By: Thesis Supervisor:

Dr. Aziz Guergachi, Information Technology Management, TRSM

Michael Mihalicz
Supervisory Committee:
MScM Candidate, Dr. Steve Gedeon, Entrepreneurship & Strategy, TRSM;
Jun. 2019 Dr. Sara Thompson, Department of Criminology

Risk and Emotion:

Measuring the effect of emotion and other visceral factors on decision making under risk.
Introduction & Objective Cumulative Prospect Theory Recruitment & Data Collection Results
Introduction Tversky and Kahneman developed cumulative prospect theory1 (CPT) as an extension of Recruitment General Patterns
Many theories have been proposed that seek to model the cognitive mechanisms underlying decision- prospect theory to meet the criteria considered to be minimal requirements for a descriptive Participants were recruited from the Ryerson Student Research Participant • Most participants were risk averse in gains and risk seeking in losses, • The median probability weighting function follows an S-shaped pattern in
making, but few attempt to specify the role of our more primal drive states. Cognitive limitations are theory of choice. Factors considered in the selection of this theory include: Pool and compensated with 2 bonus percentage points for completing all • Between 32% and 52% of the variance is due to entity-fixed effects both gains and losses,
not solely responsible for systematic deviations from rationality and there is a growing body of • Its ability to handle multiple prospects with known and unknown probabilities, three trials. 383 people participated in the first trial of the experiment, 164 indicating that stable factors are significant. • Participants were more pessimistic in the loss domain.
literature exploring the effect of visceral factors as the more dominant drivers. Visceral factors, such as • Its nonlinear treatment of outcomes and probabilities, participated in trials one and two, and 109 participated in all three trials. Losses Probability Weighting in the Gain Domain
hunger, thirst, sexual desire, physical pain and emotions, are essential to human evolution and fitness, • Its use of reference points to account for framing effects and loss aversion, Data Collection Concave Linear Convex Total 1

and understanding how and why they alter perceptions can provide valuable insights into human The following factors contributed to the development of the data collection
Concave 25 (14%) 6 (3%) 61 (34%) 92 (51%) 1
• Its adherence to the axioms fundamental to the mathematization of choice preferences, Gains Linear 2 (1%) 23 (13%) 2 (1%) 27 (15%) 0.8

behaviour. According to George Lowenstein, a central barrier in applying decision theories to real-world method: Convex 32 (18%) 6 (3%) 22 (12%) 60 (34%)

Decision Weight
applications is their failure to account for the effects of visceral factors on decision making.7 • The availability of non-parametric elicitation methods, Total 59 (33%) 35 (20%) 85 (47%) 179 (100%) 0.6

• Captures the effects of naturally occurring emotions with little-to-no 1

This study seeks to gain a deeper understanding of the limits to human rationality by investigating the • The general consensus amongst practitioners that CPT replace EU as the new standard. researcher influence,
2

Utility Values 0.4

impact of eight visceral factors – anger, sadness, happiness, anxiety, hunger, energy, tiredness and 𝑖 𝑖−1 • Avoids the tendency for people to alter their behaviour when under 0.2
𝑛 1
stress – on three distinct elements that define risk preference under cumulative prospect theory. 𝑉(𝑓) = ෍ 𝑤 ෍ 𝑝𝑗 − 𝑤 ෍ 𝑝𝑗 𝑢 𝑥𝑖 observation. 0
0

Objective 𝑖=1
𝑗=1 𝑗=1
• It is less demanding than repeated laboratory experiments, 0
Probability
1
Probability
• Preferences can be elicited under a broader range of states, Probability Weighting in the Loss Domain
The purpose of this study is to introduce a new method for collecting data, to understand how naturally
occurring visceral states affect decisions in the moment and to identify significant limiting factors to
Parametric Specification • More data can be collected from more participants. 1
1

Two-Parameter Probability • It avoids ethical implications of inducing unpleasant visceral states.

0
0.8
rationality often unaccounted for in conventional decision theories. Two questions are of primary Power Utility Function
Weighting Function Measurement Tools

Decision Weight
interest in this study: 𝑓𝑜𝑟 𝑋 > 0 0.6

𝛿 +𝑝 𝛾+ 1

1) Which visceral factors and combinations of visceral factors have an impact on risk 𝑢 𝑥 = 𝑥𝜃 • The intensity level of the visceral factors were measured using eight 2
𝑓𝑜𝑟 𝜃 > 0, 𝑤 + (𝑝) = 0.4

𝛿 +𝑝
+ 1−𝑝 𝛾+ 𝛾+ unipolar visual analogue scales.
preference? 𝑓𝑜𝑟 𝜃 = 0, 𝑢 𝑥 = 𝑙 𝑛( 𝑥)
− 𝑝𝛾− • The ‘Big Five’ personality traits were assessed using a 10-item measure9.
−1
0.2

𝛿
2) Why and through what mechanisms do visceral factors affect decision-making 𝑓𝑜𝑟 𝜃 < 0, 𝑢 𝑥 = −𝑥 𝜃 𝑤 − (𝑝) = − 𝛾−
−1 0 1 0

𝛿 𝑝 + 1−𝑝 𝛾− • 16 demographic variables were collected. 0

under risk?
• Risk preference was measured using three non-parametric elicitation
Effects of Visceral Factors 0 Probability 1 Probability

Curvature (𝜸, 𝜶) Elevation (𝜹, 𝜷) By decomposing risk preference into three independent elements elicited using non-parametric methods, I am able to show the effects of eight visceral factors
methods aggregated into a single lottery-style questionnaire.
Research Design on the subjective interpretation of gamble attributes. In particular, my findings indicate that:
• Participants were presented with numeric and visual representations of • loss aversion is most affected by visceral factors with six of the eight factors under investigation impacting the parameter estimates,
This study is designed to assess the impact of visceral factors on the way people interpret information gambles making explicit all relevant criteria. • anger, sadness, happiness, anxiety, energy and tiredness each affect five or more of the model parameters,

𝜋(𝑝)
in lottery-style decision tasks. This is determined by eliciting three distinct elements that define • sadness has the most significant effect on risk preference while hunger and stress are significant only in their interaction with other visceral factors,
individual risk preference while participants experience varying degrees of visceral states. • distinct affective states of a similar valence sometimes have opposing effects on risk preference and affect different elements of risk preference.
Visceral states and risk preferences were recorded on three occasions over a three-week period by
Utility in the Gain Domain Utility in the Loss Domain Loss Aversion
prompting participants to complete an online questionnaire. The first trial included demographic Source: Wakker, P. P. (2010). 𝑝 70% 175%

questions, a personality trait assessment, a practice exercise, eight unipolar visual analogue scales Source: Gonzalez, R. & Wu, G. (1999). 20% Anger
Utility 50%
Sadness
and a 25-question risk elicitation questionnaire. The second and third trials include only the visual Loss Aversion Index 10%
125%
Happiness
analogue scales and risk preference elicitation questions. Anxiety
30%

𝑈 𝑌0 𝑋0
Hunger
0%
Gains 𝜆= 10%
75%

This study takes a distinct-states approach that treats visceral factors as independent while allowing Losses
𝑈(𝑋0 )𝑌0
-10% Energy
for factor interaction. This experiment does not induce visceral states, but rather captures them using
-10%
25% Tiredness
• Refers to a tendency for decision makers consider Stress
self-assessment methods as they occur naturally over the course of everyday activities. The purpose of -20%
-30%

losses more heavily than equivalent gains2. Neuroticism

this is to understand how naturally occurring visceral factors affect decisions in the moment.
-25%

• Local definition of loss aversion used to measure

-30% -50%
Conscientiousness
The risk preference of participants experiencing an elevated affective or visceral state are elicited Gross Wealth
the kink in utility function at the reference point7. -40% -70% -75%

using nonparametric elicitation methods across three domains. In each trial, participants were asked
to self-assess their current state using visual analogue scales before answering a series of lottery-style Elevation in the Gain Domain Curvature in the Gain Domain Elevation in the Loss Domain Curvature in the Loss Domain
300% 200% 450% 175%
elicitation questions shown below. / /
Risk Preference Elicitation Analysis 200%
150%
350%
125%

Basic utility,2 decision weight,3 and loss aversion4 parameters are elicited using three non- To explore the relationship between visceral factors and the magnitude of the
250%
100% 75%
100%

parametric elicitation methods aggregated into a single lottery-style questionnaire. parameters that characterize risk preference under CPT, I assume the 50%
150%
25%

Participants were asked to identify indifference values between a series of paired gambles. following relationship: 0% 50%

The indifference relation under CPT implies,

0% -25%

𝜑𝑖 = 𝛽0 + 𝛽1 𝜒1,𝑖 + ⋯ + 𝛽𝑘 𝜒𝑘,𝑖 + 𝜀𝑖 -100%

-50%

-50% -75%

All seven parameter estimates, 𝜑, are assumed unique, separable and

-150%
𝑋1 , 𝛼, 𝑋0 ∼ 𝑋𝑖 , 𝛽, 𝑋𝑗 -200%
-100% -125%

⟺ logically independent factors of risk preference. /

-250%

/
+ +
𝛼 𝑈(𝑋0 ) = 𝑤 + (𝛽)𝑈(𝑋𝑖 ) + 1 − 𝑤 + 𝛽 𝑈(𝑋𝑗 )
-300% -150% -350% -175%

Seven separate OLS and fixed-effects regressions are used to explore the effects of these factors on 𝑤 𝛼 𝑈(𝑋1 ) + 1 − 𝑤
Visceral Factors Risk Preference Stable Factors
each of the model parameters. The results include a series of measurable and systematic effects on The survey questions will be grouped into three categories:
the subjective interpretations of gamble attributes and evidence supporting the proposition that a • Gain-gambles: 𝑋1 , 𝑋0 , 𝑋𝑖 , 𝑋𝑗 > 0, Angry Conclusion References
portion of the variability in human preferences unaccounted for by cognitive limitations can be • Loss-gambles: 𝑋1 , 𝑋0 , 𝑋𝑖 , 𝑋𝑗 < 0, • Individual-level risk preference is directly impacted by the type and intensity of 1.Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: cumulative
explained by interacting visceral states. • Mixed gambles: 𝑋0 , 𝑋𝑖 < 0 < 𝑋1 , 𝑋𝑗 . Sad representations of uncertainty. J Risk Uncertain., 5, 297–323
Decision Weight Personality eight visceral factors. 2.Wakker P. (2010). Prospect Theory, For Risk and Ambiguity. Cambridge University Press.
Factors of Interest Utility under the Gamble-Tradeoff (TO) Method5 Happy
Parameters Traits • The significance of interaction effects among visceral factors highlights the
3.Gonzalez, R. & Wu, G. (1999). On the Shape of the Probability Weighting Function.
Cogn Psychol., 38, 129–166
Visceral States ‘Big Five’ Personality Traits Demographic Factors • Elicits individual utilities in the gain or loss domains under CPT, importance of assessing the impact of multiple factors simultaneously. 4.Tversky, A., & Kahneman, D. (1991). Loss Aversion in Riskless Choice: A Reference-
• Anger, • Extraversion, • Age, • Yearly income, • Eliminates probability distortions. Anxious Utility Demographic Dependent Model. Q J Econ., 106(4), 1039-1061.
• This study provides evidence to support concerns regarding the overgeneralization 5.Wakker P. & Deneffe D. (1996). Eliciting von Neumann-Morgenstern utilities when
• Sadness, • Neuroticism, • Education, • Gender, Parameters Variables
Decision Weight under Abdellaoui’s Elicitation Method6 Hungry of visceral states and risk preference. probabilities are distorted or unknown. Manag Sci., 42, 1131-1150.
• Happiness • Agreeableness, • Experience investing, • Race, 6.Abdellaoui M. (2000). Parameter-Free Elicitation of Utility and Probability Weighting
• Marital Status, • Elicits individual decision weights in the gain or loss domains under CPT, • Participants are pessimistic over a greater range of probabilities in the loss Functions. Manag Sci., 46(11), 1497-1512.
• Anxiety, • Conscientiousness, • Employment status, Energetic 7.Booij, A.S. (2009). Essays on the measurement sensitivity of risk aversion and causal
• Combined income, • Elicits measurements that are unaffected by imposed parametric assumptions. Loss Aversion domain and therefore exhibit greater probabilistic risk aversion in the loss
• Hunger, • Openness to experience. • Occupation, effects in education. Amsterdam: Thela Thesis.
domain.
• Energy, • Gross wealth, • Head of household, Loss Aversion under Booij’s Elicitation Method7 Tired 8.Lowenstein, G. (1996). Out of Control: Visceral Influences on Behavior. Organ Behav
Hum Decis Process., 65(3), 272–292.
• Tiredness, • Net wealth, • # of dependents, • Elicits a measure of loss aversion that closely resembles its original specification, • Participants exhibiting high levels of neuroticism are found to be less affected by 9.Rammstedt, B. & John, O.P. (2007). Measuring personality in one minute or less: A 10-
• # of people in house. Stressed visceral factors.
• Stress. • Years until retirement, • Uses previously elicited utility values and decision weights to infer loss aversion.
item short version of the Big Five Inventory in English and German. J Res Pers., 41,
203–212.