Master Thesis Uncovska Marie
Master Thesis Uncovska Marie
Supervised by:
Written by:
Marie Uncovska
Master in Strategy, Innovation and Management Control (WU)
Master in Management (GSOM),
Student ID: h1352382 / st066154
May 24, 2018
Abstract
This paper explores the attractiveness bias in the context of HR decisions managers face.
Influencing factors – such as gender, race, age and profession – are isolated and observed
separately. The data used stems from a television show involving participants who face
trade-offs between the personality and attractiveness attributes of potential partners as a
natural experiment to measure discrimination on the basis of physical appearance. Evidence
of strong gender differences in attractiveness bias is found, with women caring significantly
more about physical attributes than men. This finding is robust to a number of other
important characteristics including age, race, occupation, and geographical region. The
potential impact on a number of HR decisions, including staffing, hiring and recompensing
as well as implications for social and organizational behaviour are discussed.
I hereby declare that I have written this Master Thesis myself, independently and without
the aid of unfair or unauthorized resources. Whenever content has been taken directly
or indirectly from other sources, this has been indicated and the source referenced. This
Master Thesis has not been previously presented as an examination paper in this or any
other form in Russia or abroad. This Master Thesis is identical with the thesis assessed by
the examiner.
Marie Uncovska
St. Petersburg on May 24, 2018.
Contents
List of Figures ii
List of Tables ii
Introduction 1
The Issue of Attractiveness Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Research Gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3 Results 30
3.1 Preliminary Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 Gender Gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3 The Effect of Participant-Partner Similarity on Attractiveness Bias . . . . . 35
4 Discussion 38
4.1 Practical Implications and Recommendations for HR Managers . . . . . . . 38
4.2 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Conclusion 46
Bibliography 47
i
List of Figures
List of Tables
ii
Introduction
Gender differences in human behaviour and social outcomes have long been studied by
social scientists including anthropologists, sociologists, psychologists, and, most recently,
economists. Economists have found men and women to elicit different preferences in con-
sumption and investment markets, the labour market, as well as the dating and marriage
markets.1 Interest in the latter social interaction follows from the seminal work by Gary S.
Becker (1973) who formalized the marriage decision as one based on assortative matching
of partner attributes such as age, education, wealth, and physical appearance. To take an
example, in the case of positive sorting over physical appearance, a beautiful female is most
likely to form a match with a good looking male, and vice versa.
In this paper, gender differences in the demand for physical appearance, or ‘attractive-
ness’, is studied. This is done by analyzing unique data from a television dating show
involving participants who at first choose a dating partner based on the personality at-
tribute only, and in the second stage are able to reject their ‘personality partner’ after the
attractiveness attribute is revealed. Given the random assignment of partner beauty to
participants (the decision makers), the television show serves as a natural experiment to
examining gender differences in attractiveness discrimination. Based on a simple discrete
choice framework, women are found to discriminate (reject their personality partners) more
than men.
This paper first and foremost aims to answer the question ‘Which gender discriminates
more based on attractivity?’. It analyzes gender differences in the dating market, where
(similar to labour hiring decisions) a number of studies have found gender differences in
1
For a recent survey on gender differences in preferences, see Croson and Gneezy (2009).
1
Introduction
preferences for physical appearance or beauty. Most existing studies find men to care more
about the beauty of their potential partner than women (Sprecher et al., 1994; Fisman et
al. 2006; Croson and Gneazy, 2009; Hitch, Hortaçsu, and Ariely, 2010). On the other hand,
a handful of recent studies find the opposite with women discriminating more on the basis
of looks than men (Belot et al., 2008). This paper attempts to synthesize these findings
and integrate them into the organizational context.
In a second step, it is examined which other personal characteristics (variables) impact
attractiveness bias and to what extent they do so. Variables analyzed include age, race, and
profession. Hereby, the individual and combined impact of these variables is scrutinized and
evaluated. Finally, the findings are extrapolated on the general organizational and labor
market context.
Research Gap
The present study is motivated by and adds to two main streams of literature. First, the
study of cognitive biases and their effect on organizational behavior; second, the research
on gender-specific differences in behavior.
While many cognitive fallacies have been extensively analyzed and covered in academic
research in relation to their effect on organizational behavior, the attractiveness (or, in some
sources, attractivity) bias has found little coverage in academic literature. Additionally, very
little research has been conducted outside the laboratory setting.
Nonetheless, attractiveness bias can have grave impacts on organizational behavior: from
sub optimal hiring and selection decisions, to questions of cooperation vs. competition
within the organization, remuneration and promotion policies, and team dynamics. Espe-
cially the hiring and selection process is said to have a fundamental effect on the success
of not only the individual in question, but also the hiring organization (Gatewood, Feild,
and Barrick, 2008). If the hiring process itself is however already beset by bias, it will lead
to inferior decisions, high employee turnover and otherwise avoidable costs. That fair and
unbiased remuneration and promotion policies are essential not only to the workings of the
firm, but also for higher aims such as social justice is without question. But also team
dynamics and questions of cooperation vs competition can be negatively affected by bias.
In order to prevent these detrimental effects, managers first need to be aware of the
existence of this mental fallacy. Only then can they take concrete action against them.
2
1 Decision-Making and Attractiveness
Bias in the Context of Organizational
Behavior
The present study relates to three main groups of studies within the economics literature.
First, it contributes to recent works on descriptive behavioral decision theory. Hereby, the
effect different cognitive biases have on organizational decisions is analyzed. Special scrutiny
is placed on decision-making and problem-solving processes.
The second group of studies this paper relates to concerns itself with gender-specific
differences in the decision-making process. The findings build on experiments such as the
Iowa Gambling Task and (Weller, Levin, and Bechara, 2010; van den Bos, Homberg, and de
Visser, 2013) and new insights from the neurochemical field (Cosgrove et al., 2007; Ashareet
al., 2013).
Finally, the third group of academic literature this paper links to is research on preferences
in partner selection (Sprecher et al., 1994; Fisman et al. 2006; Croson and Gneazy, 2009;
Hitch, Hortaçsu, and Ariely, 2010), which finds a general preference for partner homogeneity
as well as a strong male preference for physical attractivity in their partners compared to
the female preference for attributes related to social status.
3
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
former aims to provide a framework about how decisions should be made, while the latter
analyzes the decision-making process itself and seeks to answer how decisions are made
(Hansson, 2005; p. 7f). This paper focuses on the descriptive aspects of decision theory, as
it aims to shed light on one of the underlying biases affecting human choice: attractiveness
bias.
Historically, decision theory is based on three pillars: rationality, uncertainty and, conse-
quently, the utility theory. Instead of leading to a more comprehensive, perfectly rational
decision-making strategy, most recent research has however begun to cast doubt on these
three fundamentals (Buchanan and O’Connell, 2006). Over the years, both contextual
and psychological constraints have been identified by a number of authors, from Simon’s
bounded rationality theory (Simon, 1982) to Tversky’s and Kahneman’s two-system-model
(Tversky and Kahneman, 1981) and Savage’s subjective expected utility theory (Savage,
1951). In the following, these three fundamentals as well as most recent developments in
the field will be discussed more in detail.
Rationality
4
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
There are several problems with this framework: first, irrational choices are explained
only by the agent’s lack of information. However, the existence of perfect information in
real-life situations is doubtful at best. Second, the rational choice theory ignores the agent’s
cognitive limitations: appropriately analyzing all associated information (costs, benefits and
uncertainty) takes up significant resources and is prone to be affected by several biases and
behavioral heuristics. Especially the latter point gave rise to behavioral economics, a field
that aims to account for the behavior of economic actors such as it is, rather than assuming
idealized instrumental rationality.
In response to the limitations of the rational choice model, in 1982 H. A. Simon published
the bounded rationality model. The core of the model is precisely the limitation to
human rationality – be it of cognitive or temporal nature or caused by external context.
In this context, individuals no longer seek to maximize their utility by making the optimal
choice, but rather by obtaining the most satisfactory solution (‘satisficing’ ) – making choices
that are only ‘good enough’.
Empirical studies done by Tversky and Fox (1995), Thaler (1994) and Bowles and Gintis
(2002) further shed doubt on the classical rationality assumption. While Tversky and
Fox demonstrate investor irrationality1 , Thaler focuses on systematic mental biases that
color human judgment. In ‘Quasi Rational Economics’, Thaler explores cognitive biases
caused by mental accounting2 , intertemporal choices3 and the fairness bias4 . Bowles and
Gintis (2002) go a step further away from the ‘homo economicus’ concept towards a more
cooperative, reciprocal model of human behavior, coining the term ‘homo reciprocans’.
Hereby, humans are seen as cooperative agents aiming to improve their environment. In an
experimental one-shot game, Bowles and Gintis find that human decision is often framed by
social mechanisms such as ‘reward’ and ‘punishment’, leading to cooperation even though
there is no reciprocation possible.
1
Tversky and Fox analyze investor behavior in an extensive empirical study and find that investors have a
preference for risk-averse choices in gains and choose risk-seeking behavior in losses. Additionally, there
is an overall indifference for large losses compared to risk-averse behavior for small losses (Tversky and
Fox, 1995)
2
‘Coding, categorizing and assigning values to economical outcomes’, (Thaler, 1994; p. 25f.)
3
Decisions done at one point of time that influence the possibilities available at a future point in time;
mostly concerned with what and how much to do at various points of time; influenced by relative value
assigned to future payoffs (Thaler and Shefrin, 1994)
4
The concept of ‘fairness’ alters behavior: agents do not pursue utility-maximizing strategies (Kahneman,
Knetsch, and Thaler, 1986)
5
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
In order to make informed decisions, the associated risk needs to be taken into account.
While risk itself plays a prominent role in today’s research environment, it found its way
into standard business vocabulary only after World War II, when Frank Knight first distin-
guished between risk and uncertainty. The former hereby refers to a situation in which the
probability of an outcome can be calculated; the latter denotes the opposite case, in which
outcome probabilities cannot be mathematically determined (Frank, 2012).
Uncertainty in decisions was, by then, no novelty. Giacomo Cardano mused about prob-
ability puzzles during the Renaissance; Blaise Pascal and Pierre de Fermat pondered over
the likelihood of great riches in a game of dice; Bernoulli laid the groundwork for risk man-
agement by introducing binary probability distribution. In ‘An Essay towards solving a
Problem in the Doctrine of Chances’ (1763), Bayes describes the relationship between the
probability of a hypothesis before and after receiving evidence. Based on the conditional
probability axiom, it enables updating the probabilities of hypotheses when given evidence.
In relation to decision theory, Bayes’ theorem finds wide application: a Bayesian decision-
maker will make use of Bayesian probabilities to compute expected outcomes and update
his or her beliefs.
Risk and uncertainty is a key part of decision making, as it serves to asses the expected
outcomes on any given action. However, humans are notoriously bad at not only assessing
risk, but moreover interpreting probabilities.
On the one hand, an ‘extraordinary ability to compute’ (Arrow, 1982) is being ascribed to
the average decision-maker: the theories modeling decision-making assume the agents to be
able to correctly estimate and calculate the underlying probabilities, which is not necessarily
reflected in reality. On the other hand, even when probabilities are given, decision-makers
are prone to a number of different biases and do not evaluate the risk mathematically. Re-
search by Teigen and Brun (1999) shows that context alone can modify human judgment
of probabilities: people let themselves be distracted by the usage of phrases such as ‘likely’
and ‘unlikely’ to assess the probability of any given event. Bilgin and Brenner (2013) and
Tversky and Kahneman (1981) conclude that humans tend to neglect base rate probabil-
ities and underweight individual predictions5 ; Donohue and Levitt (1998) analyze how we
5
In an experimental study, Bilgin and Brenner asked respondents how likely they would be to bring an
umbrella if the weather forecast suggested 30% probability of rain in Seattle (known for its humid
climate) and Phoenix (known for rather hot and dry climate). The likeliness of a respondent bringing an
umbrella to Seattle was significantly higher than to Phoenix; respondents were adding their perceived
base probability of rain to the probability mentioned in the weather forecast. Tversky and Kahneman
show similar results in a series of studies on heuristics and fallacies in decisions under uncertainty.
6
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
Measuring Utility
In order for decision-makers to be able to assign preferences to different outcomes and make
them comparable, they first must assign values to each and every outcome. As not every
outcome can be valued in monetary terms, ‘utility’ as a measure of preferences over a
certain set of options enables them to do so and thus forms the foundation of a wide branch
of economics. The concept of utility was first mentioned by none other than Adam Smith.
In ’The wealth of nations’ (1776), he differentiated between ‘value in use’ and ‘value in
exchange, with the former referring to the utility of some particular object and the latter
denoting the ‘power of purchasing other goods which the possession of that object conveys’.
Later, Jeremy Bentham used the utility calculus to defend equality; Jevon, Menger, and
Walras developed the relationship between utility and demand and established the concept
of marginal utilities; Pareto created the concept of Pareto efficiency around utility theory
(Stigler, 1950).
The first formal application of utilities to decision theory came only in the middle of
the 20th century, when von Neumann and Morgenstern formulated the expected utility
hypothesis. Initiated by an intellectual puzzle posed by Nicholas Bernoulli in 1713 and
solved by Daniel Bernoulli in 17386 , von Neumann and Morgenstern seek to explain popular
choices that contradict the expected value criterion. Where Bernoulli utility functions (u)
only represent preferences over monetary outcomes, the von Neumann and Morgenstern
theorem (U ) denotes preferences over lotteries of monetary outcomes. In this way, Bernoulli
utility can hardly be distinguished from ordinary utility functions over consumption bundles,
6
The St. Petersburg paradox refers to a theoretical lottery game that leads to a random variable with
infinite expected value, nonetheless seems to have little value to the players. At the core of the problem
is a sequential game of ‘heads or tails’, for which a casino offers to pay 2 dollars for each ‘heads’,
doubling the amount every round. The game ends with the first ‘tails’. The expected value of the game
is EV = ∞, thus meaning that participants should be willing to pay any price asked to participate.
However, this suggestion was encountered with widespread disbelief. The solution proposed by Daniel
Bernoulli includes the introduction of utility functions and the presumption of a decreasing marginal
utility of money. Von Neumann and Morgenstern later expanded on this concept.
7
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
while the expected utility hypothesis enables the examination of the effects of uncertainty
as well as including the attitude towards risk.
Albeit clearly factoring in a wider range of variables, the expected utility hypothesis still
does not provide a complete, comprehensive decision model. First, the model is limited
to assessing choices with an objective and known probability distribution over outcomes.
Going back to Knight’s (1921) distinction between choices under risk and choices under
uncertainty, most real-life situations can be ascribed to the latter, while von Neumann and
Morgenstern’s expected utility hypothesis only is applicable to the former. Additionally,
the theorem does not consider the context of choice such as reference points7 (Rabin, 2013)
or inconsistencies within preferences among the same choice of individuals (Tversky and
Kahneman, 1981)8 .
To address the limitations of the expected utility hypothesis – namely the objectivity
of the probability distribution over choices and the lack of contextualization of options -
several alternative models have been developed. They are usually referred to as subjective
expected utility theories and encompass works by Savage (1954) as well as Kahneman
and Tversky (1979).
Leonard Savage’s representation theorem addresses choices under uncertainty rather
than risk and combines two concepts: first, a personal utility function, second a subjective
probability distribution over different choices (Karni, 2008). This way, individual beliefs re-
lated to the probability distribution over outcomes are included in the theorem. According
to Savage, individual beliefs and thus subjective probability distributions can be deducted
from their preferences and are state-dependent. While certainly a powerful tool, the core
weakness of the theory lies in this deductibility of subjective probability distributions from
preferences, as it models an output variable (preference) to asses an input variable (prob-
ability distribution). Additionally, numerous experiments have shown that individuals do
not behave in a manner consistent with Savage’s decision model, such as the Allais (1953)
or Ellsberg paradox (1961). Adding to this, in 1979 Kahneman and Tversky first published
their paper on prospect theory, which relates to the context the choice is presented in to
asses a subjective probability distribution over multiple options. Kahneman and Tversky
assume that losses and gains are valued differently, which leads to individuals making incon-
sistent choices based on whether they are based on perceived losses or perceived gains. This
7
The ‘history’ or ‘starting point’ of every individual faced with a choice. The expected utility theorem
assumes that holding wealth constant, two identical individuals will show the same risk attitude and
utilities. However, this does not consider loss aversion – if one individual had previously lost a large
sum of money, she will exhibit a more risk-averse attitude than her otherwise identical counterpart.
8
Tversky and Kahneman (1981) demonstrate that the preferences among the same choice vary ‘depending
on how these choices are presented’.
8
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
concept is often referred to as ‘loss-aversion theory’ and denotes the preference for avoiding
losses to acquiring equivalent gains. In the context of behavioral decision making, this leads
to a so-called ‘fourfold pattern of risk attitudes: risk-averse behavior when gains have mod-
erate probabilities or losses have small probabilities; risk-seeking behavior when losses have
moderate probabilities or gains have small probabilities’ (Kahneman and Tversky, 2013).
Kahneman and Tversky demonstrate this in practice, proving that when presented with
two equal choices, one shown in terms of potential losses and the other formulated in terms
of potential gains, the latter option will be chosen. However, despite presenting a compre-
hensive alternative to the expected utility hypothesis, the difficulties inherent in applying
prospect theory in economics have made an evaluation of the concept difficult (Barberis,
2013).
Aside from the critique of utility-based decision models, the concept of utility itself has
been widely criticized. Joan Robinson denoted utility as a ‘circular, non-testable concept’
(Robinson, 1964); Albert further questioned the empirical demonstrability of utility (Albert,
1998); numerous authors further postulate that ‘beliefs and desires are vague’ and thus
cannot be clearly assigned as utilities (Stigler, 1950).
Summary
Historically, economists paid little tribute to behavioral and psychological factors in decision-
making. Discrepancies and irregularities in observable events to the prevalent mathematical
models were ascribed to the agents’ incomplete information. The critical examination of
the three fundaments of decision theory - namely rationality, uncertainty and risk and
utility theory – over the past 50 years has given rise to a new field in economics: behav-
ioral decision theory. Behavioral decision research denotes an approach to judgment and
decision-making that centers around subjective expected utility and departs from normative
theories such as the classical utility theory or Bayes’ theorem. It has developed considerably
since the 1950s, and now provides many important insights into managerial behavior. A
core component of behavioral decision theory is the research into different cognitive fallacies
that color human judgment, of which the following section provides a brief overview.
9
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
Figure 1.1: Scope of organizational decision making. Reproduced from: Langton, N., Rob-
bins, S. P., and Judge, T. A. (2013). Fundamentals of organizational behaviour.
Chapter 6 Pearson Education Canada.
The logic used to systemize these decision-making models distinguishes between so-called
closed and open systems, the key difference between them being that the former assumes all
variables involved to be either known or knowable, while the latter assumes the variables
to be unknown in their specifics as well as number. This makes closed system models deter-
9
Extracted from: Nutt, P. C. (1976). Models for decision making in organizations and some contextual
variables which stipulate optimal use. Academy of management Review, 1(2), 84-98
10
ministic and proactive in their nature, while open system models are adaptive and reactive
(Nutt, 1976)10 . According to this classification, closed system models are more appropriate
for first level managers, as most data is known (or knowable), goals and responsibilities
are clearly divided, and tasks are characterized by repetitiveness. Compared to this, open
system models are more applicable in decisions with increasing uncertainty and complexity,
such as top-level managers face. This is especially true in situations with unknown (or
unknowable) goals, a strong interaction between the goals and the environment as well as
a high number of individuals involved.
10
Following this, models [1] and [2] (bureaucratic model and normative decision theory) can be characterized
as close system logic, while models [3] and [4] (behavioral and group decision model) belong to interme-
diate system logic, with models [5] and [6] (conflict equilibrium and open system model) qualifying as
open system logic.
11
Table 1.1: Organizational decision-making models
Model Decision Criteria Key Ingredients Key Assumptions
Bureaucratic Maximum effi- (1) Define decision maker’s jurisdiction Goals known
Model [1] ciency (2) Appoint experts to office and invest Master plan to judge action is
(Fayol, 1960) power in office holder a given
(3) Rules, procedures, and precedents de- Tasks repetitive or predictable
pict decision premises Environment does not influ-
(4) Refer decisions "up" hierarchy ence choices
(5) Rewards based on adherence to master Resources adequate
plan
Normative De- Maximum sub- (1) States of nature Goals known
cision Theory jective expected (2) Alternatives Information obtainable
[2] (Savage, utility
1951) (3) Probability distribution for different Resources available
12
The fundamental attribution error refers to the tendency to overemphasize the influence of
internal causes as opposed to external ones when evaluating behavior. However, the fun-
damental attribution error only occurs when evaluating other people’s behavior; it reverses
when trying to causally explain an agent’s own behavior (Maruna and Mann, 2006). Thus,
a manager might be tempted to attribute mistakes made by a subordinate to personal char-
acteristics rather than external context. Evidence found by Martinko and Gardner (1987
and Crant and Bateman (1993) suggests that this influences the way managers interact with
their subordinates. Reason (2016) suggests this bias to be circumvented by the ‘substitution
test’, which involves mentally replacing the individual in question with a fictive character
with the same experience, qualifications and background and assessing the likeliness of him
behaving in the same way under the prevailing circumstances.
14
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
Figure 1.2: General attribution sequence. Reproduced from: Glendon, A. I., Clarke, S., and
McKenna, E. (2016). Human safety and risk management, p. 82. Crc Press.
Self-Serving Bias
The self-serving bias is, in some ways, closely related to the fundamental attribution error
and describes the cognitive process by which individuals take credit for personal success
but reject responsibility for personal failures (Shepperd, Malone, and Sweeny, 2008). The
bias is caused by the need to protect and enhance self-esteem (Hughes and Beer, 2013) and
avoid blame (Bradley, 1978). In the organizational context, the self-serving bias can have
detrimental effects to team structures, promotion policies, risk management11 , and future
planning12 (Larwood and Whittaker, 1997).
11
The self-serving bias might cause an overestimation of the risk stemming from external factors and an
underestimation of risk stemming from the managers’ own shortcomings.
12
Managers tend to overestimate their own capabilities and underestimate their flaws (Larwood and Whit-
taker, 1997).
15
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
Availability Heuristics
The availability heuristics was fist described by Tversky and Kahneman (1974) and stands
for the tendency to overestimate the likelihood of occurrence of events. It is caused by
the usage of the heuristics that an event is ‘likely or frequent if instances of it can be
readily brought to mind’ (Glendon, Clarke, and McKenna 2016). This leads to individuals
exaggerating the probability of events that are either easy to imagine or recall or especially
dramatic. Additionally, Lichtenstein et al. (1978) was able to prove that the risk of rare
negative events happening tends to be overestimated, while the contrary holds for more
common adverse effects13 . This is due to a higher exposure to accounts of rare events (be it
through media or direct communication). For executives, this poses a significant challenge,
especially when it comes to risk management and planning activities. One approach to
overcoming this bias is using a ‘systemic approach to hazard evaluation, such as an expert
system or safety auditing’ (Glendon, Clarke, and McKenna 2016).
The false consensus bias (often simply referred to as ‘consensus bias’) denotes the egocentric
‘overuse of self-related knowledge in estimating the prevalence of attributes in a population’
and can be partially traced back to the availability heuristics (Krueger and Clement, 1994).
Ross, Greene, and House (1977) find evidence for this bias in four separate studies, in which
respondents consistently exhibit the tendency to think that their own beliefs are universally
shared. For managers, this highlights the necessity for clear and transparent communication
to avoid misunderstandings. Additionally, in the absence of consensus, executives need to
avoid the assumption that individuals with a differing point of view are defective in some
way.
Anchoring Bias
Anchoring bias (also known as ‘focalism’) refers to the unconscious disproportional impor-
tance placed on the first piece of information obtained (Strack and Mussweiler, 1997). In
the course of the decision-making process, anchoring occurs when an initial piece of informa-
tion is used to make subsequent judgements. Once beliefs have been established, removing
the related association is challenging, even though the initial data is proven to be inaccu-
rate or irrelevant (Hammond, Keeney, and Raiffa, 1998). Mussweiler, Strack and Pfeiffer
13
In a series of 5 experiments with 600 adults, respondents for example estimated the loss of life caused by
murders to be higher than that brought about by strokes – even though the death toll for strokes is ten
times higher than that of murders (Lichtenstein et al., 1978).
16
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
(2000) suggest that considering the opposite and enumerating reasons that speak against
the anchor help reduce the bias.
Confirmation Bias
Confirmation bias relates to ‘the seeking or interpreting of evidence in ways that are partial
to existing beliefs’ (Nickerson 1998). Simply speaking, this refers to individuals seeking out
information that confirms or supports their hypothesis, instead of applying the scientific
method of testing their assumptions. In a business setting, confirmation bias can lead to
flawed decisions in all areas: from faulty market research to regrettable hiring decisions or
bad investment choices. As one of the most detrimental fallacies executives face, it can be
avoided by deliberately seeking evidence that contradicts the viewpoints in question.
Both Samuelson and Zeckhauser (1988) and Kahneman, Knetsch, and Thaler (1991) coined
the term ‘status quo bias’ with their research in the early 1990s. This mental fallacy connotes
the disproportional preference for maintaining the current state, even under conditions
which might be less favorable than an alternative. The status quo bias is closely related
to loss aversion14 , remorse avoidance15 and the endowment effect16 . For organizations, the
biggest threat in the status quo bias is the inflexibility and inertia it causes. Due to its
entrenchment in routine processes and patterns, a suggested countermeasure to the status
quo bias is the breaking up of these structures. An introduction of an external viewpoint as
well as the usage of objective measures to assess alternatives and compare them to the status
quo would also be recommendable. Bostrom and Ord (2006) develop the so-called ‘reverse
test’ to circumvent the status quo bias: by rephrasing the question from What happens
if I change the observed parameter? to What happens if I do not change the observed
parameter?, the decision-makers are forced to defend their conservation of the status quo.
14
The preference for avoiding losses rather than receiving equivalent gains (Kahneman, Knetsch, and Thaler
(1991)
15
Remorse (also often called regret) avoidance describes the fear that a decision will be suboptimal in
retrospect and is closely associated to risk aversion. The ‘regret theory’ was first proposed by Graham
Loomes and Robert Sugden in 1982.
16
The endowment effect designates the hypothesis that individuals ascribe a higher value to items merely
because they own them (Kahneman, Knetsch, and Thaler (1991).
17
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
17
In a study with 30 college students, respondents were asked to evaluate fabricated resumes on suitability
for managerial position. Both gender and attractivity discrimination were found to be significant.
18
In this study, 96 male and 148 female respondents were asked to evaluate the qualifications of ‘unattrac-
tive’, ‘average’ and ‘attractive’ candidates. Respondents were either asked to first rate specific quali-
fications and then formulate a hiring decision or vice versa. It was found that while the order of the
rating/hiring decision did not affect the hiring decision, sex and attractiveness did.
18
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
unattractive women are found to have lower labor force participation and marry men with
less human capital. Applying status generalization theory19 , Johnson finds that individ-
uals perceived as less attractive tend to get promoted less. In addition to this, Mobius
and Rosenblat (2006) reach the conclusion that higher physical attractiveness increases
self-confidence, which in turn tends to raise perceived competence. However, even when
controlling for self-confidence, individuals with higher attractiveness exhibit greater social
and communication skills, which increase their favorable impression with employers and
thus boosts their wages.
The effect of attractivity bias on interpersonal relationships and the willingness to coop-
erate was explored by Mulford, Orbell, Shatto, and Stockard (1998). Using a version of the
prisoner’s dilemma in a laboratory setting, Mulford (et al.) find that not only are subjects
more likely both to enter a game and cooperate with individuals they perceive as attractive,
but also point out some differences in gender: men whose self-perception is ‘attractive’ co-
operate more often than ‘average’ men, while the contrary is true for women. In addition,
subjects with a higher self-perceived attractivity showed a preference for cooperation with
other attractive individuals. In the organizational context, Lee et al. propose that the ’two
fundamental types of interdependence in organizations—cooperation and competition—result
in an opposing pattern of attractiveness discrimination’. This signifies that decision-makers
discriminate in favor of individuals they perceive as attractive when the aim is to cooperate,
while discriminating against attractive individuals when they expect to compete with them.
This is especially important for team dynamics, interpersonal relationships such as between
employer and employees and all questions of collaboration.
19
Status generalization theory refers to the ‘process by which statuses of actors external to a particular
interaction are imported and allowed to determine important features of that interaction’ (Webster and
Driskell, 1978), such as for example white male executives rising to executive positions disproportionately
to their abilities.
19
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
Figure 1.3: Task-related interdependence. Reproduced from: Lee, S., Pitesa, M., Pillutla,
M., and Thau, S. (2015). When beauty helps and when it hurts: An organi-
zational context model of attractiveness discrimination in selection decisions.
Organizational Behavior and Human Decision Processes, 128, 15-28.
While these studies all yield significant results, research outside the laboratory setting
is limited. Hamermesh (2006) aims to circumvent this problem and analyzes different
photographs accompanying elections to office in the American Economic Association. He
concludes that an exogenous increase in outward perceived attractivity raises a candidate’s
chance of success, with men benefitting from a higher ‘beauty premium’. However, the
study is based on attractiveness ratings done by four PhD students, which, mainly due to
the subjective nature of attractiveness, might render a limitation of the representativeness
of the study.
All in all, it can be said that even though research on attractiveness discrimination can be
found in academic literature, it lacks two important factors: first, general insights into the
gender-specific differences in attractiveness discrimination (‘Do women discriminate more
based on attractiveness than men?’ ); second, the application of data outside the laboratory
setting. This study aims to bridge this research gap.
20
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
literature: women are more risk-averse than men (Powell and Ansic, 1997; Jianakoplos
and Bernasek, 1998; Croson and Gneezy, 2009). These findings are closely related to the
results of the Iowa Gambling Task20 , in which women on average perform more poorly than
men (Weller, Levin, and Bechara, 2010) and are found to exhibit higher loss aversion bias
than men (van den Bos, Homberg, and de Visser, 2013). Independently from the Iowa
Gambling Task, Gächter, Johnson and Herrmann (2007) find women to exhibit more low-
averse behavior even in risk-free scenarios. The Iowa Gambling study renders additional
insights into the decision-making process of both genders: while women tend to consider all
aspects of a problem, men process information more selectively and focus on global aspects
of the information (van den Bos et al, 2013). Furthermore, females tend to understand
and integrate both subjective and objective information and include all available details
(Mayers-Levy and Maheswaran, 1991). While this leads to advantages in many different
contexts, the Iowa Gambling test relies solely on objective information.
Building on findings from Cosgrove et al., Ashare, Norris, Wileyto, Cacioppo and Strasser
(2013) examine these gender-specific differences in the context of behavioral biases. In their
research, Ashare et al. analyze the evaluation of affective stimuli (such as the positivity
offset and negativity bias21 ) based on differences in serotonin-receptor genes22 . Ashare et al.
(2013) find gender differences in the functioning of the serotonin receptor, implying there
is a biological basis for gender differences in decision-making and information processing
process. Ladenburg and Olsen (2008) conduct an empirical study to assess gender-based
differences on the anchoring bias, finding that women are significantly more susceptible to
being influenced by an initial piece of information in their subsequent decision-making.
This paper aims to enrich the existing literature by examining gender-based differences in
attractiveness discrimination.
20
The Iowa Gambling Task (IGT) refers to a psychological experiment to simulate decision-making. It was
first developed by researchers at the University of Iowa, to which it owes its name. In the experiment,
respondents were presented with four virtual decks of cards composed of ‘reward’ and ‘punishment’
cards. Due to this mixture of cards, some decks can be classified as ‘bad’, while others are counted as
‘good’ decks. The goal of the game is to maximize the amount of money won over the entire course of
the game.
21
The positivity offset is an effect in psychology in which individuals rate neutral situations as moderately
more positive than they really are. This stands in direct contrast to the negativity bias, which states
that when experiencing positive and negative stimuli of equal intensity, the negative stimulus will have
a greater effect on the general state of well-being than the positive one.
22
Serotonin is a neurotransmitter that regulates the tonus of blood vessels as well as the general sense of
well-being, which is why it is sometimes also labelled the happiness hormone (incorrectly so, as it is not
a hormone)
21
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
23
For a recent survey on gender differences in preferences, see Croson and Gneezy (2009).
24
Sprecher, Sullivan, and Hatfield (1994) use nationwide data from the National Survey of Families and
Households to establish partner preferences for men and women, finding a significant tendency for male
respondents to prefer physically attractive women, while females prefer men with higher income levels.
25
Bjerk argues that individual utility is increasing both with a partner’s attractiveness and their income
while marginal utility decreases with high individual earnings. This leads to higher preferences for
attractiveness in high-earnings individuals, which, on average, still tend to be male. Additionally, the
situation reverses for low-income individuals, who attain a higher utility from a partner’s high earnings
than from their attractiveness. Hence, even if males and females have ex ante identical underlying
preferences, males and females can still have different priorities on average in the marriage market if
males generally enter the marriage market with higher earnings trajectories than females.’
22
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
In addition to this, Regan et al. study the differences in preference based on short-
term and long-term partnerships. They observe that while women indeed place more value
on characteristics pertaining to social status and men prefer physical attributes for long-
term romantic relationships, both men and women discriminate strongly based on physical
attractivity when it comes to short-term romantic relationships.
Hitch, Hortaçsu and Ariely (2010) study the economics of match formation using a data
set acquired from an online dating platform to devise a model of partner preferences. They
identify several factors determining partner selection: similarities in ‘age, education levels,
and physical traits such as looks, height, and weight’ seem to be the prevalent variables.
This is in line with Kalmijn (1998), who postulates that partner selection is largely based
on homogeneity. This however does not necessarily reflect a specific preference for a partner
with similar attributes as the decision-maker but might rather be based on the dispropor-
tional amount of time individuals spend surrounded with others similar to themselves.
Summary
Contrary to what economic models postulated up until the 1950s, the decision-making
process is laden with irrational thought patterns, cognitive fallacies and heuristics. The
three pillars decision theory is based on - rationality, risk and uncertainty, and utility theory
all have been shown not to reflect reality accurately. New, adjusted models have arisen in
their stead, trying to compensate for the flaws inherent in their predecessors: be it Simon’s
bounded rationality model, Bowles and Gintis’ homo reciprocans, Tversky and Kahneman’s
prospect theory or two-system-model or the subjective expected utility theory.
Different cognitive biases have been extensively documented and analyzed in academic
literature. From attribution errors to mental accounting, anchoring and confirmation biases
- researchers have identified countless mental traps that color the decision-making process.
The effect of these biases on organizational structures has been at the core of a multitude
of research papers aiming to aid managers and executives to make better, more informed
choices.
More recently, probably due to the high influx of women not only into the labor mar-
ket, but also into executive positions, researchers have started looking into gender-specific
differences in decision-making. While some consensus prevails on general risk attitude and
basic information processing (inclusive for women vs. selective for men), many realms of
gender-based differences have been left unexplored. One of such is the attractiveness bias,
which is at the core of this paper. There seems to be a consensus that men discriminate
more strongly according to attractiveness (Sprecher et al., 1994; Fisman et al. 2006; Croson
23
1 Decision-Making and Attractiveness Bias in the Context of Organizational Behavior
and Gneazy, 2009; Hitch, Hortaçsu, and Ariely, 2010). On the other hand, a handful of
recent studies find the opposite with women discriminating more on the basis of looks than
men (Belot et al., 2008).
Applying findings from literature on preferences in partner selection, this study aims to
synthesize these findings and extrapolate them on the organizational context. And while
these preferences in partner selection might not at a first glance relate to the organizational
context, they certainly shed light on the question which gender discriminates more strongly
based on attractiveness. Pollard (1999) and Lee (2005) both observe significant effects of
implicit or unconscious biases in professional settings, especially so in hiring and selecting
processes. Tallerico (2014) further argues that such implicit biases give rise to a systemic
bias26 in the hiring and promotion process, which is designed to favor attractive applicants.
1.7 Hypotheses
Based on the preceding academic literature, this paper aims to demonstrate that:
26
Systemic bias refers to the underlying tendency of a process within an organization/institution to be
biased towards a particular outcome. This can for example refer to a racially biased university admission
process or, on the contrary, a scholarship program that favors minorities (affirmative action).
24
2 Data and Methodology
The data in this paper are based on the reality television dating show Dating in the Dark,
first created and aired in the Netherlands (‘Daten in het Donker’ ) on April 6, 2009. The
concept and format of the show has been reproduced in a number of other countries in-
cluding Australia, Brazil, Denmark, Germany, France, Norway, Sweden, Turkey, the United
Kingdom, and the United States. The analysis is based on a total sample of 45 episodes
from Australia (7), Brazil (6), Netherlands (8), UK (13), and the USA (11).1
25
2 Data and Methodology
on a compatibility test. Similarly, depending on the episode, the host shows the male and
female groups some video footage of inside their own homes; reveals some of the personal
items brought to the show; or invites sketch artist to draw each participant’s impression of
their partner.
Following the one-on-one dates, each participant decides if they wish to observe the
physical appearance of another participant, i.e. their ‘personality’ partner. To do so, a
participant must first invite their partner (via an email message) and then wait for them
in the dark room. Participants who do not wish to reveal their appearances leave the show
immediately, without informing the other gender group. This action mainly results when
contestants dislike the personality of each potential partner, and thus cannot identify a
‘soul mate’ on the show. Those who do comply, then enter the dark room for one last
time and reveal their looks to each other in succession. While being revealed in the light,
a participant cannot observe or hear any reactions, with their personality partner standing
silently at the other end of the dark room.
Finally, after learning about the physical appearance of their personality partner, each
participant makes one final and important decision on whether or not to go on a formal date
after the show? Participants reveal their preferences by either walking to the main balcony
(‘Yes’), or by exiting the main entrance door of the house (‘No’). By choosing to meet on
the balcony signals that participants prefer each other’s physical and mental attributes. On
the other hand, walking away from their personality partner reveals distaste for the person’s
physical appearance. Thus, in the second stage of the show, participants are able to reject
their personality partners based on the attractiveness attribute alone. The latter is used in
this study as a decision variable signifying discrimination based on physical appearance.
It is also possible for participants to be found mentally attractive by (and later to reveal
their looks to) more than one member of the opposite sex (this situation will loosely be
referred to as ‘many-to-one matching’). In such situations, during the final decision, this
‘popular’ decision maker faces a multinomial set of partners with different personality and
beauty attributes available. However, each member of the multiple interested party still
faces the desired singleton choice set of partners (that is, whether or not to wait for the
popular individual) and makes a choice purely based on the partner’s physical attributes.
In such a case of many-to-one matching, the popular decision maker (‘one’) is unable to
clearly discriminate based on the personality and looks of a single partner, since differences
across partner attributes determine the final choice. To identify such choice behaviour and
preferences, it would be required to directly measure the ‘personality’ and ‘attractiveness’
attributes, which is unfortunately not possible. To this end, the choice behaviour of the
‘one’ participant when many-to-one matching is present is not recorded. Overall, this leads
26
2 Data and Methodology
to a smaller sample of decisions than possible when only one-to-one matching prevails. The
latter issue as well as participants walking out prior to revealing their looks (that is, not
finding a ‘personality partner’ at all) results in a total sample of 222 revealed choices.3
In addition to the above decision variable, a limited set of revealed participant charac-
3
The total number of participants that appeared over the sampled episodes is 281. Four percent (11)
of these individuals correspond to the ‘popular’ type described above, who attracted more than one
personality partner. The remaining 48 contestants failed to find a personality partner and hence walked
out after the first stage of the show/experiment.
27
2 Data and Methodology
teristics was recorded, including: the age of participants, gender, racial background, and
occupation. The age and occupation variables were displayed on the screen next to each
participant’s name at the start of an episode, while the gender and race characteristics
had to be visually observed and recorded (by the researcher). The variables were assigned
numeric values (1-3 for race and occupation-related variables, 1-7 for the region variable)
in order to enable a later analysis.
The average age of participants was 27, with the youngest and oldest participant being 18
and 48 years old, respectively. Within the sample, there are slightly more female decision
makers (54%), indicating that more men (than women) left during the first stage of the
experiment (could not find a personality partner), or had more than one personality partner
(see the discussion above). In terms of racial background, the majority of contestants were
white (75%), followed by the hispanic (17%) and black groups (8%). Due to the young
age of participants, the sample distribution of occupations is comprised of relatively fewer
senior professionals: low-skilled (36%), average-skilled (38%), and high-skilled (26%).
28
2 Data and Methodology
from the entire range 0 ≤ βi < 1. On the other hand, there also exist those participants of
the type “I like the personality, but I don’t like the looks” (accept pj , reject aj ), for whom
we can observe both the decision yij = 1 and yij = 0. For those contestants who decided
to go on a date with their personality partner (‘Yes’), it is known that βi < 0.5 since they
do not prefer the looks of their partner, however still promise to go on a date with them
after the show (for these individuals; partner looks are less important than personality).
Finally, participants who decide not to go on a future date (yij = 0) reveal a higher relative
preference for attractiveness βi > 0.5, i.e. looks are more important than personality. The
latter type of individuals is labeled as ‘attractiveness discriminators’.
The main focus in this paper is to test for gender differences in the choice behaviour of
the latter type of individuals. To do so, the analysis is based on regressions of the form
where the dependent variable equals 1 if participant i rejects partner j upon observing his
or her Attractiveness, and 0 otherwise; Femalei is an indicator variable that equals 1 if
participant i is a female, and 0 if a male; Xi = [ Agei , Racei , Occupationi ] and Xj = [ Agej ,
Racej , Occupationj ] are vectors of other personal characteristics describing participant i
and partner j; Zij includes show-specific variables, namely the ‘Region’ (or country) of
production; and ij is an iid random error term capturing the unobserved variables and
other idiosyncratic preference shocks. As any observed degree of gender differences in beauty
discrimination may also depend on other ‘own’ and ‘partner’ characteristics, corresponding
interaction terms to control for such heterogeneity have been included.
29
3 Results
30
3 Results
US 57 0.25 (0.43)
UK 89 0.38 (0.23) 0.04 pb(uk) > pb(us)
Australia 39 0.23 (0.43) 0.05 pb(aus) < pb(uk)
Brazil 40 0.30 (0.46) 0.18 pb(bra) < pb(uk)
Italy 6 0.17 (0.41) 0.14 pb(ita) < pb(uk)
Netherlands 40 0.25 (0.44) 0.07 pb(ned) < pb(uk)
Denmark 12 0.08 (0.29) 0.02 pb(den) < pb(uk)
‘Discrimination rate’ is the proportion of participants that chose not to date their ‘personality partner’
after the partner’s looks were revealed. Young if age ≤ 30; Mature if age > 30. Standard deviations in
parentheses. The resulting p-values from parametric Z-tests of differences between sample proportions
are reported in the fourth column. For within-region comparison, the UK is taken as a base scenario
(due to large sample size).
United States (25%), Australia (23%), and the Netherlands (25%). The 15 percentage
point (positive) difference in beauty consciousness is broadly consistent with the more than
110% increase in expenditure on cosmetic surgery and other beauty enhancing treatments
by British households between 2005 and 2010 (BAAPs 2010), compared to a notably lower
77% increase observed in the USA since the year 2000 (ASPS 2010). This type of social
development (of ’keeping up physical appearances’) in the UK is also supported by the
31
3 Results
growing number of men and especially teenagers receiving beauty treatments (with individ-
uals from the latter group entering the dating market for the first time).
32
3 Results
Table 3.2: The effect of participant and partner attributes on the probabil-
ity of attractiveness discrimination
Linear probability model; robust standard errors in parentheses. The dependent variable in all regressions is
‘Discriminate’; that is set to 1 if the participant chose not to date their ‘personality partner’ after the partner’s
looks were revealed, and 0 otherwise. ‘Female’ is an indicator variable that takes on a value of 1 if the participant
is female, and 0 if male. ‘Mature’ is an indicator variable that equals 1 if the participant is > 30 years old, and 0
otherwise. * significant at 10%; ** significant at 5%; *** significant at 1%.
33
3 Results
third model version (3)). In addition to this, significant results can be observed for the
male-only model version: men in the US and Australia discriminate less than their female
counterparts, by almost 30% in both cases. This is significant on a 5% significance level
for the specified parameters. As to the interaction term, neither occupation nor race of
the partner demonstrate significance in any of the specified model forms. However, the age
of the partner seems to matter: mature partners get rejected, on average, 15.4% more in
the general model variant and 20% more in the gender-specific model adaptation for female
decision-makers only. These results are significant on a 5% respective 10% significance level.
The gender-split models point towards two main gender-based differences: first, the
above-described regional differences in discrimination rates, with men from the US and
Australia discrimination significantly less than their female counterparts; second, the higher
age-based discrimination by male participants. Especially the second lies in line with most
academic research, which points towards males seeking, on average, younger female partners
(see e.g. Buss, 1989).
Perhaps a bit surprisingly, none of the other attributes yield significant results. Academic
research into different sociographic preferences for men and women often highlights the
female preference for partners with higher social status (Becker, 1973; Buss, 1994; Croson
and Gneezy, 2009), which does not reflect in the described model. Looking into more detail
in the gender-specific differences in participant attributes (see Table 3.3), we can see only
marginal differences in occupation levels (the average occupation level is 0.294 point higher
for male participants).
This would point towards Bjerk’s (2009) thesis that preferences in partner selection naturally
do not vary across genders but are caused by disparities in income distribution. Bjerk argues
that individual utility is increasing both with a partner’s attractiveness and their income,
while the marginal utility of a partners’ income decreases with high individual earnings. This
leads to higher preferences for attractiveness in high-earnings individuals, who, on average,
still tend to be male. Additionally, the situation reverses for low-income individuals, who
34
attain a higher utility from a partner’s high earnings than from their attractiveness. This
hypothesis will be further examined in section 3.3.
35
Table 3.4: The effect of participant-partner similarity on attractiveness discrimination
(1) (2)
Male Female All Male Female All
Both mature -0.020 -0.028 -0.030
(0.083) (0.094) (0.082)
Same occupation -0.110 0.195** -0.115
(0.083) (0.089) (0.081)
Same race -0.041 -0.100 -0.063 -0.062 -0.085 -0.132
(0.107) (0.116) (0.093) (0.106) (0.111) (0.089)
Age j > age i 0.090 0.124 0.075
(0.102) (0.093) (0.101)
Occupation j > occupation i -0.028 -0.203** -0.048
(0.105) (0.090) (0.103)
Both mature × female 0.011
36
(0.116)
Same occupation × female 0.314***
(0.119)
Same race × female -0.017 0.112
(0.093) (0.093)
(Age j > age i) × female 0.083
(0.133)
(Occup. j > occup. i) × female -0.136
(0.133)
Number of Obs. 102 120 222 102 120 222
R2 0.02 0.04 0.06 0.01 0.05 0.05
Linear probability model; robust standard errors in parentheses. Constant term included in models. The dependent variable in all regressions is ‘Discriminate’;
that is set to 1 if the participant chose not to date their ‘personality partner’ after the partner’s looks were revealed, and 0 otherwise. ‘Both mature’ is an
indicator variable that takes on a value of 1 if both the participant and partner are > 30 years old, and 0 otherwise. ‘Same occupation’ is an indicator variable
that takes on a value of 1 if the participant and partner have the same occupation level, and 0 otherwise. ‘Same race’ is an indicator variable that takes on a
value of 1 if both the participant and partner are of the same racial background (white, black, hispanic), and 0 otherwise. i denotes the participant, j denotes
the partner. * significant at 10%; ** significant at 5%; *** significant at 1%.
3 Results
Results Summary
Overall, this study sheds light on a little-researched topic and gives valuable insights on
attractiveness bias using unique data from a natural experiment. Both the first hypothesis
and sub-hypothesis,
could be confirmed. Outward physical appearance is correlated with selection decisions, with
evidence of strong gender differences in attractiveness bias confirmed. Women are found
to care significantly more about physical attributes than men. This finding is robust to a
number of other important characteristics including age, race, occupation, and geographical
region. The second hypothesis,
Hypothesis 2 The impact of the attractiveness bias is independent from individual char-
acteristics such as age, race, region and profession,
37
4 Discussion
In the following, the results will be discussed in the organizational context. While the
preferences analyzed in the previous section might not at a first glance relate to the business
context, they certainly shed light on a subconscious bias affecting daily behavior. Pollard
(1999) and Lee (2005) both observe significant effects of implicit or unconscious biases
in professional settings, especially so in hiring and selecting processes. Additionally, such
mental traps might give rise to systemic biases in companies and organizations. Biased
selection, hiring, promotion and compensation systems undermine the efficient functioning
of a firm by selecting suboptimal candidates or creating career advancement systems not
rooted in merit and accomplishments, thus giving rise to a whole range of behavior not
aligned to the organization’s goals. As Anton Chekhov said: ‘People should be beautiful in
every way - in their faces, in the way they dress, in their thoughts, and in their innermost
selves.’. Deciding based on just one of these implies a lack of appreciation for the remaining
factors, which will only rarely be aligned with the company’s interest.
To avoid the detrimental effect of such biases, practical implications for HR managers will
be provided based on the functional areas of HR management and organizational decision-
making theory. Moreover, concrete recommendations for actions and process structures
inside organizations will be provided. Lastly, limitations of the present study will be ana-
lyzed and discussed.
38
4 Discussion
all ages, occupations and races to the same degree. Looking into academic literature, this is
in line with for example findings by Marlow, Schneider, and Nelson (1996), who discover that
there is only a negligible difference between experienced and inexperienced managers when
it comes to attractiveness discrimination. For organizations, this implies the need to train
managers on all levels to counteract subconscious biases; not just these in junior positions.
Developing and instituting mandatory anti-discrimination trainings and guidelines should
thus be a given in the organizational setting. Additionally, the lack of significant differences
between occupation levels means that attractiveness bias is equally relevant for all types of
organizations, with high-skill or technical positions being no exception.
Second, weak evidence of attractiveness discrimination being influenced by geographic
region is found. This implies that attractiveness bias is, to a certain degree, ingrained in a
country’s culture. As observed in the case of the UK as well as Brazil, increased spending
on beauty products as well as cosmetic surgery and other beauty enhancing treatments
might be a good indicator for this. To put this in context of the Russian market: according
to BusinessStat (2015), within the period between 2009 and 2013, Russia’s beauty market
grew by 46%. In addition to this, the sales value of Russia’s cosmetic market in 2015
reached nearly 14 billion dollars, placing Russia 4th in Europe. This could point towards
a stronger attractiveness discrimination in Russian firms when compared to the US or the
Nordic countries analyzed in this study – however, additional research is needed to prove
this culture-specific differences in attractiveness bias.
For managers working abroad or with international teams, this highlights the need for
increased awareness of these cultural differences in cognitive biases. Lastly, women are
found to discriminate significantly more than men (on average by 15%). This is a factor
not to be overlooked: in the US as well as in Europe, HR departments are mainly dom-
inated by women. According to the Bureau of Labor Statistics data (2016), over 70% of
HR managers are female in both countries. Recruiting, which involves the attraction of
applicants for potential employment, and selection, which refers to the identification of ap-
plicants ‘with the necessary knowledge, skills, abilities, and other characteristics that will
help the organization achieve its goals’ (Noe et al., 2013), are areas crucial to circumventing
the attractiveness bias, as it is the place where an individual makes a first impression on
the organization. As Gatewood, Feild, and Barrick (2008) postulate, the first impression
is especially susceptible to attractiveness bias: the outward physical appearance (including
personal style or lack thereof and hygiene factors) might easily overweight the objective
qualification for a position.
HR managers already are aware of the importance of a uniform recruitment and selection
system (starting for example with the formulation of neutral, objective job requirements)
39
4 Discussion
and a neutralized selection process. However, how does this translate into practice?
As attractiveness bias impacts selection decision most (Huffcutt et al., 2011), the following
aims to provide recommendations specific to the selection and hiring process. From a more
practical perspective, this could for example include anonymized applications, which aim
to level the playing field in the first step of the selection process or a trimodal approach
to selection decisions. Should these measures prove difficult to implement (e.g. due to cost
or time constraints or low acceptance), minor adjustments in existing processes can serve
to counteract attractiveness bias. In the following, these three recommendations will be
discussed in more detail.
Anonymous applications target the first selection step (preselection) in personnel se-
lection procedures, that is, the sighting the application documents before the interview
invitation as compared to actual hiring decisions. In the concept of anonymous applica-
tions, applicants are to refrain from taking a photograph and stating certain biographical
and personal data, such as name, address, date of birth, age, marital status or citizenship
(see Döse, 2012). Thus, the core focus of anonymized applications are the objective factors,
such as work experience and education of a candidate. Unlike the conventional application,
no year numbers are given so that no conclusion on the age is possible. The goal is to
focus the personnel decision-makers on the first selection decision to the qualification and
professional aptitude of the candidate and avert subconscious biases, subjective feelings,
possible prejudices or stereotypes. There are several ways to realize anonymous application
procedures:
• Anonymized application forms that are made available to the applicant online or in
another form. The completed forms are returned either digitally or by post directly
to the organization
• Subsequent anonymization, e.g. a neutral body of the organization blacks out all
relevant personal information
After completing the first round of selection and deciding which candidates will be invited
to the job interview, the personnel decision-makers then receive the complete documents
of the candidates concerned. In the main selection process, the anonymity is removed and
the applicant has the opportunity to present his strengths unfiltered. The anonymization of
40
4 Discussion
the application documents deprives personnel managers of the subjective information basis
and can be regarded as a means of increasing the objectivity in the preselection decision.
Since the discrimination rate is highest in the first selection stage and falls in the course
of an application process (see Krause et al., 2010, p.4), the use of anonymous application
procedures could be a successful tool for less discriminatory job placement.
Internationally, anonymous applications have long been used in recruitment procedures
(see Krause et al., 2012) and researched through various model projects in different countries
(including Sweden, France, Switzerland, the Netherlands). As Åslund and Skans (2012) find,
such anonymous applications level the playing field in the screening process. These findings
support the results of a study conducted by Watson and Johnston in 2017, during which they
find that discrimination by attractiveness significantly decreases when companies forego an
applicant’s personal information in the first screening process.
International experience shows that the chances of being invited to a job interview are
increasing for certain groups of applicants by anonymized applications. The disadvantage
of such a process for many companies and institutions is the additional work involved in
recruitment and the delay in the selection process. There are also two major difficulties
in the implementation and practicability of anonymous application procedures: On the
one hand, more efficient methods for anonymization must be developed and, on the other
hand, the characteristics to be anonymized must be defined. However, in terms of the non-
objective candidate feature of physical attractiveness, the anonymous application process,
if handled efficiently and methodically standardized, seems to be a tried and tested means
against discrimination.
The same holds for the interview process itself, where the physical attractiveness as the
first information of an applicant can influence the assessment and thus the further course
of the conversation both positively and negatively. As a result, attractive candidates may
be perceived as more competent, more appropriate, and more likely to be persuasive, as
they are given more time to talk because of their attractiveness by the interviewer (Huffcutt
et al., 2011). Often, interviews are not well-structured enough to minimize the impact of
extraneous candidate characteristics for bias. Schuler and Marcus (2006, p. 220) therefore
advocate the use of a multimodal interview on the basis of the trimodal model of aptitude
diagnostics, which will be described more in detail below.
Personnel selection-relevant instruments can generally be divided into various categories:
on the one hand, personality-oriented procedures which in turn split into characteristics-
41
4 Discussion
oriented and simulation procedures, and, on the other hand, biography-oriented procedures.
42
4 Discussion
abilities (Krohne and Hock 2007), based on the concrete job requirements of the position
to be filled. Again, the results are usually well comparable. However, in both methods,
the required effort is significantly greater than in biography-oriented selection process, the
applicant acceptance, however, much lower.
Simulation-oriented methods focus on situation-specific behavior of an applicant in var-
ious work-typical contexts (eg assessment center, work samples, etc.). With simulation-
oriented methods, performance or performance for specific occupational requirements is to
be recorded in realistic simulations. Decisive for the validity of the procedure is the content
validity, in specific the content of the tested activity with the actual occupational activity
and its central elements. This procedure is often used by applicants (for example, new
entrants) who have few biographical information.
43
4 Discussion
Applicants themselves can play a significant role in averting attractiveness bias. Rather
mundane and obvious actions such as keeping a professional dress code and maintaining
a professional composure and atmosphere during the interview to appropriate mannerisms
and language can go a long way in deterring bias. Additionally, the focus of the interview
should in any case be steered towards and applicant’s objective qualifications. However,
these actions can only go so far in diminishing the negative effects of the attractiveness bias.
As Shanani, Dipboyle and Gehrlein (1993) find, attractiveness bias in face-to-face interviews
affects most significantly female applicants with low perceived attractiveness. Further re-
search into the so-called beauty is beastly-effect suggests that high perceived attractiveness
can be disadvantageous to female applicants applying for traditionally male positions, with
this effect reversing when they apply for typically female positions (Johnson, Podratz, Dip-
boye, and Gibbons, 2010). No such effects were found for male applicants (Johnson et al.,
2010). These findings seem to point towards the obvious advantage of keeping average looks
for female applicants; a feat easier said than done.
In a recent research paper on mitigating the effects of the attractiveness bias, Johnson,
Sitzman and Nguyen (2014) find that acknowledging appearance counteracts the negative
effects of the beauty is beastly-effect. In their paper, Johnson et al. analyze the effect of
highly attractive female applicants acknowledging their looks when applying for typically
masculine positions in three separate studies, finding a positive correlation between this
acknowledgement intervention and reduced attractiveness bias. Exempting this study, very
little academic research has focused on testing methods alleviating or even eliminating such
bias, highlighting the need for further research.
44
4 Discussion
4.2 Limitations
There are two main concerns with the present study: first, the external validity, second,
its applicability to the business context. Regarding the first, it could be argued that the
participants of the show are not representative of the population. While it is true that
primary data on attractiveness bias in organizations would be ideal, it is hard to obtain and
difficult to quantify. The underlying bias is slow-acting, which would require data collected
extensively over the course of many years, as well as implicit, which makes it very hard to
discern from a large amount of data. The clear advantage of the present study is that the
reject or accept decision of participants at the end of the show makes the bias explicit, while
naturally controlling for personality traits. As to the representativeness of participants, it
can be contended that the demographic characteristics of the participants do not greatly
differ from population means regarding race and occupation. Especially the latter proves to
be of great advantage when compared to laboratory experiments, where participants tend
to be very homogeneous (Belot et al., 2012).
As to the applicability of the results on the business context, it can be argued that deeply
ingrained behavioral heuristics and biases are not limited to personal life and greatly affect
the professional context. While there is a consensus in academic research that attractiveness
bias is prevalent in the workspace (see for example Hamermesh and Biddle, 1994; Mobius
and Rosenblat, 2006; and Johnson 2010), little effort has been put into analyzing gender-
specific differences of this particular phenomenon. Additionally, this study is unique due to
shedding light on the effect other personal characteristics, such as age, race and occupation,
have on attractiveness bias.
Lastly, it could be said that the context of a TV show leads to a form of Hawthorne
effect: participants might modify their behavior due to the fact that they are aware of being
observed. However, evidence found by Beetsma and Schotman (2001), Levitt (2004), and
Belot et al. (2012) suggests otherwise. Especially the last stipulates that while participants
might be tempted to modify their behavior away from ‘frowned upon’ actions such as racial
discrimination, attractiveness carries no such stigma.
45
Conclusion
To conclude, this present study has three main findings: first, perhaps surprisingly, women
are found to discriminate based on outward physical appearance more than men (by 15%).
This contradicts most findings in academic literature based on laboratory experiments
(Sprecher et al., 1994; Fisman et al., 2006; Croson and Gneazy, 2009; Hitch, Hortaçsu,
and Ariely, 2010); however, it is in line with recent research conducted by Belot (2012) in
TV game show settings. Second, attractiveness bias is independent from age, race, and
occupation. This is mostly in line with recent findings by Marlow, Schneider, and Nelson
(1996). Lastly, regional allocation has weak effects on attractiveness discrimination. This
implies that attractiveness bias is, to a certain degree, ingrained in a country’s culture and
highlights the need for additional research into this area.
Given the highly stylized nature of the interactions on this television show, one must
use extraordinary caution in trying to draw general conclusions from these results. How-
ever, due to the subconscious nature of cognitive biases, some general recommendations
and implications for HR managers can be formulated. These include, but are not limited
to: an implementation of a system of checks and balances and code of conduct; the neu-
tralization of the selection process; the education of employees about cognitive biases; the
usage of quantifiable performance control measures as well as promotion and compensation
tied to objective goals and milestones. An unbiased, efficient process based on merit and
accomplishments should be the logical consequence of such measures.
46
Bibliography
Agthe, M., Spörrle, M., and Maner, J. K. (2010). Don’t hate me because I’m beautiful:
Anti-attractiveness bias in organizational evaluation and decision making. Journal of Ex-
perimental Social Psychology, 46(6), 1151-1154.
Albert, H. (1998). Marktsoziologie und Entscheidungslogik: zur Kritik der reinen Ökonomik.
Mohr Siebeck.
Ariely, D. (2009). Predictably irrational (p. 71). New York: HarperCollins.
Arrow, K. J. (1982). Risk perception in psychology and economics. Economic inquiry,
20(1), 1-9.
Ashare, R. L., Norris, C. J., Wileyto, E. P., Cacioppo, J. T., and Strasser, A. A. (2013).
Individual differences in positivity offset and negativity bias: Gender-specific associations
with two serotonin receptor genes. Personality and Individual Differences, 55(5), 469-473.
Ashmore RD. (1981). Sex stereotypes and implicit personality theory. Hamilton DL
(Ed.), Cognitive processes in stereotyping and intergroup behavior (pp. 37-81).
Åslund, O., and Skans, O. N. (2012). Do anonymous job application procedures level the
playing field?. ILR Review, 65(1), 82-107.
Barberis, N. C. (2013). Thirty years of prospect theory in economics: A review and
assessment. Journal of Economic Perspectives, 27(1), 173-96.
Becker, G. S. (1973). A theory of marriage: Part I. Journal of Political economy, 81(4),
813-846.
Beetsma, R. and Schotman, P. (2001) Measuring Risk Attitudes in a Natural Experiment:
Data from the Television Game Show Lingo, Economic Journal, 111, 821-48.
Belot, M., Bhaskar, V., and Van de Ven, J. (2012) Beauty and the Sources of Discrimi-
nation, Journal of Human Resources.
Belot, M., Bhaskar, V., and van de Ven, J. (2010). Promises and cooperation: Evidence
from a TV game show. Journal of Economic Behavior and Organization, 73(3), 396-405.
Belot, M., Bhaskar, V., and Van de Ven, J. (2009) Can Observers Predict Trustworthi-
ness? Review of Economics and Statistics, forthcoming.
Bernstein, P. L., and Bernstein, P. L. (1996). Against the gods: The remarkable story of
risk (pp. 1269-1275). New York:Wiley.
47
4 Discussion
Bilgin, B., and Brenner, L. (2013). Context affects the interpretation of low but not high
numerical probabilities: A hypothesis testing account of subjective probability. Organiza-
tional Behavior and Human Decision Processes, 121(1), 118-128.
Bjerk, D. (2009). Beauty vs. earnings: Gender differences in earnings and priorities over
spousal characteristics in a matching model. Journal of Economic Behavior and Organiza-
tion, 69(3), 248-259.
van den Bos, R., Homberg, J., and de Visser, L. (2013). A critical review of sex differences
in decision-making tasks: focus on the Iowa Gambling Task. Behavioural brain research,
238, 95-108.
Bostrom, N., and Ord, T. (2006). The reversal test: eliminating status quo bias in ap-
plied ethics. Ethics, 116(4), 656-679.
Bowles, S., and Gintis, H. (2002). Behavioural science: homo reciprocans. Nature,
415(6868), 125.
Bradley, G. W. (1978). Self-serving biases in the attribution process: A reexamination
of the fact or fiction question. Journal of personality and social psychology, 36(1), 56.
Buchanan, L. and O’Connell, A. (2006) A Brief History of Decision Making, Harvard
Business Review, 84(1), 32.
Buss, D. M. (1994). Psychological sex differences: Origins through sexual selection.
Buss, D. M. (1989). Sex differences in human mate preferences: Evolutionary hypotheses
tested in 37 cultures. Behavioral and brain sciences, 12(1), 1-14.
Byrne, K. A., and Worthy, D. A. (2015). Gender differences in reward sensitivity and
information processing during decision-making. Journal of Risk and Uncertainty, 50(1),
55-71.
Cann, A., Siegfried, W. D., and Pearce, L. (1981). Forced attention to specific applicant
qualifications: Impact on physical attractiveness and sex of applicant biases. Personnel
Psychology, 34(1), 65-75.
Collins, B. E., and Guetzkow, H. S. (1964). A social psychology of group processes for
decision-making. Wiley.
Cosgrove, K. P., Mazure, C. M., and Staley, J. K. (2007). Evolving knowledge of sex dif-
ferences in brain structure, function, and chemistry. Biological psychiatry, 62(8), 847-855.
Crant, J. M., and Bateman, T. S. (1993). Assignment of credit and blame for perfor-
mance outcomes. Academy of Management Journal, 36(1), 7-27.
Croson, R., and Gneezy, U. (2009). Gender differences in preferences. Journal of Eco-
nomic literature, 47(2), 448-74.
Delbecq, A. L., and Van de Ven, A. H. (1971). A group process model for problem iden-
tification and program planning. The Journal of Applied Behavioral Science, 7(4), 466-492.
48
4 Discussion
Dipboye, R. L., Fromkin, H. L., and Wiback, K. (1975). Relative importance of applicant
sex, attractiveness, and scholastic standing in evaluation of job applicant resumes. Journal
of Applied Psychology, 60(1), 39.
Donohue, J. J., and Levitt, S. D. (1998). Guns, violence, and the efficiency of illegal
markets. The American economic review, 88(2), 463-467.
Eagly AH, Ashmore RD, Makhijani MG, Longo LC. (1991). What is beautiful is good,
but...: A meta-analytic review of research on the physical attractiveness stereotype. Psy-
chological Bulletin, 110, 109-128.
Edwards, W. (1954). The theory of decision making. Psychological bulletin, 51(4), 380.
Entman, R. M. (2007). Framing bias: Media in the distribution of power. Journal of
communication, 57(1), 163-173.
Fisman, R., Iyengar, S., Kamenica, E., and Simonson, I. (2006) Gender Differences in
Mate Selection: Evidence From a Speed Dating Experiment, Quarterly Journal of Eco-
nomics, 121, 673-97.
Fisman, R., Iyengar, S., Kamenica, E., and Simonson, I. (2008) Racial Preferences in
Dating, Review of Economic Studies, 75, 117-32.
Gächter, S., Johnson, E. J., and Herrmann, A. (2007). Individual-level loss aversion in
riskless and risky choices.
Glendon, A. I., Clarke, S., and McKenna, E. (2016). Human safety and risk manage-
ment. Crc Press. Gore, W. J. (1964). Administrative Decision-Making: A Heuristic
Model. Wiley.
Hamermesh, D. and Biddle, J (1994) Beauty and the Labor Market, American Economic
Review, 84, 1174-94.
Hamermesh, D. S. (2006). Changing looks and changing “discrimination”: The beauty of
economists. Economics Letters, 93(3), 405-412.
Hammond, J. S., Keeney, R. L., and Raiffa, H. (1998). The hidden traps in decision
making. Harvard business review, 76(5), 47-58.
Hansson, S. O. (2005). Decision theory: A brief introduction, Royal Institute of Technol-
ogy (KTH), Stockholm.
Hitsch, G. J., Hortaçsu, A., and Ariely, D. (2010). Matching and sorting in online dating.
American Economic Review, 100(1), 130-63.
Hosoda, M., Stone-Romero, E. F., and Coats, G. (2003). The effects of physical at-
tractiveness on job-related outcomes: A meta-analysis of experimental studies. Personnel
Psychology, 56(2), 431-462.
Hughes, B. L., and Beer, J. S. (2013). Protecting the self: The effect of social-evaluative
threat on neural representations of self. Journal of Cognitive Neuroscience, 25(4), 613-622.
49
4 Discussion
Jianakoplos, N. A., and Bernasek, A. (1998). Are women more risk averse?. Economic
inquiry, 36(4), 620-630.
Johnson, S. K., Podratz, K. E., Dipboye, R. L., and Gibbons, E. (2010). Physical attrac-
tiveness biases in ratings of employment suitability: Tracking down the “beauty is beastly”
effect. The Journal of social psychology, 150(3), 301-318.
Johnson, S. K., Sitzmann, T., and Nguyen, A. T. (2014). Don’t hate me because I’m
beautiful: Acknowledging appearance mitigates the “beauty is beastly” effect. Organiza-
tional Behavior and Human Decision Processes, 125(2), 184-192.
Johnston, D. (2010) Physical Appearance and Wages: Do Blondes Have More Fun?, Eco-
nomics Letters, 108, 10-12.
Judge, T. A., Hurst, C., and Simon, L. S. (2009). Does it pay to be smart, attractive,
or confident (or all three)? Relationships among general mental ability, physical attractive-
ness, core self-evaluations, and income. Journal of Applied Psychology, 94(3), 742.
Kahneman, D., Knetsch, J. L., and Thaler, R. (1986). Fairness as a constraint on profit
seeking: Entitlements in the market. The American economic review, 728-741.
Kahneman, D., Knetsch, J. L., and Thaler, R. H. (1991). Anomalies: The endowment
effect, loss aversion, and status quo bias. Journal of Economic perspectives, 5(1), 193-206.
Kahneman, D. (2011). Thinking, fast and slow. Macmillan.
Kahneman, D., and Tversky, A. (2013). Prospect theory: An analysis of decision under
risk. Handbook of the fundamentals of financial decision making: Part I (pp. 99-127).
Kalmijn, M. (1998). Intermarriage and homogamy: Causes, patterns, trends. Annual
review of sociology, 24(1), 395-421.
Karni, E. (2008). Savage’s subjective expected utility model. The New Palgrave Dictio-
nary of Economics (pp. 1-5).
Knight, F. H. (2012). Risk, uncertainty and profit. Courier Corporation.
Krueger, J., and Clement, R. W. (1994). The truly false consensus effect: An ineradicable
and egocentric bias in social perception. Journal of personality and social psychology, 67(4),
596.
Ladenburg, J., and Olsen, S. B. (2008). Gender-specific starting point bias in choice
experiments: evidence from an empirical study. Journal of Environmental Economics and
Management, 56(3), 275-285.
Langton, N., Robbins, S. P., and Judge, T. A. (2013). Fundamentals of organizational
behaviour. Pearson Education Canada.
Larwood, L., and Whittaker, W. (1977). Managerial myopia: Self-serving biases in orga-
nizational planning. Journal of applied psychology, 62(2), 194.
Lee, A. J. (2005). Unconscious bias theory in employment discrimination litigation. Harv.
50
4 Discussion
51
4 Discussion
Pollard, D. A. (1999). Unconscious Bias and Self-Critical Analysis: The Case for a Qual-
ified Evidentiary Equal Employment Opportunity Privilege. Wash. L. Rev., 74, 913.
Powell, M., and Ansic, D. (1997). Gender differences in risk behaviour in financial
decision-making: An experimental analysis. Journal of economic psychology, 18(6), 605-
628.
Rabin, M. (2013). Risk aversion and expected-utility theory: A calibration theorem.
Handbook of the Fundamentals of Financial Decision-Making: Part I (pp. 241-252).
Reason, J. (2016). Managing the risks of organizational accidents. Routledge.
Rittenberg, L., and Tregarthen, T. (2012). Principles of Microeconomics (Version 2.0).
Washington, DC: Flat World Knowledge. Chapter 6. Regan, P. C., Levin, L., Sprecher,
S., Christopher, F. S., and Gate, R. (2000). Partner preferences: What characteristics do
men and women desire in their short-term sexual and long-term romantic partners?. Jour-
nal of Psychology and Human Sexuality, 12(3), 1-21.
Regan, P. (1998) Minimum Mate Selection Standards as a Function of Perceived Mate
Value, Relationship Context, and Gender, Journal of Psychology and Human Sexuality, 10,
53-73.
Robinson, Joan (1962). Economic Philosophy. Harmondsworth, Middle-sex, UK: Pen-
guin Books.
Ross, L., Greene, D., and House, P. (1977). The “false consensus effect”: An egocentric
bias in social perception and attribution processes. Journal of experimental social psychol-
ogy, 13(3), 279-301.
Samuelson, W., and Zeckhauser, R. (1988). Status quo bias in decision making. Journal
of risk and uncertainty, 1(1), 7-59.
Savage, L. J. (1951). The theory of statistical decision. Journal of the American Statis-
tical Association, 46(253), 55-67.
Savage, L. J. (1954). The foundations of statistical inference. Methuen.
Sen, A.K. (1997). Rationality and social choice. Readings in Microeconomic Theory, 439.
Sen, A. K. (1977). Rational fools: A critique of the behavioral foundations of economic
theory. Philosophy and Public Affairs, 317-344.
Shahani, C., Dipboye, R. L., and Gehrlein, T. M. (1993). Attractiveness bias in the in-
terview: Exploring the boundaries of an effect. Basic and Applied Social Psychology, 14(3),
317-328.
Shepperd, J., Malone, W., and Sweeny, K. (2008). Exploring causes of the self-serving
bias. Social and Personality Psychology Compass, 2(2), 895-908.
Simon, H. A., and March, J. (1976). Administrative behavior organization. New York:
Free Press.
52
4 Discussion
53