospect Theory’s accuracy as a descriptive
utility function of decision-making under risk (Levy & Levy, 2002a, 2002b; Baltussen,
Post, & Van Vliet, 2006; Nwogugu, 2006), many focus on the specific mathematical
parameterizations of the function (as opposed to the general estimates described earlier).
Although that scope is not the focus of this review, their methodological qualms as to
why Kahneman and Tversky’s findings may be invalid, unreliable or merely a
“transitional stage” (Gigerenzer, 1991, p. 3) are, as an entire body of past, current and
future research may be based on misleading foundations and incorrect conclusions.
Therefore, an analysis of criticisms and counterarguments is necessary to understand the
efficacy of risk preferences; if loss aversion does not hold water, then any results of this
current project are suspect.
HEURISTICS
Spearheading the opposition to Kahneman and Tversky’s ‘Heuristics and Biases
Program’ (HBP) was Gerd Gigerenzer and his formulation of a theoretical framework
called ‘Probabilistic Mental Models’ (Gigerenzer, Hoffrage, & Kleinboelting, 1991). This
work sought to properly describe and explain probabilistic judgments within a formal,
frequentist model, a manner he argued HBP lacked. Gigerenzer’s primary critiques of
HBP (Gigerenzer, 1991) were two-fold: 1) the conjunction axiom of probability theory
cannot be rigorously applied to single-event judgments/choices. Instead, frequentist-
25
framed scenarios were more appropriate. 2) HBP lacked a coherent description of what
cognitive heuristics entailed from a mechanistic perspective. He also questioned how
“biases were actually biases” when they could be reduced and even perform optimally in
many ecological situations and within-subjects presentations.
To illustrate these differences, recall the Linda Problem and how Tversky and
Kahneman argued violations of the conjunction axiom could be explained by
representativeness. Gigerenzer disagreed with this supposition, contending that Tversky
and Kahneman’s heuristics are, “largely undefined concepts and can post hoc be used to
explain almost everything” (1991, p. 102). Moreover, he condemns using single-event
probabilities with distracting fillers between ‘Bank Teller’ and “Bank teller and is active
in the Feminist Movement” as a legitimate methodology, opting instead for frequency-
based judgments with response contiguity that are much easier to understand:
There are 100 persons who fit the description above (i.e., Lindas). How many of them
are:
A) Bank tellers
B) Bank tellers and active in the feminist movement
He cites Fiedler's (1988) results, showing that only 22% and 17% of participants violated
the conjunction axiom in this context. Gigerenzer then cites Cosmides and Tooby (1990),
who showed how this type of presentation also reduced ‘base-rate neglect’, with 76% and
92% of participants properly identifying 0.02 as the statistically correct likelihood of a
patient with a positive test actually having the disease.
26
rebuttals
Kahneman and Tversky (1996) responded to Gigerenzer’s critiques, arguing that
they themselves initially found that frequentist and within-subjects methods can reduce
these biases (Tversky & Kahneman, 1983), but this effect is not as stable as Gigerenzer
insists (Tversky & Kahneman, 1974; Slovic, Fischhoff, & Lichtenstein, 1982). They
claimed this de-legitimizes his criticisms as ignoring counterevidence, and that within-
subjects presentations aren’t ecologically valid and can too obviously cue participants to
correct their responses. They further claim the development of a theoretical model of
heuristic analyses is moot and the lack of such does not discredit the role of
representativeness in intuitive prediction:
This objection misses the point that representativeness (like similarity) can be
assessed experimentally; hence it need not be defined a priori. Testing the
hypothesis that probability judgments are mediated by representativeness does not
require a theoretical model of either concept. The heuristic analysis only assumes
that the latter is used to assess the former and not vice versa (1996, p. 585)
They also found his criticism of single-event conjunctive probabilities problematic,
suggesting it is highly improbable he disregards an individual’s subjective probability,
and more so to consider them devoid of meaning and normative rigor - individuals simply
must be able to assign an appropriate likelihood to a decision without necessarily
considering their behavior in repeated trials. They further state that the dissipation of
cognitive illusions is not the only aspect of interest to them, but rather their very
27
existence - “The Mueller-Lyer Illusion, for example, “disappears” when the two figures
are embedded in a rectangular frame, but this observation does not make the illusion less
interesting” (p. 586).
Gigerenzer (1996) responded to Kahneman and Tversky (1996), echoing PMM as
a well-founded theoretical framework of heuristic judgments and his original critiques of
HBP: principles of probability theory cannot be so strictly applied to single-event
outcomes; definitions alone cannot describe mental processes - testable and falsifiable
models must be instantiated; and within-subjects testing is a viable and ecological
research methodology. However, he does make an important clarification, in that
frequentist formats are not necessarily an end-all solution to misjudgments; only that they
aid in making proper evaluations, in part by simplifying complex single-event probability
calculations. Regardless of their disagreement, Gigerenzer stresses “understanding how
people reason under a bewildering variety of circumstances” (1996, p. 595), and implores
Kahneman and Tversky to do the same for his recommendations on researching not just if
these biases disappear, but why or why not.
SEMANTIC INFERENCE
One additional theme that arose in Gigerenzer’s (1996) reply concerning the
dichotomy between probabilistic and frequentist judgments was the linguistic aspects of
the Linda Problem. Take for example, “We invited friends and colleagues to the party”
(Mellers, Hertwig, & Kahneman, 2001). Here, the ‘and’ operator can denote membership
in one or both classes (a party attendee can be either a friend, a colleague, or both).
28
Gigerenzer argued that the semantic ambiguity of “Linda is a bank teller and is active in
the feminist movement”, coupled with the use of the word ‘probable’ when ranking
choices, could possibly lead participants to believe if ‘Linda is a bank teller” then it
excludes her activity in a feminist movement, Linda being in either of the categories
satisfies the conjunction, or interpreting it as a conditional - i.e., if Linda is a bank teller,
then she is also active in the feminist movement. Because of this he claims the behavior
observed cannot be evaluated under normative standards. Since the word ‘probable’
evokes plausibility, or having the appearance of truth, this loosens the hard constraints of
probabilistic conjunction.
Linguistic analyses (Hertwig & Gigerenzer, 1999) substantiated this claim,
yielding significant differences in how participants interpreted ‘probability’ and
‘frequency’. Participants consistently inferred non-mathematical interpretations from
‘probability’ - i.e., “possibility”, “correspondence”, “applicability” and “conceivability” -
and overwhelmingly exhibited conjunction effects (83%). Conversely, participants
consistently interpreted ‘frequency’ as mathematical - i.e., “how often”, “number”,
“percentage” and “proportion” - and “never violated the conjunction rule” (p. 288). And
this effect was then replicated without using the word ‘frequency’, demonstrating the
context, and not just the terminology, is important.
Further dispute regarding the ‘and’ operator in the Linda Problem culminated
with an “adversarial collaboration” between Hertwig and Kahneman, mediated by
Mellers (Mellers et al., 2001), attempting to reduce conjunction effects by using “and
are”, “who are” and “feminist bank teller” in lieu of “and is”. However, these results were
29
not so equivocal, showing that although conjunction effects disappeared when the filler
items were removed for “and are”, “who are” and even “and is”, only “who are”
significantly reduced conjunction effects in the presence of filler options. Ultimately,
Hertwig and Kahneman still had some different perspectives on the matter regardless of
the results (as was expected). Nonetheless, some progress was made, as Hertwig
conceded that the semantic ambiguity of “and” was not the only influence of conjunction
effects, while Kahneman accepted representativeness could be lessened in different,
equally likely linguistic contexts.
Gigerenzer (2008) neatly summarizes his work as an alternative to Kahneman,
Tversky and Slovic’s HBP, proposing common misconceptions of heuristics and an
“adaptive toolbox” likely comprised of, but not limited to 10 processes (unsurprisingly
lacking availability, representativeness, or affect) that instead promote efficacious
ecological behavior: satisficing, recognition, take the best, or default (status quo).
Gigerenzer states: “the goal of a [darwinian] organism is not to follow logic, but to
pursue objectives in its environment” (p. 25) and provides evidence these methods can
reliably outperform mathematical optimization techniques in practical scenarios. He
concludes his arguments calling for a deeper understanding of heuristics in the
environment and how to mitigate errors by altering their usage or the environment itself.
PROSPECT THEORY
While the preceding counterarguments primarily focused on the efficacy of
judgments leading to putative violations of normative models of decision-making, this
30
section presents some methodological counters to the actual implemented designs of risk
preferences (Kahneman & Tversky, 1979). Recall the S-shaped value function of
Prospect Theory (PT) - risk aversion in gains and risk-seeking in losses. These results
derived from asking participants to select between binary choice sets with congruent
outcomes in the frames, i.e., none of the gambles matched a reward against a loss. Levy
and Levy (LL - 2002a, 2002b) argued this methodology is intractable given the complex
nature of investments, involving both positive and negative consequences. To support
this claim, they ran a series of experiments that included mixed-frame gambles and results
strongly suggested participants’ choices in this respect significantly deviated from PT;
single-frame gambles replicated PT’s original results. This lead LL to adamantly reject
PT as an accurate descriptive model of decision-making under uncertainty.
rebuttal
Soon after LL’s reports came out, Wakker (2003) re-analyzed their data and fired
back, asserting their results actually do support PT and that LL had failed to properly
incorporate their data into Tversky and Kahneman’s updated (1992) model. When the
data are modeled correctly, their results agree with PT and their interpretations are
rendered asinine; Baucells and Heukamp (2004) replicated this rebuttal as well (although
Baltussen, Post, & Van Vliet, 2006 argue in favor of LL). Additionally (and this is my
argument), they explicitly state PT only uses single frame outcomes. And yet, though
they cite Tversky and Kahneman’s 1992 article, they completely ignore their use of
mixed gambles leading to the conclusions in that very article.
31
LOSS AVERSION
Despite the literature criticizing heuristics and PT, the general notion of loss
carrying a greater psychophysical weight than gains seems stable (Tom et al., 2007).
While some research has sought to define its boundaries (Novemsky & Kahneman, 2005)
or pointed out instances of reversal (Harinck et al., 2007), this phenomenon marched on.
Recently however, more poignant research has begun to show weaknesses in loss
aversion as a domain-general behavior (Ert & Erev, 2013), or an explanation of status
quo biases and risk bet premiums (Gal, 2006). Walasek and Stewart (2015) took this
criticism a step further, claiming loss aversion is merely a product of research designs
and ran a modified replication of the results from Tom and colleagues’ (2007) article,
who found a loss aversion rate of about 2:1 in their study. Their criticism of this article
came from Tom and colleagues using gain values ranging from $10-$40 and loss values
ranging from $5-$20, which they argued artificially creates a 2:1 loss aversion ratio.
To amend this asymmetry, they compared this range with a balanced range, but
also included ranges with greater loss values. Results from the condition congruent with
Tom and colleagues’ study replicated their original results, but loss aversion in the
balanced condition largely disappeared. Interestingly, risk seeking was slightly increased
in the condition with larger loss ranges. The authors conclude by acknowledging loss
aversion as a behavior phenomenon that can arise as a result of “an asymmetry in the
distribution of gains and losses...reflected in people’s memory” (e.g., small but frequent
bank withdrawals, and large but fewer bank deposits - p. 10), but restate their claim that it
can be largely elicited as a product of design.
32
RECONCILIATION
This contradicting body of research is naturally of great importance. Given that
the direction of this research project concerns the behavioral aspects of loss aversion in
single-event choices, due caution must be urged regarding the validity and reliability of
any results found and interpretations made. Nonetheless, there exist other loss aversion
paradigms that are currently still methodologically credible, along with a dual-process
cognitive framework.
Thus, the manner in which judgments are generated, whether intuitively or
reflectively (Kahneman & Frederick, 2002), is crucial to decision-making, as an over-
reliance on heuristics or failure to properly initiate corrective reasoning may cause the
retention of intuitively compelling, but discordant or false responses. Frederick’s (2005)
‘Cognitive Reflection Test’ (CRT), a three item assessment of these cognitive styles
suggests participants who respond more intuitively to the questions display risk
preferences with greater loss aversive tendencies and steeper rates of delay discounting
(replicated by Hardisty & Weber, 2009; Shenhav, Rand, & Greene, 2017). Furthermore,
research has shown how cognitively intuitive proclivities are not only linked with greater
propensity to hold paranormal/religious beliefs (Aarnio & Lindeman, 2005, 2007), but
that inducing intuitive or reflective mindsets, in both positive and negative frames, can
also significantly alter these beliefs (Shenhav, Rand, & Greene, 2012).
33
CHAPTER 5
RELIGION
The reach and strength of religion’s influence on people and cultures around the
world is profound. Whether it is an individual engaging in religious practices or societal
gatherings to partake in rites and customs, these beliefs permeate through one’s
perspectives about the world and affect the decisions they make in the environment.
Hommel and Colzato (2010) offer a slightly more provocative framework of this
relationship though: religions can act as a cognitive control-guide permanently altering
perceptual and attentional processing of information, and that re-directing attention to or
away from important states helps individuals follow the principles of their belief. The
reinforcement of these values may instill them as deeply ingrained, long-lasting default
control styles, facilitating judgments and choices when faced with uncertainty.
In a review analyzing religious identification and behaviors, this sentiment seems
to be well supported, with religiosity strongly relating to self-control and regulation but
not necessarily self-monitoring (McCullough & Willoughby, 2009). These relationships
present an interesting case then. Most doctrines encourage moderation (or abstinence) of
resource consumption, charitable donations, sexual reticence, et cetera. And all of these
behavioral modulators have a clear purpose: some form of eternal salvation or karmic
reward meant to trump any immediately pleasurable activities an individual may engage
in. By not indulging in immediate, smaller rewards, they maximize future rewards.
Connecting this with risk preferences, will there be observable differences in discount
rates, future planning and risk aversion as a function of religiosity?
34
RELIGIOSITY AND RISK PREFERENCES
delay discounting
Indeed, previous research has demonstrated how participants with greater
religious beliefs (or fundamentalism) and conviction tend to be more future oriented
(Oner-Ozkan, 2007) and discount future rewards less than non-religious participants
(Carter et al., 2012). However, Paglieri and colleagues (2013) studied Dutch Calvinists (a
highly conservative Protestant sect), Italian Catholics and Atheists from both countries,
and found that while Calvinists were significantly more future oriented than Catholics
and Atheists, Catholics were significantly less future oriented than Atheists. The authors
attribute this disparity to Calvinists believing in predestination and the permanence of
actions across their lifespan, while Catholics believe in confession, permitting more
flexibility in short-term behavior. These denominational differences are vital to
understanding and disentangling between-subjects from within-subjects variability, as
observed differences may be incorrectly ascribed to the general ideology rather than the
intensity of individual factors. For example, Shenhav and colleagues’ (2017) studies
contrast this general notion, showing a steeper discount rate for individuals who self-
identify as more religious and with greater conviction, although religious demographics
were not recorded. These differences are quite stimulating and serve as a reminder of the
complexities of constructs under examination, wherein a multi-perspective approach must
ultimately be the goal. However, these nuanced analyses demand greater sample sizes
and are therefore more time consuming and expensive.
35
risk aversion
With empirical evidence observing these tentative relationships between
religiosity, cognitive styles and delay discounting that warrants credence, will similar
associations exist with risk aversion? First, in a cross-cultural analysis (Miller, 2000),
there was a tendency for both Western and Eastern cultures to display risk-averse
behavior, although significant differences were present. Participants from traditional
western cultures (U.S. & Italy) with a greater proportion of Christians, and Turkey (Islam
- Turkey is difficult to classify as it straddles Eurasia) displayed significantly stronger
negative correlations between levels of religiosity and risk-taking than traditional eastern
cultures (Japan [Buddhist and Shinto] and India [Hindu]). Moreover, between these
countries, regression coefficients were significantly stronger for the U.S. compared to all
other countries, stronger for Italy compared to India and Japan, and stronger for Turkey
compared to Japan.
Next, Noussair and colleagues (2012) measured the relationship between risk
propensity and Christian denominations, Catholic or Protestant, and initial tests
demonstrated a positive association between religiosity and risk aversion. However,
results showed no significant differences between these groups in either hypothetical or
real stake games. Furthermore, religious factors such as frequency of church attendance
and praying were significantly correlated with risk aversion, but religious belief and
conviction were not. So overall, while mixed results are present, there seems to be at least
a general trend as to how particular religions, their denominations and elements of belief
or practices may or may not play a role in risk-based behavior. But these studies were
36
largely confined to lab-based testing of surveys and designs; do these results hold in an
ecological context?
GAMBLING
While the previous studies primarily utilized general self-report and lab-based
designs, there is a closely related economic and social behavior individuals may
frequently entertain: gambling. Gambling consists of staking portions of one’s own (or
another’s) money (or objects of value) for a probabilistic chance to win greater amounts.
Gambling can take many forms, from sports/track betting to games of luck (roulette,
slots) or skill (poker), but the prevailing notion is the same. Recalling back that most
religions tend to encourage familial responsibilities and moderation of resource
consumption, and recognize that the excessive pursuit of material wealth can interfere
with these, do major religions then have explicit stances regarding gambling?
According to Kevin King (2013), they do, and negatively so. He states that
Hinduism and Buddhism frown upon gambling as non-virtuous and leading to moral and
financial destitution, by placing far too much intrinsic value on external conditions.
Abrahamic traditions (Judaism and Christianity) also discourage and even outright forbid
gambling (Islam), denouncing it as laborless or effort-free wealth acquisition and stealing
from those whom one wins against. However, different sects of these religions hold
varying gradations of attitudes towards behaviors, and therefore it is helpful to examine
gambling behavior across Christian sects.
37
Beginning with participants hailing from Las Vegas itself, Diaz’s (2000) analyses
between Christian denominations found significant negative relationships between
frequency of gambling and religious service attendance and importance of faith.
Gambling frequency also significantly differed across denominations, with Catholics
gambling the most often and Mormons the least often. Furthermore, the average amount
gambled was negatively related to service attendance and importance of faith, but did not
differ across denominations. However, Diaz states many of these correlations were only
weakly significant or of small effect size and could not substantially establish these
relationships. Additionally, Lam (2006), using data from the National Gambling Impact
Commission’s 1999 survey, found similarly consistent results. Gamblers were
significantly less likely to attend religious services than non-gamblers, and the number of
games an individual has played (0-4) emerged as a significant, negative predictor of
service attendance. However, importance of faith was not significantly different between
gamblers and non-gamblers; and while it significantly predicted gambling frequency,
only a small-moderate effect size was observed.
Ellison and McFarland (2011) did a denominational analysis of gambling
behaviors across Christian sects and found significant differences as a function of
religious conservatism. First, conservative Protestants were less likely to engage in mid-
high frequency gambling compared to mainline Protestants, Catholics and non-believers.
Furthermore, conservative Protestants had significantly stronger negative relationships
between gambling and service attendance and percentage of one’s social network
affiliated with their religious congregation - friendships with fellow members and a
38
religious leader acted as a stronger deterrent of gambling propensities than for mainline
Protestants, Catholics and Jews. Lastly, frequency of prayer was not related to gambling
behaviors.
Given the intriguing empirical connections (or lack thereof) between an
individual’s cognitive style, religiosity and risk preferences, little research has focused on
the individual behavioral components of loss aversion as it pertains to religiosity beyond
Rieger and colleagues’ (2011) initial reports on cultural dispositions. Therefore, the
purpose of this study was twofold: to further explore if Christian participants exhibit
greater levels of loss aversion across a series of hypothetical lotteries compared to
Nonreligious participants; and to examine if/which other factors of religiosity may play a
role in differential response patterns.
CHAPTER 6
PRESENT STUDY
METHODS
participants.
250 participants were initially recruited from Amazon’s Mechanical Turk
(mTurk) online survey platform and redirected to the tasks on Qualtrics via an
anonymous link. A power analysis estimated that to detect a moderate effect size (0.5)
with a power of 0.9 and significance level of 0.05 using independent samples, two-tailed t
tests, at least 86 subjects per group were necessary. More participants than recommended
were recruited to adjust for demographically asymmetric response rates and deletion of
39
poor quality data. The total sample used (reduction methods in Appendix A) in the final
analyses was 180 (50% Christian, 54.2% male, Mage = 33.97). Participants were drawn
from a U.S. population, aged 18+, who identified as either ‘Christian’ or ‘Nonreligious’,
and no restrictions on their mTurk worker status (master or non-master classification)
were used. The assessments lasted approximately 15 minutes and participants were
compensated $2.00. This study was approved by the Institutional Review Board at
Arizona State University (see Appendix C for approval documentation).
design and procedure
Participants were first presented with a series of 10 uniformly randomized,
hypothetical lottery tables (Figure 4). Each lottery table consisted of a 50-50 chance to
lose $5-$140 (intervals of $15) and asking the participants to state the minimum amount
they would require to potentially win, in order to make each lottery acceptable to play.
Figure 4. - An example lottery table presented to participants.
This specific method was used as it replicated Rieger and colleagues’ (2011)
design eliciting a simple, yet robust loss aversion coefficient termed the ‘Gain/Loss
Ratio’ (Tversky & Kahneman, 1991). Recall how the participant’s monetary value to
40
make the gamble acceptable is divided by the loss amount originally presented; and the
larger the quantitative coefficient, the greater the inferred loss aversion. G/L ratios of
each loss value and their overall average were calculated and log transformed to
normalize the data.
Next, participants were asked to provide demographic information pertaining to
their age, sex, education status and religious identification. Participants self-identifying as
‘Christian’ were then asked to provide their denomination, how often they attended
religious services, and their frequency of prayer, and then answered the
Intrinsic/Extrinsic (I/E) Religiosity Scale (Gorsuch & McPherson, 1989), three subscales
(Intrinsic, Extrinsic-Personal, Extrinsic-Social) measured on a Likert scale ranging from
0 (Strongly Disagree) to 4 (Strongly Agree). Participants who self-identified as
‘Nonreligious’ were only asked to state if they were ‘Atheist’, ‘Agnostic’, or
‘Unaffiliated’. Both groups of participants then took the nonreligious subscale of the
Nonreligious Nonspiritual Scale (NRNSS; Cragun, Hammer, & Nielsen, 2015), a
questionnaire measured on a Likert scale from 0 (Strongly Agree) to 4 (Strongly
Disagree), developed to validly and reliably measure observable differences of religiosity
and spirituality between religious and nonreligious individuals. As this study was only
interested in religiosity of participants, and for time/compensation purposes, the
spirituality subscale was not used.
The final task consisted of a series of 40 short, uniformly randomized and
individually presented statements sampled from Harris et al.’s (2009) experimental
41
stimuli, in which participants were asked to say whether they believed each statement to
be ‘True’, ‘False’, or ‘Unsure’. The statements were equally split into four categories:
Religiously True/Nonbeliever False - “Jesus really rose from the dead.”
Religiously False/Nonbeliever True - “The Biblical story of creation is false.”
Factually True - “Eagles really exist.”
Factually False - “Superman really exists.”
For reasons of time/compensation, we used this abridged version of their original
240 statements. As all of the religiously oriented statements were confirmed to
significantly differ by religious identity by Harris and colleagues in a pilot study prior to
their main experiment, and these statements do not comprise a conglomerated larger
scale, this subset can be acceptably included for exploratory purposes in relation to the
certainty of ideological affirmation without sacrificing validity or reliability. In addition
to the statement responses, response times were measured to assay whether religious and
nonreligious participants differed in how quickly they affirmed or rejected, or were
unsure of, doctrinal statements [in]congruent with their religious identity.
RESULTS
data analysis
As a manipulation check, a two-tailed t test for independent samples was run in
SPSS V.24 to determine if religious and nonreligious participants differed in their mean
42
responses to the NRNSS. The test revealed significant differences between religious (M =
1.62) and nonreligious (M = 3.48) participants in the expected direction, t(178) = -
13.539, p < .001, 95% CI [-2.14, -1.59].
Reliability analyses using Cronbach’s alpha were conducted for the ten log
transformed G/L Ratios (ɑ = .984), NRNSS (ɑ = .963), I/E Intrinsic subscale (ɑ = .855),
I/E Extrinsic-Social subscale (ɑ = .898), and the I/E Extrinsic-Personal subscale (ɑ =
.677). Because the I/E Extrinsic-Personal subscale is only three items, deletion of items to
improve reliability may cause issues with data analysis.
religious identity and loss aversion
Even after a log transformation correction and much more acceptable levels of
skewness and kurtosis, the Shapiro-Wilk test of non-normality was significant. Therefore,
independent samples Mann-Whitney U tests were used that yielded non-significant
differences between religious and nonreligious participants for every lottery, as well as
the overall averaged G/L Ratio, (p > .05). Table 1 and Figure 5 show the median values
of each lottery and the overall median, as a function of religious identity.
43
Table 1. - Median Loss Aversion Coefficients separated by Religious Identity.
Lottery Value Nonreligious G/L Ratio Religious G/L Ratio
$5 2.00 2.00
$20 2.50 2.50
$35 2.36 2.42
$50 3.00 3.00
$65 2.31 2.46
$80 2.50 2.50
$95 2.53 2.34
$110 2.49 2.73
$125 2.40 2.40
$140 2.50 2.47
Overall Median 2.49 2.47
44
Figure 5. – Median Loss Coefficients by Religious Identity.
Bivariate correlations (Table 2) assessing the relationship between loss aversion
and religious service attendance, frequency of prayer, intrinsic religiosity, extrinsic social
religiosity and extrinsic personal religiosity were all nonsignificant (p > .05). Figure 6
shows the distribution of log Gain/Loss ratios as a function of responding on the NRNSS,
separated by religious identification. These data were analyzed to ascertain if a quadratic
relationship existed, but the R-squared value was only 0.0014.
45
Table 2. - Correlation Table of Average Loss Coefficients and Religious Factors
Variable 1 2 3 4 5 6
1. Log G/L
Ratio
-
2. Service
Attendance
.00990 -
3. Freq. Prayer .06090 .33390** -
4. I/E Intrinsic .02490 -.34090** -.65190** -
5. I/E ext. social -.06690 -.50790** -.05890 .08590 -
6. I/E ext.
personal
.00390 -.06490 -.11390 .144 .15890 -
7. NRNSS avg. .033180 .31490** .46490** -.73490** -.08590 -.2790*
Pairwise N’s are provided in subscript
** Correlations significant at the .001 level (two-tailed)
* Correlations significant at the 0.05 level (two-tailed)
46
Figure 6. - Log Gain/Loss Ratio Religiosity and Identity.
religious statement responding
Pearson Chi-square difference tests in crosstabs analyses replicated Harris and
colleagues’ (2009) results of religious participants significantly affirming doctrinally true
statements and rejecting doctrinally false statements compared to nonreligious
participants. Both groups also correctly validated and rejected factually true and factually
false statements, respectively, at equivalent rates. Although Mann-Whitney U tests
showed no significant difference in the median response time, averaged across all
statements, between religious (Med. = 3.67s) and nonreligious participants (Med. =
3.40s), nonreligious participants responded significantly faster in 10 (50%) religiously
47
oriented statements, and three (7.5%) factual statements (p < .05; see Appendix B for
more details on chi-square and response time analyses).
DISCUSSION
Previous research using monetary gambles has reliably evinced a behaviorally
salient aversion to losses (Kahneman & Tversky, 1979, 1984; Tversky & Kahneman,
1981; Tom et al., 2007), the repercussions of which have been attributed to individuals
deviating from economic rationality in decision-making (Novemsky & Kahneman, 2005).
The reliance on, or slacked monitoring of, intuitive judgments susceptible to emotional
reflexivity (Kahneman, 2003) may lead to this overweighting of negative outcomes, and
has also been linked to the beliefs one holds, such as in the existence of supernatural
beings (Aarnio & Lindeman, 2005, 2007; Shenhav et al., 2012). Experiments testing the
reverse relationship of religiosity predicting risk preferences congruent with intuitive
thinking, while limited, have suggested a similar pattern (Rieger et al., 2011). Thus, the
present study set out to explore how religious identity and its corresponding factors might
relate specifically to loss aversion.
Despite replicating the overall nature of loss aversion within the sample collected
(Med. θ = 2.6), no significant main effect was found between religious and nonreligious
participants. Nor did religious factors significantly predict levels of this aversion. This is
a puzzling result; religious participants responded significantly different from their
nonreligious counterparts on the NRNSS and doctrinal statements, but actually tended to
show lower average loss coefficients. In addition, there was no quadratic relationship
48
between loss aversion and religious responding on the NRNSS present, an analogous
observation Kahan (2012) found when testing political ideology and cognitive style.
Kahan’s results suggested participants with greater political convictions (both
conservative and liberal) provided more reflective answers on Frederick’s (2005)
cognitive reflection test compared to political moderates, and he interpreted this as
ideologues being able to initiate deliberative cognition to rationalize one’s beliefs.
There appears to be no simple explanation for these results. Every study has a
likelihood of yielding null results based on the methods, sample, or merely as a
probabilistic blip, and this present study was not without its limitations. First, previous
studies have linked Christian orthodoxy (Rieger et al., 2011) and intuitive cognition
(Frederick, 2005) to increased loss aversion coefficients, and intuitive cognition to a
general religiosity (Shenhav et al., 2012), but this current experiment did not include
metrics of either religious fundamentalism or intuitive inclinations. Therefore, while the
factors of religiosity utilized did not elicit any significant predictive qualities, it may be
premature to speculate any further without understanding the cognitive strategies
participants implemented when performing this task. Second, a slight modification of the
wording used to present the lotteries was made to fit the constraints of Qualtrics’ API,
possibly leading to some participants interpret the task differently. Third, the overtness of
the recruitment script requiring a [non]religious identity could have also contributed to
this result.
Of course, it is also entirely possible that there are no practical differences
between these groups in terms of loss aversion. Therefore, more research is necessary to
49
either suggest this finding as an outlier or substantiate that the ideology and motivational
factors themselves are not predictive of loss aversion, but rather something the
fundamentalist mindset ideologies can shape. Future studies could incorporate response
times of decisions in the lottery or cognitive style items, as Pennycook, Cheyne, Koehler
and Fugelsang (2013) have shown lengthened response latencies and higher accuracy
rates in analytic reasoning tasks for religious skeptics. Finally, there are a multitude of
possible designs used in loss aversion literature, such as a dichotomous
acceptance/rejection of a gamble, and non-monetary (Asian Disease Problem) or riskless
scenarios, and integrating these along with cognitive attitudes into a religious context
may paint a better picture of this behavior.
50
REFERENCES Aarnio, K., & Lindeman, M. (2005). Paranormal beliefs, education and thinking styles.
Personality and Individual Differences, 39(7), 1227–1236. http://doi.org/10.1016/j.paid.2005.04.009.
Aarnio, K., & Lindeman, M. (2007). Religious People and Paranormal Believers. Journal
of Individual Differences, 28(1), 1–9. http://doi.org/10.1027/1614-0001.28.1.1 Allais, M. (1953). L'extension des théories de l'équilibre économique général et du
rendement social au cas du risque. Econometrica, Journal of the Econometric Society, 269-290.
Baltussen, G., Post, T., & Van Vliet, P. (2006). Violations of cumulative prospect theory
in mixed gambles with moderate probabilities. Management science, 52(8), 1288-1290.
Baucells, M., & Heukamp, F. H. (2004). Reevaluation of the results of Levy and Levy
(2002a). Organizational Behavior and Human Decision Processes, 94(1), 15-21. Baumard, N., & Boyer, P. (2013). Religious Beliefs as Reflective Elaborations on
Intuitions: A Modified Dual-Process Model. Current Directions in Psychological Science, 22(4), 295–300. http://doi.org/10.1177/0963721413478610
Bernoulli, D. (1954). Exposition of a new theory on the measurement of risk.
Econometrica, 22(1), 23–36. (Original work published in 1738). http://doi.org/10.2307/1909829
Bloom, P. (2007). Religion is natural. Developmental Science, 10(1), 147–151.
http://doi.org/10.1111/j.1467-7687.2007.00577.x Boyer, P., & Bergstrom, B. (2008). Evolutionary Perspectives on Religion. Annual
Review of Anthropology, 37(1), 111–130. http://doi.org/doi:10.1146/annurev.anthro.37.081407.085201
Casscells, W., Schoenberger, A., & Graboys, T. B. (1978). Interpretation by physicians of
clinical laboratory results. New England Journal of Medicine, 299(18), 999-1001.
51
Carnevale, P. J. (2008). Positive affect and decision frame in negotiation. Group Decision and Negotiation, 17(1), 51–63. http://doi.org/10.1007/s10726-007-9090-x
Chung, S. H., & Herrnstein, R. J. (1967). Choice and delay of reinforcement. Journal of
the experimental analysis of behavior, 10(1), 67-74. Cosmides, L., & Tooby, J. (1990, August 17). Is the mind a frequentist? Paper presented
at the Second Annual Meeting of the Human Behavior and Evolution Society, Los Angeles, CA.
Cragun, R. T., Hammer, J. H., & Nielsen, M. (2015). The NonReligious-NonSpiritual
Scale (NRNSS): Measuring Everyone from Atheists to Zionists. Science, Religion & Culture, 2(3), 36–53. http://doi.org/10.17582/journal.src/2015/2.3.36.53
De Martino, B., Kumaran, D., Seymour, B., & Dolan, R. J. (2006). Frames, Biases and
Rational Decision-Making in the Human Brain. Science, 313(5787), 684–687. http://doi.org/10.1038/020493a0
Diaz, J. D. (2000). Religion and gambling in sin-city: A statistical analysis of the
relationship between religion and gambling patterns in Las Vegas residents. The Social Science Journal, 37(3), 453-458.
Ellison, C. G., & McFarland, M. J. (2011). Religion and gambling among US adults:
Exploring the role of traditions, beliefs, practices, and networks. Journal for the Scientific Study of Religion, 50(1), 82-102.
Ellsberg, D. (1961). Risk, Ambiguity, and the Savage Axioms. The Quarterly Journal of
Economics, 75(4), 643–669. Ert, E., & Erev, I. (2013). On the descriptive value of loss aversion in decisions under
risk: Six clarifications. Judgment and Decision Making, 8(3), 214-245 Evans, J. S. B. T. (2008). Dual-processing accounts of reasoning, judgment, and social
cognition. Annual Review of Psychology, 59, 255–278. http://doi.org/10.1146/annurev.psych.59.103006.093629
Fiedler, K. (1988). The dependence of the conjunction fallacy on subtle linguistic factors.
Psychological Research, 50, 123–129.
52
Frederick, S. (2005). Cognitive Reflection and Decision Making. Journal of Economic Perspectives, 19(4), 25–42. http://doi.org/10.1257/089533005775196732
Friedman, N. P., Miyake, A., Young, S. E., Defries, J. C., Corley, R. P., & Hewitt, J. K.
(2008). Individual differences in executive functions are almost entirely genetic in origin. Journal of Experimental Psychology. General, 137(2), 201–25. http://doi.org/10.1037/0096-3445.137.2.201
Gal, D. (2006). A psychological law of inertia and the illusion of loss aversion. Judgment
and Decision Making, 1(1), 23. Gervais, W. M., Willard, A. K., Norenzayan, A., Henrich, J. (2011). THE CULTURAL
TRANSMISSION OF FAITH Why innate intuitions are necessary, but insufficient, to explain religious belief. Religion, 41 (3), p. 389-410.
Gigerenzer, G. (1991). How to Make Cognitive Illusions Disappear: Beyond “Heuristics
and Biases.” European Review of Social Psychology, 2(1), 83–115. http://doi.org/10.1080/14792779143000033
Gigerenzer, G. (1996). On narrow norms and vague heuristics: A reply to Kahneman and
Tversky. Psychological Review, 103(3), 592–596. http://doi.org/10.1037/0033-295X.103.3.592
Gigerenzer, G. (2008). Work Why Heuristics. Perspectives on Psychological Science,
3(1), 20–29. Gigerenzer, G., Hoffrage, U., & Kleinboelting, H. (1991). Probabilistic Mental Models -
A Brunswikian Theory of Confidence. Psychological Review, 98(4), 506–528. http://doi.org/10.1037/0033-295X.98.4.506
Gigerenzer, G., & Regier, T. (1996). How do we tell an association from a rule?
Comment on Sloman (1996). Gilbert, D. T., Morewedge, C. K., Risen, J. L., & Wilson, T. D. (2004). Looking forward
to looking backward: the misprediction of regret. Psychological Science : A Journal of the American Psychological Society / APS, 15(5), 346–350. http://doi.org/10.1111/j.0956-7976.2004.00681.x
53
Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley, T. P. (1998). Immune Neglect: A Source of Durability Bias in Affective Forecasting. Journal of Personality and Social Psychology, 75(3), 617–638. http://doi.org/10.1037/0022-3514.75.3.617
Gorsuch, R. L., & Mcpherson, S. E. (1989). Intrinsic/Extrinsic Measurement : I/E-
Revised and Single-Item Scales. Journal for the Scientific Study of Religion, 28(3), 348–354.
Green, L., & Myerson, J. (1996). Exponential versus hyperbolic discounting of delayed
outcomes: Risk and waiting time. American Zoologist, 36(4), 496-505. Hardisty, D. J., & Weber, E. U. (2009). Discounting future green: money versus the
environment. Journal of Experimental Psychology: General, 138(3), 329. Harinck, F., Van Dijk, E., Van Beest, I., & Mersmann, P. (2007). When gains loom larger
than losses: Reversed loss aversion for small amounts of money. Psychological Science, 18(12), 1099–1105. http://doi.org/10.1111/j.1467-9280.2007.02031.x
Harris, S., Kaplan, J. T., Curiel, A., Bookheimer, S. Y., Iacoboni, M., & Cohen, M. S.
(2009). The neural correlates of religious and nonreligious belief. PLoS ONE, 4(10), 1–10. http://doi.org/10.1371/journal.pone.0007272
Henrich, J. (2009). The evolution of costly displays, cooperation and religion. credibility
enhancing displays and their implications for cultural evolution. Evolution and Human Behavior, 30(4), 244–260. http://doi.org/10.1016/j.evolhumbehav.2009.03.005
Hertwig, R., & Gigerenzer, G. (1999). The “conjunction fallacy” revisited: How
intelligent inferences look like reasoning errors. Journal of Behavioral Decision Making, 12(4), 275–305.
Hommel, B., & Colzato, L. S. (2010). Imag(in)ing the Buddhist Brain. Zygon, 45(3), 2–8.
http://doi.org/10.1007/bf00937808 Hui, C. H., Chan, S. W., Lau, E. Y., Cheung, S. F., & Mok, D. S. Y. (2014). The role of
religion in moderating the impact of life events on material life goals: Some evidence in support of terror management theory. Mental Health, Religion & Culture, 17(1), 52-61.
54
Ioannou, C. A., & Sadeh, J. (2016). Time preferences and risk aversion: Tests on domain
differences. Journal of Risk and Uncertainty, 53(1), 29-54. James, W. (1950). The Principles of Psychology. 2 vols. New York: Henry Holt and Co.
Reprint. Kahan, D. M. (2012). Ideology, motivated reasoning, and cognitive reflection: An
experimental study. Judgment and Decision Making, 8, 407-24. Kahneman, D. (2003). A perspective on judgment and choice - Mapping bounded
rationality. American Psychologist, 58(9), 697–720. http://doi.org/10.1037/0003-066x.58.9.697
Kahneman, D., & Frederick, S. (2002). Representativeness revisited: Attribute
substitution in intuitive judgment. Heuristics and biases: The psychology of intuitive judgment, 49.
Kahneman, D., & Frederick, S. (2007). Frames and brains: elicitation and control of
response tendencies. Trends in Cognitive Sciences, 11(2), 45–46. http://doi.org/10.1016/j.tics.2006.11.007
Kahneman, D., Slovic, P., & Tversky, A. (Eds.). (1982). Judgment under uncertainty:
Heuristics and biases. Cambridge, UK: Cambridge University Press. Kahneman, D., & Tversky, A. (1972). Subjective probability: A judgment of
representativeness. Cognitive Psychology, 3, 430–454. http://doi.org/10.1016/0010-0285(72)90016-3
Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under
Risk. Econometrica, 47(2), 263–292. Kahneman, D., & Tversky, A. (1984). Choices, values, and frames. American
Psychologist, 39(4), 341. http://doi.org/10.1037/0003-066X.39.4.341 Kahneman, D., & Tversky, A. (1996). On the reality of cognitive illusions. Psychological
Review, 103(3), 582–591. http://doi.org/10.1037/0033-295X.103.3.582
55
Kelemen, D. (2004). Are Children Intuitive Theists’’? Reasoning About Purpose and Design in Nature. Psychological Science, 15(5), 295–301.
Kelemen, D., & Rosset, E. (2009). The Human Function Compunction: Teleological
explanation in adults. Cognition, 111(1), 138–143. http://doi.org/10.1016/j.cognition.2009.01.001
Keren, G., & Schul, Y. (2009). Two is not always better than one a critical evaluation of
two-system theories. Perspectives on psychological science, 4(6), 533-550. Kermer, D. A., Driver-Linn, E., Wilson, D., & Gilbert, D. T. (2006). Loss Aversion Is an
Affective Error Forecasting. Psychological Science, 17(8), 649–654. http://doi.org/10.1111/j.1467-9280.2006.01760.x
King, K. (2013). How Different Religions View Gambling.
http://www.gambleonline.co/religious-views-of-gambling/ Knight, F. H. Risk, Uncertainty, and Profit. 1921. Library of Economics and Liberty.
Retrieved December 10, 2015 from: http://www.econlib.org/library/Knight/knRUP7.html
Kruglanski, A. W., & Gigerenzer, G. (2011). Intuitive and deliberate judgments are based
on common principles. Psychological review, 118(1), 97. Kühberger, A., Schulte-Mecklenbeck, M., & Perner, J. (1999). The effects of framing,
reflection, probability, and payoff on risk preference in choice tasks. Organizational behavior and human decision processes, 78(3), 204-231.
Lam, D. (2006). The influence of religiosity on gambling participation. Journal of
Gambling Studies, 22(3), 305-320. Levy, M., & Levy, H. (2002a). Prospect theory: much ado about nothing?. Management
Science, 48(10), 1334-1349. Levy, H., & Levy, M. (2002b). Experimental test of the prospect theory value function: A
stochastic dominance approach. Organizational Behavior and Human Decision Processes, 89(2), 1058-1081.
56
Lichtenstein, S., & Slovic, P. (1971). Reversals of preference between bids and choices in gambling decisions. Journal of experimental psychology, 89(1), 46.
Markowitz, H. (1952). The utility of wealth. Journal of political Economy, 60(2), 151-
158. Mazur, J. E. (1987). An adjusting procedure for studying delayed
reinforcement. Commons, ML.; Mazur, JE.; Nevin, JA, 55-73. McCullough, M. E., & Willoughby, B. L. (2009). Religion, self-regulation, and self-
control: Associations, explanations, and implications. Psychological bulletin, 135(1), 69.
McDermott, R. (2010). Prospect Theory in Political Science : Gains and Losses From the
First Decade. Political Psychology, 25(2), 289–312. Mellers, B., Hertwig, R., & Kahneman, D. (2001). Do frequency representations
eliminate conjunction effects? An exercise in adversarial collaboration. Psychological Science, 12(4), 269-275.
Mellers, B. A., & McGraw, A. P. (2001). Anticipated Emotions as Guides to Choice.
Current Directions in Psychological Science, 10(6), 210–214. http://doi.org/10.1111/1467-8721.00151
Mellers, B., Schwartz, A., & Ritov, I. (1999). Emotion-based choice. Journal of
Experimental Psychology: General, 128(3), 332–345. http://doi.org/10.1037/0096-3445.128.3.332
Miller, A. S. Going to Hell in Asia : The Relationship between Risk and Religion in a
Cross Cultural Setting. (2016). Review of Religious Research, 42(1), 5–18. von Neumann, J., & Morgenstern, O., Theory of Games and Economic Behavior.
Princeton, NJ. Princeton University Press, 1953. Norenzayan, A., Shariff, A. F., Gervais, W. M., Willard, A. K., McNamara, R. a.,
Slingerland, E., & Henrich, J. (2014). The Cultural evolution of Prosocial Religions. Behavioral and Brain Sciences, 1–86. http://doi.org/10.1017/S0140525X14001356
57
Noussair, C. N., Trautmann, S. T., van de Kuilen, G., & Vellekoop, N. (2013). Risk aversion and religion. Journal of Risk and Uncertainty, 47(2), 165–183. http://doi.org/10.1007/s11166-013-9174-8
Novemsky, N., & Kahneman, D. (2005). The Boundaries of Loss Aversion. Journal of
Marketing Research, 42(2), 119–128. Nwogugu, M. (2006). A further critique of cumulative prospect theory and related
approaches. Applied mathematics and computation, 179(2), 451-465. Öner-Özkan, B. (2007). Future Time Orientation and Religion. Social Behavior and
Personality: An International Journal, 35(1), 51–62. http://doi.org/10.2224/sbp.2007.35.1.51
Paglieri, F., Borghi, A. M., Colzato, L. S., Hommel, B., & Scorolli, C. (2013). Heaven
can wait. How religion modulates temporal discounting. Psychological Research, 77(6), 738–747. http://doi.org/10.1007/s00426-012-0473-5
Pennycook, G., Cheyne, J. A., Koehler, D. J., & Fugelsang, J. a. (2013). Belief bias
during reasoning among religious believers and skeptics. Psychonomic Bulletin & Review, 20(4), 806–11. http://doi.org/10.3758/s13423-013-0394-3
Pew Forum on Religion & Public Life. (2012). The Global Religious Landscape:
Religious beliefs and practices. Diverse and politically relevant. Washington, DC: Author.
Pew Forum on Religion & Public Life. (2014). U.S. Religious Landscape Survey:
Religious beliefs and practices. Diverse and politically relevant. Washington, DC: Author.
Preston, J., & Epley, N. (2005). Explanations Versus Applications The Explanatory
Power of Valuable Beliefs. Psychological Science, 16(10), 826–832. Preston, J., & Epley, N. (2009). Science and God: An automatic opposition between
ultimate explanations. Journal of Experimental Social Psychology, 45(1), 238–241. http://doi.org/10.1016/j.jesp.2008.07.013
58
Rachlin, H., Raineri, A., & Cross, D. (1991). Subjective probability and delay. Journal of the Experimental Analysis of Behavior, 55(2), 233–44. http://doi.org/10.1901/jeab.1991.55-233
Rieger, M. O., Wang, M., & Hens, T. (2011). Prospect Theory Around the World. NHH
Dept. of Finance & Management Science Discussion Paper No. 2011/19. Available at SSRN: https://ssrn.com/abstract=1957606
Shenhav, A., Rand, D. G., & Greene, J. D. (2011). Divine intuition: Cognitive style
influences belief in God. Journal of Experimental Psychology. General, 141(3), 423–8. http://doi.org/10.1037/a0025391
Shenhav, A., Rand, D. G., & Greene, J. D. (2017). The relationship between
intertemporal choice and following the path of least resistance across choices, preferences, and beliefs. Judgment and Decision Making, 12(1), 1-18.
Simon, H. A. (1955). A behavioral model of rational choice. The quarterly journal of
economics, 69(1), 99-118. Sloman, S. A. (1996). The empirical case for two systems of reasoning. Psychological
Bulletin, 119(1), 3–22. http://doi.org/10.1037/0033-2909.119.1.3 Slovic, P., Finucane, M. L., Peters, E., & MacGregor, D. G. (2007). The affect heuristic.
European Journal of Operational Research, 177(3), 1333–1352. http://doi.org/10.1016/j.ejor.2005.04.006
Slovic, P., Fischhoff, B., & Lichtenstein, S. (1982). Facts versus fears: Understanding
perceived risk. In D. Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncertainty: Heuristics and biases (pp. 463-489). Cambridge, England: Cambridge University Press.
Stanovich, K. E., & West, R. F. (2000). Individual differences in reasoning: implications
for the rationality debate? The Behavioral and Brain Sciences, 23, 645-665-726. http://doi.org/10.1017/S0140525X00003435
Tom, S. M., Fox, C. R., Trepel, C., & Poldrack, R. a. (2007). The neural basis of loss
aversion in decision-making under risk. Science (New York, N.Y.), 315(5811), 515–518. http://doi.org/10.1126/science.1134239
59
Tversky, A. (1969). Intransitivity of Preferences. Preference, Belief, and Similarity, 433. Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and
probability. Cognitive psychology, 5(2), 207-232. Tversky, A., & Kahneman, D. (1974). Judgment under Uncertainty: Heuristics and
Biases. American Association for the Advancement of Science, 185(4157), 1124–1131.
Tversky, A., & Kahneman, D. (1981). The Framing of Decisions and the Psychology of
Choice. Science, 211(4481), 453–458. Tversky, A., & Kahneman, D. (1983). Extensional versus intuitive reasoning: The
conjunction fallacy in probability judgment. Psychological review, 90(4), 293. Tversky, A., & Kahneman, D. (1991). Loss Aversion in Riskless Choice: A Reference-
Dependent Model. The Quarterly Journal of Economics, 16(4), 1039–1061. Tversky, A., & Kahneman, D. (1992). Advances in Prospect-Theory - Cumulative
Representation of Uncertainty. Journal of Risk and Uncertainty, 5(4), 297–323. http://doi.org/Doi 10.1007/Bf00122574
Wakker, P. P. (2003). The Data of Levy and Levy (2002) “Prospect Theory: Much Ado
About Nothing?” Actually Support Prospect Theory. Management Science, 49(7), 979–981. http://doi.org/10.1287/mnsc.49.7.979.16383
Walasek, L., & Stewart, N. (2015). How to make loss aversion disappear and reverse:
Tests of the decision by sampling origin of loss aversion. Journal of experimental psychology: general, 144(1), 7-11.
60
APPENDIX A
DATA COLLECTED OCTOBER 2016
61
Data from 250 participants were initially collected. However, the final data
analyzed did not reflect this amount as attention tests were failed, and the unconstrained
lottery values were subject to extreme variability resulting in ratios approaching and
occasionally surpassing 100; other participants responded in values below a ratio of 0.50.
To combat this, while still maintaining an essence of the data collected and conforming to
Rieger et al.’s (2011) methodology and power analyses, participants with an average G/L
ratio greater than or equal to 100, or below 0.50 were entirely deleted, as well as any who
failed at least one of the two attention tests. Out of the 250 original participants, 228 were
retained. A second data reduction method was also run to balance religious affiliation
groups by randomly dropping Nonreligious subjects. After this round of deletion, 180 (90
each group) participants were used in the analysis.
Timing data for the affirmation statements were deleted on a case-wise basis. As
the average ‘page submit’ response time was typically around 5 seconds, any RT above
30 seconds will reflect instances above and beyond the required time to read and answer
the question, especially when the participants’ RT for other statements more closely
resembles the average. Overall, only less than 8% of timing data were deleted.
62
APPENDIX B
STATISTICAL ANALYSES
63
CHI-SQUARE DIFFERENCE TESTS
Examples of each of the four stimuli conditions showing the expected response
[in]congruency as a function of religious or factual statements:
Religiously True: “People who believe in God do so on good evidence”
Christians significantly more likely to affirm this statement, 𝜒2 = 58.397, p < .001
Religiously False: “Jesus really did not rise from the dead”
Christians significantly more likely to reject this statements, 𝜒2 = 109.515, p <
.001
Factually True: “Eating wrong foods can negatively affect health”
Both groups equally likely to affirm this statement, 𝜒2 = .994, p > .05
Factually False: “A high rate of crime is very good for society”
Both groups equally likely to reject this statement, 𝜒2 = 1.006, p > .05
64
TIMING DATA
Significant results (medians reported; C = Christian, N = Nonreligious; **denotes a
religiously false statement):
Q16** - Humans are a product of the natural world, just like all other animals.
C = 4.76 N = 3.68 p < .001
Q25** - The Biblical story of creation is purely a myth.
C = 3.98 N = 3.10 p < .001
Q32 - Jesus Christ will literally save all human beings who accept his grace.
C = 4.41 N = 3.42 p < .001
Q43** - Jesus Christ is very unlikely to return to Earth in a Second Coming
C = 4.07 N = 3.64 p = .021
Q44 - A high rate of crime is very good for society.
C = 3.72 N = 3.10 p = .013
Q51 - The Bible is the most important book we have.
C = 3.55 N = 2.72 p < .001
Q52 - Some Christians will probably go to heaven after death.
65
C = 3.96 N = 3.65 p = .032
Q54 - Superman is a fictional character.
C = 3.21 N = 2.67 p = .002
Q55 - The airplane is a human invention.
C = 3.04 N = 2.51 p = .001
Q61** - If Jesus existed, he was conceived naturally, like every other human being.
C = 5.11 N = 4.21 p = .005
Q62 - Human cultures are all exactly the same.
C = 3.66 N = 3.02 p = .002
Q65 - The Bible is free from error.
C = 3.38 N = 3.02 p =.042
Q68 - Jesus Christ really performed the miracles attributed to him in the Bible.
C = 3.67 N = 3.04 p = .017
Q73** - Original Sin is nothing more than a myth.
C = 3.54 N = 3.03 p = .011
66
APPENDIX C
IRB APPROVAL
67
EXEMPTION GRANTED Elias Robles-Sotelo Social and Behavioral Sciences, School of 602/543-4515 [email protected] Dear Elias Robles-Sotelo: On 10/6/2016 the ASU IRB reviewed the following protocol:
Type of Review: Initial Study Title: Factors of Religiosity and Loss Aversion Investigator: Elias Robles-Sotelo IRB ID: STUDY00005101 Funding: None Grant Title: None Grant ID: None Documents Reviewed: • Consent Form, Category: Consent Form;
• Stimuli, Category: Measures (Survey questions/Interview questions /interview guides/focus group questions); • Demographics, Category: Measures (Survey questions/Interview questions /interview guides/focus group questions); • mTurkRecruitmentScript.pdf, Category: Recruitment Materials; • Howatt-Robles 2016.docx, Category: IRB Protocol;
The IRB determined that the protocol is considered exempt pursuant to Federal Regulations 45CFR46 (2) Tests, surveys, interviews, or observation on 10/6/2016. In conducting this protocol you are required to follow the requirements listed in the INVESTIGATOR MANUAL (HRP-103). Sincerely, IRB Administrator