DELIBERATIVE AND NON-DELIBERATIVE PROCESSES
IN STRATEGIC INTERACTION
AN EXPERIMENTAL ANALYSIS
a dissertation
submitted to CIFREM
in partial fulfillment of the requirements
for the degree of
doctor of philosophy
in Economics and Management
Dominique Cappelletti
April 2009
ii
Advisors
Prof. Luigi Mittone
Prof. Jonathan W. Leland
Doctoral Committee
Prof. Michele Bernasconi
Prof. Lucia Piscitello
Prof. Stefano Zambelli
iii
iv
Acknowledgments
This thesis owes its existence to the help, support, and inspiration of many people.
My first thanks goes to my advisor, Prof. Luigi Mittone, who has been a source
of support and encouragement over many years. I would also like to express my
gratitude to my co-advisor, Prof. Jonathan W. Leland, for invaluable insights and
thoughtful discussions.
My deepest gratitude and appreciation goes also to Prof. Dr. Werner Güth for
providing stimulating comments and for instilling confidence in me.
I also benefited greatly from the researchers at Max Planck Institute of Economics
in Jena, especially Maria Vittoria Levati, Birendra Kumar Rai, and Christoph Vanberg, who provided me with helpful suggestions and comments.
I owe sincere gratitude to Mark Beittel, who has not only been an excellent writing
teacher, but also a helpful psychotherapist.
I am very grateful to Marco Tecilla for his constant assistance and friendship.
Warmest thanks are also due to my fellows in Cifrem, especially Stefania Bortolotti, Riccarda Moser, Mariana Doria, Ivan Soraperra, Andrea Gentili, Giulia Canzian,
and Sridhar Thapa, for making my time in the PhD programme so great and keeping
up the spirits in the difficult times.
I dearly thank my parents, my sister, and my best friend Roberta Felicetti for
their understanding, patience, and love.
Finally, a very special thanks goes to Matteo Ploner for continuous and unconditional support.
v
vi
Contents
Acknowledgments
v
1 Introduction
1
2 Extending the standard account of economic behaviour
11
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
2.2
Limitations to human behaviour
. . . . . . . . . . . . . . . . . . . .
12
2.2.1
Bounded Rationality . . . . . . . . . . . . . . . . . . . . . . .
12
2.2.2
Bounded Selfishness
. . . . . . . . . . . . . . . . . . . . . . .
14
2.2.3
Bounded Willpower . . . . . . . . . . . . . . . . . . . . . . . .
15
The influence of affect on behaviour . . . . . . . . . . . . . . . . . . .
16
2.3.1
Direct influences of affective states . . . . . . . . . . . . . . .
18
2.3.2
Indirect influences of affective states
. . . . . . . . . . . . . .
19
Dual-system theories . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
2.4.1
Dual-system theories in psychology . . . . . . . . . . . . . . .
20
2.4.2
Dual-system theories in economics . . . . . . . . . . . . . . . .
22
2.4.3
Stimulating the affective system . . . . . . . . . . . . . . . . .
25
2.4.4
Weakening the deliberative system . . . . . . . . . . . . . . .
26
2.3
2.4
3 Affective processes in the Ultimatum Game
31
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3.2
The Ultimatum Game . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.3
A dual-system approach . . . . . . . . . . . . . . . . . . . . . . . . .
35
3.4
Behavioural Predictions . . . . . . . . . . . . . . . . . . . . . . . . .
37
vii
3.5
Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
3.5.1
Participants and Procedures . . . . . . . . . . . . . . . . . . .
39
3.5.2
Treatments . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
3.5.3
Interaction Structure . . . . . . . . . . . . . . . . . . . . . . .
41
Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
3.6.1
Proposers’ Behaviour . . . . . . . . . . . . . . . . . . . . . . .
44
3.6.2
Responders’ Behaviour . . . . . . . . . . . . . . . . . . . . . .
49
3.6.3
Other Findings . . . . . . . . . . . . . . . . . . . . . . . . . .
54
3.7
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
58
3.8
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
3.6
4 Defaults and public goods provision
71
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
71
4.2
Default effects in non-strategic settings . . . . . . . . . . . . . . . . .
72
4.2.1
Why do default effects occur? . . . . . . . . . . . . . . . . . .
74
4.3
The Public Goods Game . . . . . . . . . . . . . . . . . . . . . . . . .
76
4.4
The Threshold Public Goods Game . . . . . . . . . . . . . . . . . . .
82
4.4.1
Equilibria of the game . . . . . . . . . . . . . . . . . . . . . .
83
4.5
The present study . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
4.6
Experiment I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
4.6.1
Treatments . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
4.6.2
Participants and Procedures . . . . . . . . . . . . . . . . . . .
89
4.6.3
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
4.6.4
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
Experiment II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95
4.7.1
Treatments . . . . . . . . . . . . . . . . . . . . . . . . . . . .
98
4.7.2
Participants and Procedures . . . . . . . . . . . . . . . . . . . 100
4.7.3
Results and Discussion . . . . . . . . . . . . . . . . . . . . . . 100
4.7
4.8
Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5 Defaults as recommendations in public goods provision
5.1
119
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
viii
5.2
The experimental design . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.2.1
Treatments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.2.2
Participants and Procedures . . . . . . . . . . . . . . . . . . . 125
5.3
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.4
Discussion and Concluding remarks . . . . . . . . . . . . . . . . . . . 133
6 Summary and Concluding Remarks
149
6.1
Summary and future research . . . . . . . . . . . . . . . . . . . . . . 149
6.2
Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
ix
x
Chapter 1
Introduction
Given their background as moral philosophers, classical economists considered psychological aspects of economic behaviour in their writings. Adam Smith provided
several insights on what today are called anomalies, that is behaviours that are inconsistent with mainstream economic theory. In The Theory of Moral Sentiments,
Adam Smith included aspects of human behaviour such as loss aversion, self-control,
overconfidence, altruism, and fairness (Ashraf et al., 2005).
The interest in the psychological underpinnings of economic behaviour persisted
in economic theory until the marginal revolution. The aim of developing a formal
body of theory based on mechanical laws necessitated the adoption of a more stylized
concept of utility, i.e., ordinal utility. Formal modelling sustained by an ordinal
approach made psychology disappear from economics. Thus, modern mainstream
economics is characterized by selfish and fully rational decision-makers maximizing a
given utility function, which encompasses their goals and motivations.
Accumulating evidence of anomalous behaviour, i.e., departures from what is predicted by standard economic theory, made economists start questioning the descriptive validity of the standard economic model of individual behaviour and seeking
alternatives. This dissatisfaction with mainstream economics gave rise to what Sent
1
2
Chapter 1. Introduction
(2004) defines old behavioral economics, which attempted to bring psychology back
into economics. One of the most prominent contributors to this movement is Herbert Simon. Simon’s critique of the use of full rationality assumption in standard
economic models was based on the idea that decision-makers have limitations of both
knowledge and computational capacity and, thus, can be only boundedly rational. A
direct implication of bounded rationality is a satisficing strategy, which is thought as
a substitute for the maximizing one. What Simon proposed was a procedural theory
of decision, aimed “to capture the actual process of decision as well as the substance
of the final decision itself” (Simon, 2008).
In the 1970s, new developments in cognitive psychology led to the emergence of
the behavioral decision research. The two most influential contributors were Amos
Tversky and Daniel Kahneman, who were able to bring this new branch of psychological research to the attention of economists, perhaps thanks to their formalism typical
of economic theories (Angner and Loewenstein, 200x). Most relevant to behavioural
economics are their contributions on decision-making heuristics — the idea that people use simple short-cuts when making probability judgments — and prospect-theory,
with the core ideas that preferences depend upon a reference point and that losses
loom larger than gains. These developments in cognitive psychology constitute the
roots of new behavioural economics (Sent, 2004). As Camerer and Loewenstein (2004)
emphasize, “at the core of [new] behavioral economics is the conviction that increasing the realism of the psychological underpinnings of economic analysis will improve
economics on its own terms – generating theoretical insights, making better predictions of field phenomena, and suggesting better policy” (p. 3). Compared to old
behavioural economics, new behavioral economics represents a less radical departure
from the neoclassical approach. Much of behavioral economics, for example, does not
abandon utility maximization, but adapts the content of the utility function. This
practice, defined by Güth (1995) as neoclassical repair shop, has been subject to
critiques that point toward a more truly behavioural approach (Güth, 2008).
3
Mullainathan and Thaler (2000) identify three main departures of new behavioral
economics from the standard economic approach, which involve limits to rationality,
willpower, and self-interest. Bounded rationality incorporates the idea that individuals have limited cognitive abilities that prevent them from solving difficult problems optimally. Bounded willpower reflects the fact that individuals have self-control
problems and may sometimes undertake impulsive behaviour that goes against their
long-run interest. Finally, bounded self-interest indicates that individuals take into
account the utility of other individuals when making their decisions.
Much of behavioral economics has tended to explain anomalies in cognitive terms,
while, up to recently, the role of affect in economic behaviour did not receive much
attention. Just few years ago, Camerer and Loewenstein (2004) defined the study of
affect in economic decisions as a promising new direction.
The first attempts to incorporate emotions in economic theories focused on expected emotions, i.e., emotions that are expected to occur when outcomes will be experienced. At the moment of choice expected emotions are merely cognitions about
future emotions, which do not undermine the consequentialist nature of economic
models of choice (Rick and Loewenstein, 2007). However, considering only rational
anticipation of emotions does not treat the influences of emotions on economic decisions fully, since also immediate emotions, i.e., emotions experienced at the moment of
choice, can have a great impact on choices. For example, when making decisions people can rely on the informational value carried by immediate emotions (Schwarz and
Clore, 1983; Damasio, 1994; Finucane et al., 2000) and be guided by these emotions in
selecting the information to consider in specific environmental circumstances (Ketelaar and Todd, 2000). In addition, immediate emotions influence people’s learning
processes, information retrieval from memory, and choice of decision-making strategy
(Loewenstein and Lerner, 2003). New lines of research have begun to take greater
account of immediate emotions.
4
Chapter 1. Introduction
More recent developments propose that behaviour is the product of two different systems, which activate different processes and can have conflicting motivations.
Following Loewenstein and O’Donoghue (2005)’ terminology, the deliberative system
involves deliberation, thus is maximally demanding of cognitive resources; in addition,
it is goal-oriented, forward-looking, and affect free. In contrast, the affective system
involves instincts and intuition, thus is minimally demanding of cognitive resources; in
addition, it is myopic and primarily driven by affective states. The affective system is
thought to be the default mode, which gives a preliminary response that may or may
not be adjusted by the deliberative system, depending on the specific circumstances
in which the decision-maker operates. Any factors such as time pressure, cognitive
load, and mental depletion, which tax scarce cognitive resources on which the deliberative system relies, make difficult its intervention in the decision-making process.
The final decision depends on which of the two systems prevail (Lobel and Loewenstein, 2005). This dual-system perspective, which was already present in Adam Smith
(1982) in the forms of “passions” and “impartial spectator”, has become influential
in economics, as demonstrated by the publication of dual-system models of behaviour
(Bernheim and Rangel, 2004; Benhabib and Bisin, 2005; Fudenberg and Levine, 2006)
in top economic journals such as The American Economic Review and Games and
Economic Behavior. Recent studies in the field of neuroeconomics give support to the
presence of two different systems also at a neural level (Sanfey et al., 2006). However,
this dual-system perspective has received also critiques, being deemed biologically
inaccurate (Glimcher et al., 2005) or not providing a unified theory across different
choice environments (Rustichini, 2008). Although it is probably an oversimplification,
this dual-system approach is useful for a deeper understanding of the complex human
mind (Lee et al., 2006).
This thesis experimentally investigates anomalies in strategic interaction utilizing
a dual-system approach in the most part of the analysis. The analysis will be focused
on two typical strategic situations: ultimatum bargaining and public goods provision.
5
About ultimatum bargaining, a great amount of experimental research has evidenced
systematic non-normative behaviour. The present work adds to this body of research
by investigating the affective aspects of non-normative ultimatum behaviour. About
public goods provision, the present work investigates whether default biases, which
have been identified in several non-strategic situations, are present also in strategic
situations like the public goods provision and, thus, whether these can influence
cooperative behaviour. The study will draw also on the evidence from psychological
and neurological research on the influence of affective states. As Hanoch (2002) says,
“this is one area where, by utilizing existing research and findings, economists can
economize” (p. 21).
Outline of the thesis
In this thesis I experimentally investigate non-normative behaviour in strategic interaction with economically relevant consequences. I focus on two kinds of strategic
interactions: ultimatum bargaining and public goods provision.
Chapter 2 is composed of three main parts. The first reviews the main examples of
non-normative behaviour accumulated by decades of experimental research. For the
sake of clarity of exposition, these evidences are presented following the three bounds
identified by Mullainathan and Thaler (2000), i.e., bounded rationality, bounded
willpower, and bounded self-interest. The second part offers a brief account of how
affective states can either directly or indirectly influence economic decisions. The third
part provides an overview of the dual-system theories proposed in psychology and
economics and describes the conditions under which the operations of the two systems
are favoured. This part is particularly relevant for the methodological approach used
in the present work, since it gives support to the manipulations made in three of the
four experiments presented in this thesis.
6
Chapter 1. Introduction
Chapter 3 focuses on ultimatum bargaining decisions. Literally hundreds of experimental studies consistently reported non-normative behaviour in this kind of social
interaction. Using a dual-system approach, this study adds to this body of research by
investigating how affect influences non-normative ultimatum bargaining behaviour.
The Ultimatum Game is a simple two-party game in which one party, the proposer,
makes an offer to the other party, the responder, about how to split a sum of money
between them. If the offer is accepted, the sum of money is split as agreed. If
the offer is rejected, both players earn nothing. The standard economic prediction
is that responders should accept all non-zero offers because, as maximizing agents,
they prefer some money to none. Knowing this, proposers should offer a very small
amount and keep almost all of the money for themselves. Instead, a great deal
of empirical evidence indicates the existence of systematic non-normative behaviour
both on proposers and on responders side: proposers offer high amounts, often close
to equal splits, and responders reject relatively low offers. A series of studies support
the idea that rejections can be attributed to an emotional reaction to being treated
unfairly. However, these studies have investigated the role of affect mainly as a “hot”
reaction to a specific unfair real offer. The experimental study reported in this chapter
adds to this body of research by showing that affect activates underlying processes
that are able to influence behaviour also in “cold” environments.
Another aspect of originality resides in the conceptual and methodological approach used. We adopt a dual-system approach as described above. The deliberative
system is taxed through time pressure and cognitive load to give the opportunity to
the affective system to prevail. A simple and clean kind of interaction such as the one
involved in the Ultimatum Game offers an ideal opportunity to test the effectiveness
of the manipulation used at the experimental level to inhibit the deliberative system
in favour of the affective one.
A final primary contribution made by this study is the investigation of how affect
influences proposers’ behaviour. While extensive research has been conducted on the
7
affective aspects of responder behaviour, the affective aspects of proposer behaviour
have been almost entirely disregarded. The main reason for this neglect seems to be
of methodological nature. While, in the case of responders, affect resulting from the
receipt of an offer can be examined through self-reported and physiological measures,
or through the measurement of activity in the brain areas associated with it, in the
case of proposers there is not an event or a stimulus to react to. The dual-system
approach adopted here allows us to shed light on the proposer side.
Chapter 4 focuses on contribution behaviour in Public Goods Game. An impressive amount of experimental evidence shows systematic non-normative behaviour in
this kind of social interaction. In a standard Public Goods Game, subjects have to
decide how much to contribute to a common project, whose income is equally divided
among all subjects regardless of their personal contribution. A subject motivated only
by her own monetary payoff should contribute nothing to the project and free ride
on the efforts of others. What happens, instead, is that subjects contribute positive
amounts, lying on average between 40 and 60% of their initial endowment. Extensive
research has been conducted to understand what factors improve cooperation. To
name just a few, these factors include the opportunity to communicate, the use of
positive frames, and the opportunity to punish.
The present study adds to this body of research by investigating whether the
presence of a default contribution has an impact on contribution choices. A default
contribution is a predefined amount that subjects automatically contribute unless
they specify a different amount, and we suspect that they may have an impact on
contribution choices due to the default bias that decision-makers very often display.
Default bias refers to an exaggerated preference for the default option and has been
found in many non-strategic domains, such as retirement savings, organ donation,
and Internet privacy policy adoption, to name just a few. However, to the best of our
knowledge, no laboratory, natural nor field studies have explored the role of defaults in
domains of strategic interaction. The present study is a first step towards redressing
8
Chapter 1. Introduction
this lacuna.
Among other explanations, it has been proposed that limited cognitive resources
may account for default effects. Changing the default is cognitively demanding, while
keeping it is not. Therefore, sticking with the default represents a good solution
when resources are constrained, but also when they are unconstrained and people
attempt to economize on them. To test whether this explanation is relevant in this
kind of contexts, we tax cognitive resources through cognitive load and see whether
the incidence of default bias increases. In a dual-system perspective, taxing cognitive resources makes the intervention of the deliberative system difficult, leading to
greater reliance on the affective system. The affective system relies on heuristics, and
retaining the default option is a simple heuristic. In addition, it is sensitive to salient
contextual features of the decision-making situation, such as the presence of a default
option.
Chapter 5 extends the research of default bias in public goods provision by testing a different hypothesis for the occurrence of default effects. Specifically, it has
been proposed that the information conveyed by the default option can account for
the default bias: people may interpret the default option as a recommendation from
those who set the default. In this case, default effects may not be the product of
the impulsive affective system, but instead may constitute the result of a deliberate
decision to follow the recommendation. Previous research (Croson and Marks, 2001)
has explored the effect of recommended contributions in the provision of public goods.
However, in this research the recommended contributions were not the default option and were given by the experimenter. The experiment reported in this chapter
is designed to specifically test for default effects, thus the recommended contribution
assumes the form of default contribution. In addition, it is set by some participants,
avoiding this way potential demand effects and conflicts of interest. Default options
can have an impact both on contribution choices, i.e., influencing people’s preferences, and on beliefs, i.e., influencing expectations about others’ contributions. In
9
the experiment reported here beliefs were elicited in order to shed light on processes
underlying default effects.
Chapter 6 summarizes the results of the experimental chapters, identifies lines
of future research, and offers some brief concluding thoughts about the importance
of considering affective processes as a determinant of economic behaviour and the
usefulness of a dual-system approach for investigating economic behaviour.
10
Chapter 2
Extending the standard account of
economic behaviour
2.1
Introduction
Through years, several contributions from the psychological literature have entered
the growing body of behavioral economics (for an extensive review see, Camerer and
Loewenstein, 2004). These contributions have progressively relaxed the strict assumptions of standard models and have enriched the characterization of economic agents.
This chapter presents a brief overview of the principal behavioural theories that have
tried to accommodate the large amount of anomalies emerged both in the laboratory and in the field. Following an early classification provided by Mullainathan and
Thaler (2000), they can be organized under three main areas, according to the standard assumption they revise: bounded rationality, bounded selfishness, and bounded
willpower.
From this brief overview, it emerges that emphasis has been put on cognitive
aspects, while affective aspects have not been properly acknowledged. The second
part of this chapter illustrates the influences that affect can have on behaviour, and
the third part overviews more recently developed theories that propose behaviour as
the product of two different systems and represent a more fundamental challenge to
11
12
Chapter 2. Extending the standard account of economic behaviour
the standard account of economic behaviour.
2.2
2.2.1
Limitations to human behaviour
Bounded Rationality
Bounded rationality incorporates the idea that individuals have limited cognitive abilities that prevent them from solving difficult problems optimally. The seminal work
of Simon (1955) has open the field to revisions of the standard economic assumptions of perfect rationality. The extensive survey by Starmer (2000) focuses on such
revisions in the domain of risk and uncertainty. The reference model for decisions
involving probabilistic outcomes is represented by Expected Utility Theory (EUT).
The theory delivers an elegant and parsimonious description of choices under risk.
However, for the theory to be a valid descriptive tool, observed choices must respect
a set of axioms (i.e., completeness, transitivity, independence, and continuity). Over
years, several pieces of evidence about violations of such axioms have been collected
(e.g., Allais’ Paradox). The work of Tversky and Kahneman (1974) illustrates some
decisional shortcuts, called heuristics, which may lead to systematic violations of the
predictions of EUT. A The representativeness heuristic exacerbates the matching
between some features of an event and the stereotyped representation of a class of
events. The heuristic is likely to produce biases because the likelihood that an event
actually belongs to a certain class of events is overestimated when is perceived as
very representative of that class (e.g., gambler’s fallacy). According to the availability heuristics, the estimated likelihood of an event is function of the easiness of recall
of previous analogous events. Thus, the expected frequency of vivid or recent events
is likely to be overestimated. Finally, humans seem to apply a two-step procedure
in the evaluation of unknown events. The authors label this heuristic anchoring and
adjustment. The first step is constituted by the identification of a reference point for
2.2 Limitations to human behaviour
13
the evaluation; the second step is represented by an adjustment of the initial evaluation. However, evidence suggests that the adjustment is only partial and a strong
path dependency in choices is likely to be observed.
The evidence of biases and mistakes in risky decision making has lead to the
formulation of an alternative theory to EUT. The theory goes under the name of
Prospect Theory (Kahneman and Tversky, 1979). The theory aims at providing a
general descriptive framework for decisions under uncertainty. The main constituents
of the theory are the value function and the probability weighting function. The assessment of a risky event (i.e., a prospect) is the result of the interaction between the
two evaluation functions. In more details, the former assigns a psychological value
to the payoffs associated to an outcome, while the latter is employed to evaluate the
likelihood of the outcome. Three main features characterize the value function: it
is defined over variations from a reference point, it is concave for gains and convex
for losses, and it is steeper in the domain of losses than in the domain of gains (i.e.,
loss aversion). The probability weighting function produces a transformation of objective probabilities that tends to overestimate small probabilities and to produce
sub-additivity in the sum of the probabilities of an event and of its complement.
Prospect Theory has shown to explain several anomalies observed in laboratory experiments. Among the most studied anomalies there are the Endowment Effect and
the Status Quo Bias (Kahneman et al., 1991). The former refers to the tendency to
ask an higher price to give up a good than what is offered to purchase it. The latter
refers to the fact that the current state is systematically preferred to an alternative
one. Both these biases seem to be largely explained by the concept of loss aversion
embedded in the Prospect theory. While the largest part of evidence in support of
Prospect Theory comes from laboratory experiments, Camerer (2000) reviews several
successful applications of the theory to real-life phenomena.
14
Chapter 2. Extending the standard account of economic behaviour
2.2.2
Bounded Selfishness
Bounded selfishness incorporates the idea that individuals take into account the utility of other individuals when making their decisions. One of the first contributions
considering bounded self-interest is offered by Becker (1974). The work of Becker
introduces the concept of social income as an attempt to encompass others’ welfare
into one’s own welfare. What Backer proposes is a model of pure altruism, since
others’ welfare directly enters the utility of the decision-maker. Since then, several
contributions have dealt with other-regarding preferences, and distinct types of social
preferences have been investigated. Akerlof (1982) has theoretically investigated the
relevance of reciprocity concerns for labor contracts. Laboratory experiments have
shown that, in a principal-agent setting similar to that conceived by Akerlof (1982),
reciprocity-based contracts can be sustained and deliver a Pareto-superior outcome
(Fehr et al., 1998).
Limits to selfishness may explain behavior in prisoner’s dilemma like strategic
interactions. In situations of this kind, the pure strategy Nash equilibrium of the game
is a Pareto-inferior solution to alternative solutions based on cooperation. Ledyard
(1995) presents a survey of several studies about Public Goods Games and identifies
some stylized facts. A thorough discussion about the Public Goods Game and the
factors that have been found to improve cooperation is provided in Chapter 4.
Limits to selfishness has been also proposed as an explanation for non-normative
behaviour consistently reported in studies on the Ultimatum Game (Güth et al.,
1982) and the Dictator Game (Kahneman et al., 1986). The experimental evidence
collected in these two games has stimulated the development of social preferences
models. The models of Fehr and Schmidt (1999) and Bolton and Ockenfels (2000)
both focus on preferences for equity. Subjects are assumed to care for the distribution
of payoffs among players. In particular, a subject incurs a psychological costs both
when being better off than the others and when being worse off than the others.
2.2 Limitations to human behaviour
15
Another class of models extends the framework provided by psychological game theory
(Geanakoplos et al., 1989) and focuses on intentions behind subjects’ actions. As an
example, Rabin (1993) presents a theoretical account of reciprocity-driven fairness.
In a recent contribution, Charness and Dufwenberg (2006) show that decision makers
are reluctant to let down others’ expectations in a trust-based game.
2.2.3
Bounded Willpower
A direct implication of the standard economic approach is the intertemporal consistency of choices. The assumption of stable and well-defined intertemporal preferences
together with that of perfect rationality implies a full coherence between planning and
implementation of actions. However, actual behavior of humans seems to be better
described as an internal agency problem between two components, the planner and
the doer (Thaler and Shefrin, 1981). The former has a long-time horizon and resembles the standard economic agent, while the latter is myopic and subject to temptations. This dyad in intertemporal decision-making may lead to self-control problems
that are of main interest for economics (O’Donoghue and Rabin, 2000). Typical examples of self-control problems that have attracted the attention of economists are
intertemporal transfers of wealth (e.g., retirement saving) or health-related choices
(e.g., smoking).
The survey of Frederick et al. (2002) provides a critical assessment of the inadequateness of the standard discounted utility model, employing an exponential discount
function, in describing intertemporal choices involving self-control. A descriptively
appealing alternative to the standard discounted utility model is represented by the
class of hyperbolic discount models. The basic assumption of models falling into this
class is that the discount rate is not constant over time, like in the exponential model,
but tends to decrease over time. This implies a strong present-bias in consumption
choices that may lead to intertemporal inconsistencies (Laibson, 1997).
The existence of self-control problems and their negative impact on the overall
16
Chapter 2. Extending the standard account of economic behaviour
welfare of individuals has interesting policy implications. Potentially, detrimental
consequences of self-control problems can be avoided by the introduction of some
commitment devices (Benartzi and Thaler, 2004). The behavioral economics literature has progressively moved from a descriptive approach to a more normative one.
Policy interventions, which acknowledge the relevance of human limitations and suggest mechanisms to alleviate their negative impact in terms of welfare, have been
grouped under the conceptual framework of libertarian paternalism (Thaler and Sunstein, 2003). Guiding principles of this approach are the preservation of individual
freedom of choice and the design of choice mechanisms that will help individual make
better choices. In Thaler and Sunstein (2003), several real-life applications of policies
inspired by libertarian-paternalism principles are surveyed. Among others, the authors devote attention to the beneficial impact of default-choices in retirement saving
decisions.
2.3
The influence of affect on behaviour
As it can be seen from the brief overview of the limits to rationality, self-interest,
and self-control and the models that attempt to accommodate them, non-normative
behaviour has been largely explained in cognitive terms, while the role of affect has
been marginalized. This section illustrates the influences that affect can have on
behaviour and, thus, the importance to consider it in an attempt to better understand
economic behaviour.
A growing amount of research has shown that affective factors, which include
emotions, mood, drive states such as hunger and sexual desire, and motivational
states such as physical pain and drug craving (Camerer et al., 2005), can greatly
influence economic behaviour.
Affective states can enter deliberative processes as factors that influence utility.
This is the way some economists have attempted to incorporate some measure of the
2.3 The influence of affect on behaviour
17
emotional dimension into their theories of decision-making. A common feature of
the models proposed by these scholars is that emotions that have been taken into
account are anticipated emotions, which are not experienced at the moment of choice
but are expected to occur when outcomes are experienced. For example, Loomes and
Sugden (1982, 1986) and Bell (1982, 1985) proposed theories focused on two counterfactual emotions, namely disappointment and regret, which result from unfavourable
comparisons between alternative outcomes of the same option and between outcomes
of alternative options respectively. Subjects rationally anticipate how they will feel
when they will experience the outcome of their decisions and take these emotions into
account at the moment of choice. The core issue of these theories is that individuals
are motivated to avoid these negative emotions. As a consequence, subjects averse
to regret or disappointment make decisions in a way that minimizes the likelihood
of experiencing them. Other models of decision-making including anticipated emotions have focused on other types of emotions, such as envy (Kirchsteiger, 1994) and
anxiety (Caplin and Leahy, 2001).
An implicit assumption of these models is that people are able to correctly predict
their future affective states. However, there is plenty of evidence showing that, on
the contrary, people systematically mispredict them. (see Loewenstein and Schkade,
1999, for an extensive discussion). Loewenstein and Schkade (1999) identify three
major mechanisms that explain why people are not very good at predicting future
affective states. First, people may have wrong beliefs about the determinants or
about the impact of these determinants on specific affective states. Second, when
making predictions, people focus on some aspects that may become less salient when
experiencing a particular affective state. Finally, people have difficulties in predicting
how they would feel in a state that is different from their current one, so that predicted
future affective states tend to disproportionately reflect the current affective states.
Although including rationally anticipated affective states represents an improvement in terms of descriptive validity, much remains out of conventional theories of
18
Chapter 2. Extending the standard account of economic behaviour
economic decision-making. Non-deliberative factors that are at work at the time of
decision-making and that often occur beyond consciousness have been found to influence decision behaviour both directly and indirectly by impacting on judgments of
expected consequences and related expected emotions, as well as on the nature and
the depth of information processing (Loewenstein and Lerner, 2003).
2.3.1
Direct influences of affective states
The nature of direct influences of affective states on behaviour, Loewenstein and
Lerner (2003) argue, depends on the intensity and on the qualitative character of the
affective states. At low and moderate levels of intensity, affective states can have
an informational value that people take into account when deciding. The feelingsas-information hypothesis (Schwarz and Clore, 1983, 2003) and the affect heuristic
(Finucane et al., 2000) are among the theories that have suggested the consultative
function of affective states. Both these theories posit that people consult their affective feelings when making decisions. At higher levels of intensity, affective states exert
a stronger influence on behaviour, till they overcome deliberation and take control
over behaviour at extreme levels (Loewenstein, 1996).
The influence of affective states on behaviour depends also on their specific type,
since different affective states involve different types of action tendencies, i.e., impulses
to act in a certain way (Frijda, 1986). For example, envy may give rise to an urge
to destroy, and anger may have the action tendency to punish. Action tendencies
may be translated into real actions or be modified through self-regulatory processes,
which can occur either at conscious or unconscious level(Elster, 1998). The impact
of affective states on decisions can be greatly reduced when some actions are taken
to alleviate them. For example, Xiao and Houser (2005) showed that the behavioural
consequences of negative emotions can be mitigated by allowing subjects to discharge
their emotions through emotional expression: responders who were allowed to express
their emotions directly to proposers were significantly less likely to reject unfair offers.
2.3 The influence of affect on behaviour
19
Affective states can extend their influence beyond the situation that generated
them. Psychologists use the term integral affect to refer to situation-induced affective
states and the term incidental affect to refer to affective states that are independent
of the situation at hand.
2.3.2
Indirect influences of affective states
A large body of research has focused on the indirect impact that affective states
can have on decision-making. From this body of research, the following three main
findings have emerged. First, feelings affect subjects’ learning processes by making
individuals focus attention on aspects of the situation that are congruent with their
mood. Subjects in a negative affective state were found to acquire more negative
than positive information to which they have been exposed (Bower and Cohen, 1982;
Blaney, 1986).
Second, feelings affect what information is retrieved from memory. Tversky and
Kahneman (1973, 1974) have suggested that the ideas that come to mind first or
most easily may influence judgment. People in a positive affective state were found
to be more likely to think about positive possibilities and to be optimistic in their
decisions (Isen et al., 1978; Isen and Shalker, 1982). Wright and Bower (1992) found
that happy people are optimistic, in the sense that they report higher probabilities
for positive events and lower probabilities for negative events. The inverse pattern
has been found for subjects in a negative affective state.
Third, feelings influence the choice of decision-making strategy. Subjects in a
positive affective state, when compared with subjects in a negative affective state,
tend to reduce the complexity of the decision task through the choice of a simpler
process of information retrieval. They disregard irrelevant information, consider fewer
dimensions, recheck less information, and take significantly less time to make their
choices (Isen and Means, 1983). This kind of processing could either facilitate or
impair subjects’ performance, depending on the circumstances. From a normative
20
Chapter 2. Extending the standard account of economic behaviour
point of view, the authors suggest that when feedback on a task is provided or when a
task is extremely important, people in a positive affective state would not be expected
to engage in the type of processing described above.
2.4
2.4.1
Dual-system theories
Dual-system theories in psychology
The idea that human thinking and decision-making are governed by two different
but interacting systems has been increasingly recognized as influential in psychology.
These two systems have been variously identified as rational and experiential systems
(Epstein, 1994), associative and rule-based systems (Sloman, 1996), implicit and explicit systems (Evans and Over, 1996), hot and cool systems (Metcalfe and Mischel,
1999), System 1 and System 2 (Stanovich, 1999; Kahneman and Frederick, 2002), reflexive and reflective systems (Lieberman, 2003), and impulsive and reflective systems
(Strack and Deutsch, 2004; Strack et al., 2006).
System 1 (to use Stanovich’s more generic terminology) is evolutionarily old, and
its processes are generally described as fast, automatic, associative in nature, emotionally charged, and minimally demanding of cognitive resources. System 1 is governed
by habits and emotions and includes intuition and innately programmed instinctive
elements; therefore, its responses are difficult to control or modify. In contrast, System 2 is recent in evolutionary terms, and its processes are generally described as
slow, deliberately controlled, analytical, affect free, and maximally demanding of
cognitive resources. System 2 is potentially rule-governed and allows people to make
hypothetical thinking and abstract reasoning (Evans, 2003; Kahneman, 2003a).
Psychological research has accumulated evidence for dual process in reasoning.
An example is provided by the Wason selection task (Wason, 1968), which is a reasoning problem largely investigated in psychological literature. In this task, subjects
are presented with a conditional statement and have to ascertain whether or not the
2.4 Dual-system theories
21
statement is being violated. In the abstract version of the task,1 in which the conditional statement is expressed using letters and numbers, only about 10% of subjects
answer correctly. Stanovich and West (1998) found that giving the correct answer
is positively correlated to cognitive ability,2 which is related to System 2 function
(Stanovich, 1999). The modal incorrect answer suggests the influence of a System 1
heuristic, termed by Evans (1972) as matching bias, i.e., the tendency to give relevance
to the values mentioned in the conditional statement. When the task is contextualised
and expressed in realistic terms,3 the rate of correct answers increases dramatically.
The concrete inputs may facilitate performance by evoking prior knowledge and beliefs, related to System 1 processes (Evans, 1998). Stanovich and West (1998) did
not find any significant relation between correct answers and cognitive ability in the
enriched task.
The two systems operate in parallel and compete for control of overt behaviour.
However, while System 1 is the default mode, in the sense that it is always engaged,
System 2 may be disengaged. Typically, System 1 leads to a preliminary response,
which may or may not be adjusted by System 2 (Gilbert et al., 1988). Stimulating
functions of System 2 may increase the ability of System 2 to override or inhibit the
responses of System 1.
Recently, Kogler and Kühberger (2007) argued that the diversification bias, i.e.,
1
In the abstract version of the task, subjects are shown four cards, each with a letter on one
side and a number on the other. The top sides of the cards show “A”, “D”, “3”, and “7”, and the
conditional statement is “If there is a D on one side of any card, then there is a 3 on its other side”.
Subjects have to select those cards that need to be turned over to determine whether or not the
statement is being violated. The correct answer is “D” and “7”, while the modal incorrect answers
are “D” or “D” and “3”.
2
Cognitive ability was proxied by the SAT score, i.e., the score obtained in a test for college
admission in the U.S.
3
An example of enriched selection task is the Drinking-age Problem (Griggs and Cox, 1982).
Subjects are asked to imagine to be a police officer on duty, checking whether the drinking laws are
in effect in a local bar. The rule is “If a person is drinking beer, then the person must be over 21 years
of age”. Subjects are shown four cards, each with a person’s age on one side and what the person is
drinking on the other side. The top sides of the cards show “drinking beer”, “drinking Coke”, “22
years of age”, and “16 years of age”. Similarly to the abstract version of the task, subjects have to
select those cards that need to be turned over to determine whether or not the statement is being
violated.The correct answer is “drinking beer” and “16 years of age”.
22
Chapter 2. Extending the standard account of economic behaviour
the tendency to diversify choices in identical choice situations, is caused by System 1,
which fails to be corrected by System 2. In their study, the authors employ the card
task of Rubinstein (2002). In this task, five cards are randomly drawn from a deck
of 100 coloured cards composed of 36 Green, 25 Blue, 22 Yellow, and 17 Red cards.
The five cards are placed into five separate envelopes, and subjects have to predict
the colour of the card in each envelope. The authors argue that the widely observed
strategy of choosing a mixture of colours in proportions similar to the sample, i.e.,
the probability matching, is a product of System 1, since similarity is an automatically
assessed attribute. The authors found that, when the corrective functions of System 2
were encouraged through some procedures suggested by Kahneman (2003b) – namely
increasing the vigilance of the monitoring activities and providing strong cues to
the relevant rules4 –, the probability-maximizing strategy was employed significantly
more often at the cost of the probability-matching strategy.
2.4.2
Dual-system theories in economics
Recently, a number of dual-process models have been proposed also in economics,
with applications to intertemporal choice (Thaler and Shefrin, 1981; Bernheim and
Rangel, 2004; Loewenstein and O’Donoghue, 2005; Benhabib and Bisin, 2005; Fudenberg and Levine, 2006), risk preferences (Loewenstein and O’Donoghue, 2005),
labour supply (Goette and Huffman, 2007), and social preferences (Loewenstein and
O’Donoghue, 2005). For example, Benhabib and Bisin (2005) propose a new model
of consumption-saving decisions, in which the decision-makers have a consumptionsaving plan that requires the exertion of self-control to be implemented. The agent
can invoke either automatic or controlled processes: automatic processes are susceptible to impulses or temptations, which lead to immediate consumption, while
controlled processes, which are unaffected by temptations, encourage the agent to
4
Specifically, the task was described as a statistical test rather than a lottery, and it was made
clear that the aim of the test was to measure the statistical competence of the subjects. In addition,
there was no time pressure, and subjects were invited to carefully revise their decisions.
2.4 Dual-system theories
23
implement the determined consumption-saving plan. Controlled processes allow the
agent to actively maintain attention to the plan and to inhibit automatic responses.
Agent’s behaviour is determined by the processes that are relatively more activated.
An example in which automatic processes are likely to prevail is when the agent has
lower cognitive control abilities.
In Loewenstein and O’Donoghue (2005), human behaviour is modelled as the
outcome of an interaction between two systems: a deliberative and an affective
system, which may have different motives. The deliberative system can over-ride
the affective one by exerting cognitive effort, referred to as willpower. More precisely, willpower represents a scarce internal resource – subject to depletion after
its repeated use – that attempts to dominate affectively-motivated behaviours when
they conflict with cognitively-deliberated motives (Loewenstein, 2000b). The cost
of exerting willpower depends on its current stock and all the factors that weaken
the deliberative system. In formal terms, as a result of the decision-making process, the chosen option x will be the one which maximises the following utility
function: V (x, s) = U (x, c(s), a(s)) − h(W, σ) ∗ M (xA , a(s)) − M (x, a(s)) where
U (x, c(s), a(s)) identifies the utility function of the deliberative system; c(s) represents the cognitive states induced by the environmental stimuli s; a(s) represents the
affective states induced by the environmental stimuli s; h(W, σ) is the cost to the deliberative system of exerting willpower, which depends on the current stock of willpower
W and on other factors σ that weaken the deliberative system. M (x, a(s)) identifies
the motivational function of the affective system, while xA ≡ arg maxx∈X M (x, a(s))
represents the affective optimum.
All these models view economic behaviour as determined by the interaction between two different systems, an affective and a deliberative system (Loewenstein and
O’Donoghue, 2005). The affective system is the counterpart of System 1 in psychological models and is considered to be fast,5 myopic, activated by environmental
5
Rubinstein (2007) convincingly demonstrates that most biases in the sense of deviations from
24
Chapter 2. Extending the standard account of economic behaviour
stimuli, and primarily driven by affective states. The deliberative system is analogous
to System 2 in psychological models and is generally described as goal-oriented and
forward-looking.
This dual-system view is also supported at a neural level. Recent neuroimaging
evidence indicates that affective and deliberative processes share some common neural
components, but activate distinct neural areas. Deliberative processes are associated
with the outer part of the brain (neocortex), in particular with anterior and dorsolater
regions of prefrontal cortex, while affective processes are associated with the inner
part of the brain (the limbic system), which includes anterior and posterior cingulate
cortex, insular cortex, orbitofrontal cortex, and the amygdala (Dolan, 2002; Cohen,
2005; Sanfey et al., 2006).
What are the implications of these dual-process models of decision-making and
the brain? Behaviour is seen as determined by the interaction between affective and
deliberative processes which evaluate the same circumstances differently. For example, Knutson et al. (2005) showed that, unlike the deliberative system, the affective
system more likely disregards probabilities. Using functional magnetic resonance
imaging, they found that activity in subcortical regions (in particular, the nucleus accumbens) was proportional to the reward magnitude, while the activation of cortical
regions (in particular, the mesial prefrontal cortex) was related to both magnitude
of gain and probability. Different evaluations can generate conflicting motivations,
and behaviour depends on which kind of processes is prevailing. For example, in a
neuroimaging study of intertemporal choice, McClure et al. (2004) found that the limbic system is particularly activated when the decision involves an immediate reward,
while neocortical regions associated with deliberative processes remain unvaryingly
activated in all decisions. Moreover, choices are predicted by the relative level of
activation of the two systems.
non-opportunistic or non-expected utility maximizing choices rely on fast decision-making.
2.4 Dual-system theories
25
Affective reactions are likely to have a larger influence on decision making than deliberative reactions when the affective system is stimulated or the deliberative system
is weakened.
2.4.3
Stimulating the affective system
The affective system can be stimulated in different ways. For example, affective
processes are particularly sensitive to temporally, spatially, and socially near environmental stimuli (Loewenstein, 1996). For example, in a study examining intertemporal
choice in children, Mischel et al. (1989) found that children were more impatient when
either the smaller immediate reward or the larger delayed reward was placed in front
of them. In an experiment investigating the effect of affective reactions on consumers’
preferences, Shiv and Fedorikhin (1999) asked subjects to choose between an option
that was superior on the affective dimension but inferior on the cognitive dimension
(chocolate cake) and another option that was superior on the cognitive dimension
but inferior on the affective dimension (fruit salad). Under conditions favouring the
affective control of behaviour, i.e., under cognitive load, subjects were more likely to
choose the chocolate cake when presented with the real cake then when presented
with a photograph of it.
Many studies have tried to stimulate the affective system through affective state
induction techniques (for a meta-analysis, see Westermann et al., 1996). These techniques includes procedures such as giving small presents, presenting stories, pictures,
and movies, giving subjects fake feedback about their performance on a test, and
asking subjects to provide a detailed report of a life-event. For example, Kirchsteiger
et al. (2006) manipulated the affective state of second movers in a gift-exchange
game by presenting one group with a funny movie and another with a depressing
one. They found that a positive affective state is associated with more generosity,
while a negative affective state is associated with more reciprocity. In an experiment
investigating the impact of specific emotions on the endowment effect, Lerner et al.
26
Chapter 2. Extending the standard account of economic behaviour
(2004) presented subjects with a sad, disgusting, or neutral clip, depending on the
experimental condition, and asked to project themselves in the situation depicted
in the clip. The results show that, compared to subjects in the neutral condition,
subjects in the disgust condition set lower buying and selling prices in a way that
eliminates the endowment effect, while subjects in the sadness condition set higher
buying prices and lower selling prices in a way that reverses the endowment effect.6 In
a study investigating the influence of positive affect on risk taking, Isen and Patrick
(1983) induced positive affect by giving subjects an unexpected gift certificate as a
token of appreciation for their participation in the experiment. The authors found
that, compared to subjects in the neutral affect condition, subjects in a positive affective state show more risk-aversion when facing high stakes, while they show more
risk-seeking in the presence of low stakes.7
2.4.4
Weakening the deliberative system
The deliberative system works slowly and relies on scarce processing resources. Therefore, factors such as time pressure, mental depletion, and cognitive load will tend to
weaken deliberative processing in decision-making (Lobel and Loewenstein, 2005). It
has to be noted that the deliberative system plays a role in self-regulation, including
emotion regulation (e.g., Ochsner and Gross, 2005). Therefore, when it is weak, it
is not only less involved in the decision-making process, but also loses control of the
affective system.
6
Lerner et al. (2004) argue that the pattern of choices observed in the disgust condition is consistent with an “expel” goal: since a behavioural component of disgust is taking distance from some
object, event, or situation (Rozin et al., 2000), subjects tend to get rid of the owned object by
lowering the selling prices and also to avoid acquiring a new one by lowering the buying prices.
In contrast, the pattern of choices observed in the sadness condition is consistent with a “change
circumstances” goal: sadness creates a general orientation toward seeking change, and both buying
and selling represent an opportunity to attain the goal.
7
The observed behavioural pattern is consistent with a state-maintenance hypothesis: subjects
are motivated to preserve their positive affective state and thus tend to avoid large losses and benefit
from the gain without risking too much (Mano, 1994).
2.4 Dual-system theories
27
Time pressure
Time pressure impacts on the role of the deliberative system in decision-making
mainly in two ways. First, since deliberation takes time, a shortage of time tends
to reduce deliberative processing. Second, when time is constrained, it needs to
be monitored. This activity absorbs a part of central processing resources (Zakay,
1993), crowding out deliberation and self-regulation. In addition, time pressure has
an impact on the affective system, by increasing the level of arousal. For example,
in a study investigating the effect of time pressure on risk-taking decision-making,
Maule et al. (2000) found that, when choosing under a time constraint, subjects were
more anxious and more energetic.8 Although the arousal component may have an
important role, most empirical studies have focused on the cognitive-process component. Many systematic changes of cognitive processes under time pressure have been
reported in the literature, such as an increased selectivity of input of information,
and an increased weighting of more important attributes and of negative information
(Edland and Svenson, 1993; Zakay, 1993).
Mental depletion
A variety of factors, such as stress, exhaustion, sleep deprivation, and decision fatigue contribute to the depletion of mental resources. When these factors are at
work, less resources are available for decision-making and self-regulation. Previous
studies showed that performance in a task requiring mental resources, in particular
working memory, decreases with time spent on the task (Dewitte et al., 2003); prior
exertion of self-regulation impairs cognitive performance on a reasoning task (Schmeichel et al., 2003); making a series of choices in an effortful, deliberate manner impairs
the subsequent exertion of self-control (Vohs et al., 2008) and increases attraction to
affective aspects of products (Bruyneel et al., 2006). Some studies also showed that
8
However, Maule et al. (2000) are cautious about the effect of energy: unlike for anxiety, different
time pressure conditions, such as prolonged time pressure, could have even an opposite effect on
energy.
28
Chapter 2. Extending the standard account of economic behaviour
biological processes, such as the circadian rhythm (being a “morning” or a “night”
person), influence the effectiveness of mental resources. Specifically, mental capacity,
and thus also the ability to self-regulate, is diminished at times that diverge from
one’s own optimum circadian time. In a study about stereotypic biases in judgment,
Bodenhausen (1990) found that people were more susceptible to a judgment bias such
as the conjunction fallacy 9 when making probability judgments at a nonoptimal time
of day, i.e., in the evening for morning people and in the morning for night people.
The conjunction fallacy is a product of an intuitive judgment of System 1, which fail
to be corrected by the intervention of System 2 (Kahneman, 2003b). An example of
the influence of the circadian rhythm on economic behaviour is provided by (Gonzalez and Loewenstein, 2004). In a repeated trust game (centipede game10 ), the two
authors found that subjects who were off their circadian time cooperated less and,
consequently, made less money than subjects who were on their circadian time. This
result was due to the fact that “off peak” subjects responded more aggressively to
their opponents’ uncooperative moves and continued to play non-cooperatively also
with new opponents. Although emotions were not measured, the authors attributed
the observed behaviour to the reduced ability of “off peak” subjects to regulate their
affective responses.
Cognitive load
Cognitive load refers to the amount of mental activity imposed on cognitive resources,
in particular working memory, at a point in time. It is usually manipulated through a
dual-task procedure in which subjects have to complete another task while performing the task of primary interest. Frequently used secondary tasks include memory
9
The conjunction fallacy (Tversky and Kahneman, 1983) occurs when two events occurring together are judged as more probable than either one occurring alone.
10
The centipede game (Rosenthal, 1981) is a finite move extensive form game in which two players
alternately get a chance to either take a larger share of an increasing pile of money and terminate
the game or pass and continue the game, producing a social gain. The payoffs are structured so that
passing decreases one’s payoff if the opponent takes the larger share and terminate the game on the
next move.
2.4 Dual-system theories
29
tasks, tracking task, and vocal or manual reaction-time tasks. Memory tasks are
the most commonly used. Subjects are required to memorize strings of numbers
(Shiv and Fedorikhin, 1999; Roch et al., 2000; Hinson et al., 2003; Benjamin et al.,
2006; Cornelissen et al., 2007), lists of words (Drolet and Luce, 2004), or series of
slides (Ward and Mann, 2000) and to reproduce them at the end of the experiment.
Other techniques are more elaborated. For example, Skitka et al. (2002) manipulated cognitive load by having subjects listening a tape tones at varying notes and
counting the number of tones sounded before a note change. Cognitive load tends to
weaken deliberative processing in decision-making because scarce resources must be
allocated to different simultaneous tasks. In addition, less resources are available for
self-regulation. For example, Ward and Mann (2000) showed that cognitive load has
a disinhibiting effect on normally restrained eaters: these people relaxed more their
restraining food consumption rules under high cognitive load than under low cognitive
load. Shiv and Fedorikhin (1999) found that, compared to those under low cognitive
load, subjects under high cognitive load were more likely to choose an affect-laden
option (a chocolate cake) over a cognitive-laden option (a fruit salad). Other studies investigated the impact of cognitive load on economic behaviour. These studies
showed that people under higher cognitive load discount delayed monetary rewards
at higher rates (Hinson et al., 2003; Benjamin et al., 2006), offer more in the Dictator
Game (Cornelissen et al., 2007), and are more likely to request an equal amount from
a common resource pool (Roch et al., 2000).
30
Chapter 3
Affective processes in the
Ultimatum Game1
3.1
Introduction
In economics it is mostly ignored how emotions affect decision-making. However,
recent evidence suggests an interplay between emotion and deliberation in economic
decision making (e.g., Sanfey et al., 2003; McClure et al., 2004; Bechara and Damasio,
2005). Here we investigate how affective processes influence proposers’ and responders’ behaviour in the Ultimatum Game. We address this question by using a dualsystem approach to decision-making, which assumes that cognitive resources, which
are scarce, are needed for both implementing deliberative processes and overriding
affective processes (e.g., Bernheim and Rangel, 2004; Loewenstein and O’Donoghue,
2005; Benhabib and Bisin, 2005; Fudenberg and Levine, 2006).
To enhance the influence of affective processes on behaviour, we tax cognitive
resources through time pressure and cognitive load. Many economic decisions are
taken under intense time pressure, e.g. when trading on the stock market. It can
be cognitively demanding when simultaneously engaging in multiple tasks, especially
in tasks like those concerning business and investments, which are characterized by
1
This chapter is based on: Cappelletti, Güth, & Ploner (2008). Being of two minds: an ultimatum
experiment investigating affective processes. Jena Economic Research Papers, No. 2008-048
31
32
Chapter 3. Affective processes in the Ultimatum Game
high complexity and information load. Understanding the impact of time pressure
and cognitive load on decision-making is therefore quite important.
Our workhorse is the familiar Ultimatum Game. We are interested in both proposer and responder behaviour. While extensive research has been conducted on the
affective aspects of responder behaviour, the affective aspects of proposer behaviour
have been almost entirely disregarded. As regards responders, affect has been previously investigated mainly as a “hot” reaction to a specific unfair real offer. Instead, we
adopt the “cold” strategy vector method, which, if anything, would tend to mitigate
the observed effect.
The impact of cognitive load and time pressure on the deliberative system has been
investigated in several studies, but none of them, to the best of our knowledge, considered both factors simultaneously. The experimental design employed here allows
us to disentangle the effect of cognitive load and time pressure on decision behaviour.
Our results show that proposers offer more under time pressure. Acceptance
thresholds of proposer participants who were asked what they would accept as a responder and proposers’ beliefs about the acceptance of their offers let us conjecture
that the increasing generosity of proposers is due to strategic considerations rather
than to other-regarding concerns. Consistent with previous results (Sutter et al.,
2003), we find that responders are more likely to reject under time pressure. Interestingly, our research shows that both proposer’s and responder’s decisions appear to
be unaffected by cognitive load manipulation.
The remainder of the chapter is organized as follows: the next section explains the
Ultimatum Game and briefly review the main findings emerged from the vast experimental literature that investigated this game; Sections 3.3 and 3.4 briefly illustrate
the dual-system approach used in this study and outline the behavioural predictions;
Section 3.5 describes the experimental design and procedures; Section 3.6 presents
the results; Section 3.7 discusses the findings; Section 3.8 concludes.
3.2 The Ultimatum Game
3.2
33
The Ultimatum Game
The Ultimatum Game (Güth et al., 1982) is a simple game in which two parties bargain for dividing a positive amount of money, denoted by e. One party, the proposer,
determines a demand x for herself, where 0 ≤ x ≤ e. The other party, the responder, can either accept or reject. If she accepts, the proposer gets x and she gets the
residual amount (e − x). If she rejects, both parties get nothing. According to the
standard economic reasoning, the responder is purely guided by payoff maximization.
Thus, she should accept any offer, as long as x < e. Therefore, the game theoretic
solution predicts that the proposer demands x∗ = e − , where is the smallest unit
of currency available, and the responder accepts any amount (e − x) > 0. However,
decades of experimental research have reported anomalies in the Ultimatum Game:
both the proposer and the responder behave in a way that is inconsistent with the
theory. On average, proposers offer about 30-40% of the endowment, with 40-50%
being the modal offer. On the other side, responders reject offers of less than 20%
of the endowment about half the time (Camerer, 2003). The Ultimatum Game has
been widely tested for robustness with respect to various factors. A first group of
factors includes some methodological variables, such as the size of the endowment, the
knowledge of the endowment size, repetition, communication, social distance, the role
assignment method, and the elicitation method. The main results reported in these
studies showed that increasing the monetary stakes has no effect on proposers’ offers,
while makes responders more willing to accept (Hoffman et al., 1996; Cameron, 1999;
Munier and Zaharia, 2003; Carpenter et al., 2005); when the size of the endowment
is unknown to the responder and there is common knowledge about this lack of information, proposers decrease their offers and responders accept lower offers (Straub
and Murnighan, 1995; Kagel et al., 1996; Croson, 1996a); repetition with strangers
seems to have no significant effect when the stakes are low, while it tends to slightly
move both proposer and responder behaviour toward the Nash equilibrium when the
34
Chapter 3. Affective processes in the Ultimatum Game
stakes are high (Slonim and Roth, 1998; List and Cherry, 2000); allowing the responder to make a non-binding request to the proposer has a negative impact on offers
and, for a given offer, it reduces the probability of rejection (Rankin, 2003); reducing
social distance by providing participants with the family name of their counterparts
has no significant effect on the offers2 (Charness and Gneezy, 2008); obtaining the
right to be the proposer through a competition instead of a random procedure makes
proposers lower their offers (Hoffman et al., 1994); compared to the use of the play
method, the use of the strategy method tends to increase both offers and rejection
rates (Oosterbeek et al., 2004).
A second group of factors includes some characteristics of the subjects, such as gender, age, race, physical appearance, field of study, and groups versus individuals. The
main results reported in these studies showed that there are no differences between
men and women when playing the role of proposers, and women receive lower offers
than man (Eckel and Grossman, 2001; Solnick, 2001);3 children appear to be more
self-interested than adults: they propose less and reject less often (Murnighan and
Saxon, 1998; Harbaugh et al., 2003); blacks offer more and reject more often than
nonblacks (Eckel and Grossman, 2001); attractive and unattractive people behave
similarly both when playing as proposer and when playing as responder, and attractive people receive higher offers than unattractive people (Solnick and Schweitzer,
1999); studying economics has an inconsistent effect across experiments4 (Carter and
Irons, 1991; Stanley and Tran, 1998); groups offer and accept lower amounts than
individuals (Bornstein and Yaniv, 1998).5
2
In contrast, this information has a positive impact on offers in the Dictator Game (Charness
and Gneezy, 2008).
3
These two studies found sharply different results about gender effect on responder behaviour.
Since the two studies present some methodological differences (i.e., one-shot design vs. repeated-play
design and strategy method design vs. play method design), the responder role seems to be more
sensitive to the experimental context.
4
Specifically, Carter and Irons (1991) found that economists are more self-interested, since they
offer and accept lower amounts than noneconomists, while Stanley and Tran (1998) found that
economists are less self-interested than other students.
5
The Ultimatum Game is one of the most studied game in the experimental research. Several
other variations than those reported here have been made in the standard version of the game, for
3.3 A dual-system approach
35
This study adds to research on the Ultimatum Game by investigating the affective aspects of ultimatum decisions. Both fair offers by proposer participants6 and
rejection of unfair offers by responder participants are typically viewed as more emotionally motivated. The next section briefly illustrates the dual-system approach used
in this study.
3.3
A dual-system approach
In this study we investigate affective decisions in the Ultimatum Game in a dualprocess perspective, which considers behaviour as the product of two systems that
work according to different rules. Following Loewenstein and O’Donoghue (2005)’ terminology, the deliberative system involves deliberation, thus is maximally demanding
of cognitive resources; in addition, it is goal-oriented, forward-looking, and affect free.
In contrast, the affective system involves instincts and intuition, thus is minimally
demanding of cognitive resources; in addition, it is myopic and primarily driven by
affective states. This duality is also reflected at a neural level, since the two systems
consistently involve distinct neural areas.
The implication of this duality is that the two systems may have different motivations and, thus, clash with one another. The Ultimatum Game is an example of
situations in which the two systems appear to conflict. The final behaviour will be
closer to the goal of the system that prevails. Which system is likely to prevail in a
specific situation depends upon the conditions of the situation. The affective system
is thought to be the default mode, which gives a preliminary response that may or
example the use of different frames (Hoffman et al., 1994; Larrick and Blount, 1997), the presence
of an inactive third party (Güth and van Damme, 1998; Shupp et al., 2006), the presence of an
outside option (Knez and Camerer, 1995; Schmitt, 2004), the provision of additional information
about what others did (Duffy and Feltovich, 1999; Bohnet and Zeckhauser, 2004), or the provision
of advice from predecessors (Schotter and Sopher, 2007). However, an exhaustive survey of the
experimental results in the Ultimatum Game is beyond the scope of this chapter. The most recent
survey available can be found in Camerer (2003).
6
Cool anticipation of rejection of an unfair offer may also induce proposers to abstain from making
unfair offers.
36
Chapter 3. Affective processes in the Ultimatum Game
may not be adjusted by the deliberative system. Thus, the affective system is likely
to have a larger influence on final behaviour either when it is stimulated or when
the deliberative system is weakened (or, naturally, when the two conditions occur
simultaneously).
The affective system can be stimulated by manipulating temporally, spatially,
and socially near environmental stimuli (Loewenstein, 1996) or by inducing affective
states through affective-state-induction techniques such as as giving small presents,
presenting stories, pictures, and movies, giving subjects fake feedback about their
performance on a test, and asking subjects to provide a detailed report of a lifeevent. In this study we do not utilize this emotion-induction approach for several
reasons. There is a lack of consensus about the effectiveness of these procedures
(for a meta-analysis, see Westermann et al., 1996). In addition, these techniques
are often characterised by the use of deceptive procedures (e.g., bogus feedback or
use of confederates), a lack of scripts for participants, a lack of financial incentives,
and explicit instructions to enter a specified mood state, which are likely to trigger
demand effects. These practices go against the principles of experimental economics
(Hertwig and Ortmann, 2001). Finally, this approach does not seem necessary, as the
focus of this study is not on the effect of specific emotions or the valence (positive or
negative) of emotions on ultimatum bargaining decisions.
The deliberative system is responsible for deliberation and control of the affective
system. Thus, when it is weak, it is not only less involved in the decision-making
process, but also has difficulties in exerting control over the impulsive responses of
the affective system. Since the deliberative system works slowly and relies on scarce
processing resources, any factors such as time pressure, cognitive load, and mental
depletion due to stress, exhaustion, sleep deprivation, and decision fatigue will tend
to weaken deliberative processing in decision-making (Lobel and Loewenstein, 2005).
In this study we try to weaken the deliberative system by taxing its resources
3.4 Behavioural Predictions
37
through cognitive load and time pressure,7 so that the affective system exerts greater
control over ultimatum bargaining decisions.
The next section outlines our be-
havioural predictions for both proposers and responders.
3.4
Behavioural Predictions
On the responder side, we expect to observe higher rates of rejection when the deliberative system is weaker. When facing an unfair offer, the cognitive goal of gaining
money and the affective goal of resisting unfairness are in conflict (Sanfey et al., 2003).
When the deliberative system is weaker, the affective goal is more likely to prevail.
Emotional responses to Ultimatum offers, thus, result in higher rejection rates. In
particular, previous studies found that the probability of rejection is positively correlated with the intensity of negative self-reported emotions (Pillutla and Murnighan,
1996; Bosman et al., 2001), the level of physiological emotional response (van´ t Wout
et al., 2006), and the activation of anterior insula (Sanfey et al., 2003), a brain area
associated with negative emotional states. Further, Xiao and Houser (2005) found
that rejection rates fall when responders can express their negative emotions directly
to proposers. Koenigs and Tranel (2007) found that subjects with damaged ventromedial prefrontal cortex, a key brain area for emotion regulation, reject unfair offers
at a higher rate than subjects in the control group. Finally, Sutter et al. (2003) found
that time pressure is associated with higher rejection rates.
On the proposer side, predictions are more difficult. Previous studies almost
entirely disregarded the affective underpinnings of proposer’s behaviour. Proposers
may be either strategically deliberating or intrinsically fair. Strategic considerations
require time and cognitive resources. If these are constrained, proposers may fail
to form definite expectations about what responders would accept and, thus, may
7
In order to make the experimental design less complex, we decided to investigate the effect of
time pressure and cognitive load, leaving the analysis of the effect of mental depletion for future
research.
38
Chapter 3. Affective processes in the Ultimatum Game
opt for a “safe” equal split. Two approaches to other-regarding concerns have been
suggested. According to the intuitionist approach (Haidt, 2001), moral decisions are
primarily driven by quick, automatic, effortless affective processes. According to van
Winden (2007), “it is probably not so much cognition but emotion that plays a major
role in the individual enforcement of, as well as the compliance with, norms like
fairness” (p. 50). The findings of Roch et al. (2000) and Cornelissen et al. (2007)
support the intuitionist hypothesis. In addition, assuming that instinctive responses
require less response time than cognitive responses, Rubinstein (2007) found that
equal division is the more instinctive choice in the Ultimatum Game. Following the
intuitionist approach, disrupting deliberative processing should increase ultimatum
offers. Considering both strategic and other-regarding concerns, the net effect of
inhibiting deliberative processes should thus lead to more generosity by proposers.
In contrast, the rationalist approach views moral decisions as resulting from reasoning and reflection. As Moore and Loewenstein (2004) maintain, self-interest is
automatic, unconscious, and viscerally compelling, whereas considering others generally requires thoughtful processing. The results reported by van den Bos et al.
(2006) and Knoch et al. (2006) support the rationalist hypothesis. In van den Bos
et al. (2006)’s experiments, subjects are more satisfied with advantageous unequal
outcomes when their cognitive processing is limited (through either a cognitive-load
or a time-pressure manipulation). Knoch et al. (2006) found that the disruption of
the right dorsolateral prefrontal cortex, a brain area associated with deliberative processes, increases the acceptance rate of unfair ultimatum offers. This suggests that
deliberation is needed to override self-interested impulses.8 Following this approach,
disrupting deliberative processing should lower ultimatum offers. Since the strategic
and the other-regarding components push offers in opposite directions, the net effect
8
However, subjects’ fairness judgments are not influenced by this manipulation, indicating that
this area of the brain is crucial for the implementation of fairness-related responses.
3.5 Method
39
of impairing deliberative processes seems unclear. We have tried to test these predictions experimentally. Our experimental design and procedures are detailed in the
next section.
3.5
3.5.1
Method
Participants and Procedures
376 students (154 males and 222 females) at the Friedrich Schiller University in Jena
(Germany) participated in the experiment. They were randomly assigned to one of
the 8 conditions described in the next section. Participants were recruited through
the online recruitment system ORSEE (Greiner, 2004). On their arrival at the laboratory, participants were seated in computer-equipped cubicles that do not allow
communication or visual interaction among the participants. In order to prevent the
use of external aids (e.g., paper and pencil, cellphone) during the experimental tasks,
participants were asked to leave their personal belongings at the entrance.
The experiment was programmed and conducted using the Z-tree software (Fischbacher, 2007). Participants received written instructions, which were first read individually by the participants and then aloud by a German-speaking collaborator to
establish common knowledge. Understanding of the instructions was tested through
an on-screen questionnaire that subjects were asked to answer before the experiment.
Including payment, sessions lasted for about 50 minutes, and participants earned, on
average, e 9.57 (including a show-up fee of e 2.50).
3.5.2
Treatments
The experiment has a 2 (cognitive load: load vs. no load) × 2 (time pressure: high
vs. low) × 2 (incentives: low vs. high) between-subject design. The 8 treatments are
summarized in Table 3.1.
Cognitive load was manipulated through a dual-task procedure. Participants in
40
Chapter 3. Affective processes in the Ultimatum Game
Table 3.1: Treatments
Treatment code
Cognitive Load
Time Pressure
Incentives
cl1.tp0.i0
Yes
Low
Low
cl1.tp0.i1
Yes
Low
High
cl1.tp1.i0
Yes
High
Low
cl1.tp1.i1
Yes
High
High
cl0.tp0.i0
No
Low
Low
cl0.tp0.i1
No
Low
High
cl0.tp1.i0
No
High
Low
cl0.tp1.i1
No
High
High
the cognitive-load condition (cl1) were asked to memorize five 3-digit numbers and
keep them in mind while deciding in a Ultimatum Game (hereafter UG). In contrast,
participants in the no cognitive-load condition (cl0) did not confront the memory task
while deciding in the UG.
Time pressure was manipulated by setting a limit for deciding in the UG. Participants in the high time-pressure condition (tp1) had 15 seconds to decide as proposers and 30 seconds to decide as responders. In contrast, participants in the low
time-pressure condition (tp0) had 180 seconds for both kinds of decisions. A postquestionnaire confirmed that the manipulation was successful.9
In addition, we introduced monetary rewards to incentivise participants to exert
effort in the memory task. Most research examining cognitive load did not employ
real incentives (for an exception, see Benjamin et al. (2006)). In order to test the
effect of financial incentives on performances in the memory task, we set two levels of
incentives: e0.30 per digit in the high-incentive condition (i1) and e0.03 per digit in
9
Participants indicated on a 5-point Likert scale (ranging from “not at all” to “very much”)
whether they felt under time pressure while making decisions in the UG. The average scores (3.34 and
1.51 for participants in the high time-pressure and in the low time-pressure condition respectively)
were significantly different (Wilcoxon Rank-Sum Test, p-value < 0.001).
3.5 Method
41
the low-incentive condition (i 0). The payment rule for the memory task is detailed
in the next subsection.
3.5.3
Interaction Structure
Four distinct stages can be identified in the experiment. In Stage 1, each participant
is asked to mentally solve five multiplication problems.10 The problems are presented
successively11 and involve two 2-digit numbers such that they result in a 3-digit number (e.g., 14 × 16 = 224). Each participant is asked to memorize the results of the
problems and to keep them in mind until Stage 3, where she will be asked to recall
them. If a participant calculates more than two problems correctly, she is given a
provisional endowment of e15; otherwise, she is given a provisional endowment of
e7. These endowments constitute the amount of money to be divided in the UG. To
prevent participants from making their decisions in advance, they are informed about
the size of these endowments only when decisions are made in the UG. To incentivise
more choices, the actual performance in the multiplication task is revealed only at
the end of the experiment.
In Stage 2 a UG is played. The strategy vector method is employed to collect
choices in the game. Since the participants’ role in the game is revealed only at the
end of the experiment, each participant is asked to report her preferred options for
each of the two roles in the game. As the proposer (referred to as role A), a participant
has to state the offer she intends to make and, as a responder (referred to as role B),
she has to state her reaction (i.e., acceptance or rejection) for each of the possible
offers. The options available to the proposer depend on the endowment available
(e15 or e7). Each natural number between 1 and the endowment available could be
10
The purpose of the multiplication task is to establish endowment legitimacy (Cherry et al., 2002).
One might argue that this analytic task favours the activation of analytical deliberative processing,
affecting this way subsequent decision-making. If this is indeed the case, it would mitigate the effect
of our manipulation, not enhance it.
11
Once the participants enter the result, they pass to the next problem and cannot return to
previous screens.
42
Chapter 3. Affective processes in the Ultimatum Game
selected as an offer. Since at this stage they do not yet know their performance in the
multiplication task, and thus their endowment, participants have to make a decision
for both possible endowments. To summarize, subjects have to enter four distinct
action profiles in the following order: proposer in the high endowment condition,
proposer in the low endowment condition, responder in the high endowment condition,
and responder in the low endowment condition.12
In Stage 3, each participant is first asked for an assessment of her own and her
partner’s performance in the computations and in the recall task. Thus, four estimates are collected. Each correct guess is rewarded with e0.50. The participant is
then asked to recall the results of the multiplication problems in the same order of
appearance as in Stage 1. Real incentives are provided for the recall task; specifically, when the recalled number is equal to the computed number, each digit equal
to the digit in the correct solution is rewarded with a certain amount of money. This
payment procedure ensures against the introduction of “ad hoc” values to simplify
recall. For example, if the correct solution of a problem is 350, and the value entered
is 358, but the value recalled is 350, then the participant earns nothing for this recall.
If for the same problem the value recalled is 358, the participant is paid for 2 out of
3 digits.
In Stage 4, each participant is asked to estimate how likely the offer she made
is accepted by the responder. In other terms, the participant is asked about her
beliefs of acceptance of the offer made. This question is asked for both endowment
levels. The task was incentivised and the payoffs for each combination of estimated
probability and action of the responder are detailed in Table 3.2.
The payoffs in Table 3.2 are defined according to a quadratic scoring rule (for a
detailed explanation of the rule, see Schotter and Sopher, 2007). The probabilities of
acceptance are defined over the values {0, 20, 40, 60, 80, 100} According to the rule the
Participants failing to submit one or more choices pay a flat penalty of e1 to be subtracted from
the show-up fee.
12
3.5 Method
43
Table 3.2: Quadratic Scoring Rule for the Acceptance Beliefs
Certainty of acceptance
0%
20%
40%
60%
80%
100%
Earning when accepted
0.00
0.70
1.30
1.70
1.90
2.00
Earning when rejected
2.00
1.90
1.70
1.30
0.70
0.00
payoffs associated to certain beliefs (πb ) are defined as follows: when the responder
1
A
2
accepts, πb = 2 − 10000 × 2 × (100 − pa ) ; similarly, when the responder rejects,
1
πbR = 2 − 10000
× 2 × (100 − pr )2 . The rule penalizes both the situation in
which less than full probability was assigned to an event when it happens and the
situation in which some probability was assigned to an event when it does not happen.
This mechanism should induce true beliefs over the alternative events. The payoffs
obtained following this procedure are rounded and presented to the participants in
a table resembling Table 3.2. In order to simplify the task and avoid mistakes, the
participants are only confronted with the table and not with the equations for πbA and
πbR .
After having stated their acceptance beliefs, the participants are informed about
their actual role in the game – proposer or responder – and about the relevant endowment for the UG – high or low. The payoffs for each of the stages13 and overall
are then communicated to each participant.
The sequence of stages just described refers to the conditions with cognitive load
(i.e., cl1). The conditions without cognitive load (i.e., cl0) differ only in the order of
Stage 2 and Stage 3, so that the memory task and the UG are not concurrent.
13
If one or both participants in a pair fail to submit the strategy profile relative to their actual
role and the relevant endowment, both receive nothing for the UG task, since payoffs cannot be
calculated. The earnings from Stage 4 are paid only to the proposers.
44
Chapter 3. Affective processes in the Ultimatum Game
3.6
3.6.1
Data Analysis
Proposers’ Behaviour
Offers
Table 3.3 provides descriptive statistics for offers in the UG in the high-endowment
condition (e15).
Table 3.3: Proposer Offers: high endowment condition
Treatment
N*
mean
std.dev.
median
Pooled
346
6.194
1.696
7.000
cl1.tp0.i0
48
5.729
1.567
6.000
cl1.tp0.i1
46
6.217
2.240
7.000
cl1.tp1.i0
40
5.800
2.210
7.000
cl1.tp1.i1
36
6.417
1.610
7.000
cl0.tp0.i0
48
6.000
1.414
6.000
cl0.tp0.i1
48
6.229
1.134
6.000
cl0.tp1.i0
43
6.628
1.415
7.000
cl0.tp1.i1
37
6.676
1.617
7.000
The mean offer is e6.19 and the median offer is e7.00, revealing a strong concern of
proposers for equity. The mean offer represents about 41% of the large pie of e15,
and the median offer corresponds to one of the two nearly equal splits available to
the proposer.14
Table 3.3 allows us to compare the offers in the UG under different experimental
14
Given that the available endowment is not even, no symmetric equitable splitting is available to
the decision maker. Thus, in the high endowment condition the allocation of 8 to oneself and 7 to
the other and the allocation of 7 to oneself and 8 to the other may be interpreted as fair splits. The
same reasoning applies to the low endowment condition, with the two fair options equal to (4,3) and
(3,4).
3.6 Data Analysis
45
treatments. When comparing treatments differing only in time pressure, a tendency
to offer more under higher time pressure is observed. Moreover, in 3 out of the 4
treatments with low time pressure the median offer (e6.00) is lower than that of the
pooled observations (e7.00).
The impact of time pressure on some of the offers is confirmed by non-parametric
tests (Wilcoxon Rank-Sum Test). In the absence of cognitive load and with incentives
kept constant across comparisons, a statistically significant difference in the distribution of offers is registered (cl0.tp0.i0 vs. cl0.tp1.i0, p-value=0.0025; cl0.tp0.i1 vs.
cl0.tp1.i1, p-value=0.031).
Table 3.4 provides descriptive statistics for offers in UG in the low-endowment
condition (e7). The mean and the median offers are very close to the “equal split”
of offering e3 to the counterpart. As had to be expected, less variability across
treatments is observed in the low-endowment condition (compare Tables 3.3 and 3.4).
Table 3.4: Proposer Offers: low-endowment condition
Treatment
N*
mean
std.dev.
median
Pooled
374
2.957
0.917
3.000
cl1.tp0.i0
48
2.688
0.803
3.000
cl1.tp0.i1
46
3.022
1.238
3.000
cl1.tp1.i0
47
2.681
0.837
3.000
cl1.tp1.i1
46
3.065
0.929
3.000
cl0.tp0.i0
48
2.875
0.703
3.000
cl0.tp0.i1
48
3.146
0.922
3.000
cl0.tp1.i0
45
2.978
0.941
3.000
cl0.tp1.i1
46
3.217
0.786
3.000
46
Chapter 3. Affective processes in the Ultimatum Game
Beliefs
Proposers were asked to estimate the probability that their offer would be accepted
by responders. These beliefs are reported in Tables 3.5 (high-endowment condition)
and 3.6 (low-endowment condition). According to the procedure described in section
3.5.1 above, the values in the tables represent the degree of certainty that the offer
made will be accepted by the responder.
Table 3.5: Proposer Acceptance Beliefs: high-endowment condition
Treatment
N*
mean
std.dev.
median
Pooled
346
77.688
19.122
80.000
cl1.tp0.i0
48
77.083
17.005
80.000
cl1.tp0.i1
46
76.957
23.839
80.000
cl1.tp1.i0
40
74.500
19.209
80.000
cl1.tp1.i1
36
77.222
18.610
80.000
cl0.tp0.i0
48
83.333
17.665
80.000
cl0.tp0.i1
48
76.667
15.065
80.000
cl0.tp1.i0
43
76.279
20.589
80.000
cl0.tp1.i1
37
78.919
20.519
80.000
As the pooled figures in the first row indicate, proposers’ confidence of acceptance is higher in the high-endowment condition than in the low-endowment condition (77.69% vs. 72.19%). Statistically significant differences at a level of 0.05 in
the treatments cl1.tp1.i1, cl0.tp0.i1 and cl0.tp1.i1 are detected by pairwise Wilcoxon
Signed-Rank Tests (p-value=0.013; p-value=0.001; p-value=0.027, respectively). The
mean and median values reported in Table 3.5 indicate a small variance of beliefs
across treatments. The few differences detected by the Wilcoxon Rank-Sum Test do
not suggest a systematic effect of one of the experimental factors on the observed
variable.
3.6 Data Analysis
47
Table 3.6: Proposer Acceptance Beliefs: low-endowment condition
Treatment
N*
mean
std.dev.
median
Pooled
374
72.193
23.040
80.000
cl1.tp0.i0
48
73.750
21.500
80.000
cl1.tp0.i1
46
73.043
23.178
80.000
cl1.tp1.i0
47
66.383
26.079
60.000
cl1.tp1.i1
46
75.217
23.546
80.000
cl0.tp0.i0
48
80.417
19.125
80.000
cl0.tp0.i1
48
64.583
23.786
60.000
cl0.tp1.i0
45
69.333
22.401
80.000
cl0.tp1.i1
46
74.783
21.678
80.000
In comparison, the corresponding data in Table 3.6 are characterized by higher
variability. Wilcoxon Rank-Sum tests detect some significant differences across treatments. In particular, with no cognitive load, time pressure increases the certainty
of beliefs when incentives are low (cl0.tp0.i0 vs. cl0.tp1.i0, p-value=0.013) and decreases the certainty of beliefs when incentives are high (cl0.tp0.i1 vs. cl0.tp1.i1,
p-value=0.038). There are no statistically significant differences in the other comparisons.
Data about the earnings in the belief-elicitation stage provide additional information on the correctness and certainty of beliefs. Indeed, the higher the earnings from
the belief stage, the more correct the belief for a given level of certainty and vice versa.
On average, belief earnings are higher in the high-endowment condition than in the
low-endowment condition (e1.71 vs. e1.60). The difference is statistically significant
according to a Wilcoxon Signed-Rank Test (p-value < 0.001). Thus, proposers in
the high-endowment condition are, on average, more accurate and certain about the
acceptance of their offer.
48
Chapter 3. Affective processes in the Ultimatum Game
Regression Analysis
Table 3.7 reports the results of a Tobit Regression of the offers in UG. This specification has been chosen to account for the limits imposed to the offers in the experiment.
The dependent variable Offers is regressed on the explanatory treatment variables Time Pressure, Cognitive Load, and Incentives - and on the number of correct recalls (Correct Recall ) in the memory task.15 This last variable provides a proxy of
the actual effort in the memory task and a control on possible wealth effects in the
game. The interactions between Cognitive Load and Time Pressure (TPCL), Cognitive Load and Incentives (IncCL), and Cognitive Load and Correct Recall (CRCL)
are also included in the model.
Table 3.7: Proposer Behaviour (Tobit regression)
Acceptance
Coeff (Std. Err.)
High endowment
Time Pressure
0.507 (0.257)**
Cognitive Load
0.009 (0.594)
Incentives
0.126 (0.255)
0.242 (0.140)*
Correct Recall
-0.038 (0.036)
-0.033 (0.020)*
TPCL
-0.337 (0.366)
-0.029 (0.198)
IncCL
0.510 (0.365)
0.155 (0.198)
CRCL
-0.042 (0.051)
0.005 (0.027)
cons
6.436 (0.438)***
0.057 (0.140)
-0.318 (0.326)
3.210 (0.241)***
Obs
346
374
Prob > F
0.009
0.005
Pseudo R2
0.014
0.020
∗∗∗
15
Low endowment
(0.01);
∗∗
(0.05); ∗ (0.1) significance level
Here we do not employ the number of rewarded recalls, but the number of correct recalls regardless of the correctness of the computation. The latter provides a better measure of actual cognitive
effort in the memory task.
3.6 Data Analysis
49
In the high-endowment condition (first column of Table 3.7), a significant positive
impact of time pressure on the offer is registered. The other explanatory variables
do not have a significant impact on the dependent variable. In the low endowment
condition (second column of Table 3.7), only incentives and correct recalls have a
marginally positive, respectively negative significant impact (i.e., 0.1 level of significance) on the amount offered. In both endowment conditions, the model as a whole
is statistically significant, but the explanatory power of the two estimations is very
low (Pseudo R2 ≤ 0.02).
The lack of a systematic impact of the experimental factors on the certainty of
beliefs is confirmed by an Ordinary Least Squares analysis in which the certainty of
beliefs is regressed on the explanatory variables employed in the model estimation
reported in Table 3.7. In both endowment conditions, the joint null hypothesis that
all the explanatory factors have no impact on beliefs is not rejected at the conventional
significance levels (p-value > 0.1). Consequently, the full estimation results are not
reported.
3.6.2
Responders’ Behaviour
Responses
The acceptance rates of each potential offer in each of the experimental treatments
are illustrated in Figure 3.1 for the high-endowment condition and in Figure 3.2 for
the low-endowment condition.
In both endowment conditions, the rate of acceptance decreases nearly linearly for
offers lower than the asymmetric equitable splitting (e8 - e7 in the high-endowment
condition and e4 - e3 in the low-endowment condition). This is in line with previous findings and goes against the standard economic prediction, based on material
opportunism, that all positive offers should be accepted.
Heterogeneity in the behaviour across treatments is observed in both endowment
conditions. In the high-endowment conditions (Figure 3.1), focusing on values lower
50
Chapter 3. Affective processes in the Ultimatum Game
0.6
0.5
0.4
% Acceptance
0.7
0.8
0.9
1.0
Figure 3.1: Acceptance Frequency: high-endowment condition
cl1.tp0.i0 (N=48)
0.3
cl1.tp0.i1 (N=46)
0.2
cl1.tp1.i0 (N=43)
cl1.tp1.i1 (N=40)
cl0.tp0.i0 (N=48)
0.1
cl0.tp0.i1 (N=48)
cl0.tp1.i0 (N=39)
0.0
cl0.tp1.i1 (N=39)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Offer
0.6
0.5
0.4
% Acceptance
0.7
0.8
0.9
1.0
Figure 3.2: Acceptance Frequency: low-endowment condition
cl1.tp0.i0 (N=48)
0.3
cl1.tp0.i1 (N=46)
0.2
cl1.tp1.i0 (N=47)
cl1.tp1.i1 (N=46)
cl0.tp0.i0 (N=48)
0.1
cl0.tp0.i1 (N=48)
cl0.tp1.i0 (N=46)
0.0
cl0.tp1.i1 (N=44)
1
2
3
4
5
6
7
Offer
than e7, it can be observed that the highest rate of acceptance is registered in the
treatment cl1.tp0.i0 and the lowest is registered in the treatment cl0.tp1.i0. A Fisher’s
Exact Test shows that the difference between the two distributions is significant at
least at the 0.1 level for all offers lower than e7. The same test also shows that the
3.6 Data Analysis
51
highest number of statistically significant differences in acceptance across treatments
is registered for offers of e5. Among the significant pairwise differences registered for
this offer, there is evidence of a systematic impact of time pressure in the treatments
with low incentives (cl1.tp0.i0 vs. cl1.tp1.i0, p-value=0.009; cl0.tp0.i0 vs. cl0.tp1.i0,
p-value=0.001).
Similarly, in the low-endowment condition (Figure 3.2) the rejection rates in treatments cl1.tp0.i0 and cl0.tp1.i0 are, respectively, the highest and the lowest across
treatments. A Fisher’s Exact Test shows that the difference between the two distributions is statistically significant at least at 0.05 level for all offers lower than e4.
The highest number of differences across treatments occurs when the offer is e1.
At this offer, the experimental treatments with high time pressure are characterized
by lower acceptance rates. Furthermore, a statistically significant difference exists
between cl0.tp0.i0 and cl0.tp1.i0 (Fisher’s Exact Test, p-value=0.037).
Minimum thresholds of acceptance (hereafter MTA) exhibit a monotonic rejection
pattern.16 The mean MTA is e3.45 in the high-endowment condition and e2.20
in the low-endowment condition. This implies that, on average, offers lower than
23% of the initial endowment are rejected in the high-endowment condition, while
offers lower than 31.4% of the initial endowment are rejected in the low-endowment
condition. Focusing only on main factor effects, a statistically significant difference
in MTA is registered between cl0.tp1.i0 and cl0.tp0.i0 (Wilcoxon Rank-Sum Test, pvalue=0.038) in the high-endowment condition, and between cl1.tp1.i0 and cl1.tp0.i0
(Wilcoxon Rank-Sum Test, p-value=0.027) in the low-endowment condition. For both
these comparisons, the MTA registered under high time pressure is higher than that
under low time pressure.
16
The number of non-monotonic choice vectors is equal to 11 in the high-endowment condition
and equal to 13 in the low-endowment condition. All the 11 responders expressing a non-monotonic
pattern of rejection in the high-endowment condition also exhibit non-monotonic rejections in the
low-endowment condition.
52
Chapter 3. Affective processes in the Ultimatum Game
Regression Analysis
Table 3.8 reports the estimated coefficients of a Random Effects Logit Regression of
acceptance behaviour. The baseline in the regression is represented by the acceptance
of “hyper-fair” offers (i.e., offers that are greater than the proposer-favoring equitable
splitting).17
In both the high-endowment and the low-endowment conditions, higher time pressure reduces acceptance.18 The number of correct recalls has a positive, but weakly
significant, effect. As expected, the data show that the lower the proposers’ offers
the lower the likelihood of acceptance. Finally, the model specification is statistically
significant in both endowment conditions.
This regression analysis provides an overview of responder behaviour for all possible offers in the game. The logistic regressions reported in Tables 3.9 and 3.10 offer
a more detailed description of responder behaviour to a given “unfair” offer. The
Tables report only the estimations that have at least a statistical significance at the
0.1 level.
In the high-endowment condition (Table 3.9), no strongly significant effect is found
for offers of e7. For offers of e6, a positive significant effect of the number of correct
recalls is registered. The model estimation conditioned on offers of e5 provides some
interesting results. First, a strong negative effect of time pressure is observed. Second,
the number of correct recalls positively affects the likelihood of acceptance. Third,
cognitive load increases acceptance but the effect is statistically significant only at the
0.1 level. Finally, the interaction between cognitive load and the number of correct
recalls is significantly negative. This suggests that for offers equal to 1/3 (i.e., e5) of
the endowment, exerting effort in the memory task produces an effect similar to that
In the high-endowment condition, all decisions to accept offers equal to or greater than e8 are
aggregated in the baseline. Similarly, in the low-endowment condition, all decisions to accept offers
equal to or greater than e4 are included in the baseline.
18
The odds ratios computed from the coefficients reported in Table 3.8 are equal to 0.298 for the
high-endowment condition, and to 0.308 for the low-endowment condition.
17
3.6 Data Analysis
53
Table 3.8: Responder Behaviour (Random-Effects Logistic Regression)
Acceptance
Coeff (Std. Err.)
Time Pressure
High endowment
Low endowment
-0.829 (0.347)**
-0.753 (0.384)**
Cognitive Load
0.333 (0.791)
1.285 (0.891)
Incentives
0.226 (0.348)
0.243 (0.382)
Correct Recall
0.101 (0.051)*
0.103 (0.055)*
TPCL
0.446 (0.482)
-0.322 (0.541)
IncCL
-0.506 (0.482)
-0.493 (0.542)
CRCL
-0.008 (0.067)
-0.070 (0.075)
offer.7
-1.029 (0.342)***
offer.6
-2.245 (0.272)***
offer.5
-3.425 (0.246)***
offer.4
-5.339 (0.245)***
offer.3
-6.504 (0.256)***
-2.160 (0.272)***
offer.2
-7.182 (0.265)***
-5.114 (0.277)***
offer.1
-7.659 (0.272)***
-6.144 (0.300)***
5.399 (0.641)***
4.677 (0.693)***
cons
Obs
351(15)
Prob > chi2
∗∗∗
373(7)
< 0.001
(0.01);
∗∗
< 0.001
∗
(0.05); (0.1) significance level
produced by time pressure.
In the low-endowment condition (Table 3.10) only time pressure has a significant
impact on the likelihood of acceptance when the offer is equal to e3. The effect is
similar, both in magnitude and in direction, to that observed in the high-endowment
condition.
An OLS regression of the MTA on the explanatory variables represented by the
54
Chapter 3. Affective processes in the Ultimatum Game
Table 3.9: Responder Behaviour (Logistic Regression): high-endowment condition
Acceptance
Coeff (Std. Err.)
Offer=7
Offer=6
Offer=5
Time Pressure
-0.706 (0.753)
-1.057(0.574)*
-1.248 (0.422)***
Cognitive Load
-1.521 (1.602)
1.120 (1.227)
1.731 (0.966)*
Incentives
-0.474 (0.754)
-0.508(0.558)
0.478 (0.411)
Correct Recall
0.122 (0.103)
0.171 (0.078)**
0.119 (0.058)**
TPCL
2.323 (1.363)*
-0.043(0.799)
0.370 (0.590)
IncCL
-0.490 (1.190)
0.560 (0.763)
-0.609 (0.576)
CRCL
0.163 (0.154)
-0.159 (0.105)
-0.160 (0.081)**
cons
2.551 (1.201)**
1.611 (0.887)*
0.770 (0.677)
Obs
351
351
351
Prob > chi2
0.067
0.041
0.004
∗∗∗
(0.01);
∗∗
(0.05); ∗ (0.1) significance level
treatment variables (Time Pressure, Cognitive Load, and Incentives), the number of
correct recalls (Correct Recall ) in the memory task, and the interactions between Cognitive Load and Time Pressure (TPCL), Cognitive Load and Incentives (IncCL), and
Cognitive Load and Correct Recall (CRCL) shows that time pressure has a significant
positive effect on the MTA, both in the high-endowment (coefficient =0.575, p-value
=0.095) and in the low-endowment (coefficient =0.311, p-value =0.067) conditions.
3.6.3
Other Findings
Offers and Minimum Threshold
The strategy vector method allows us to collect decision data of the same individual
in both roles, proposer and responder. In the high-endowment condition, 14.4% of
the participants with a monotonic acceptance pattern set their acceptance threshold equal to their offer in the UG. The mean spread between the MTA and the
3.6 Data Analysis
55
Table 3.10: Responder Behaviour (Logistic Regression): low-endowment condition
Acceptance
Coeff (Std. Err.)
Offer=3
Offer=1
Time Pressure
-1.089 (0.510)**
-0.538 (0.310)*
Cognitive Load
0.903 (1.072)
0.937 (0.714)
Incentives
0.456 (0.474)
-0.064 (0.308)
Correct Recall
0.104 (0.066)
0.066 (0.045)
TPCL
0.222 (0.703)
-0.096 (0.431)
IncCL
-0.502 (0.662)
-0.194 (0.430)
CRCL
-0.089 (0.089)
-0.053 (0.060)
cons
1.481 (0.783)*
-0.856 (0.540)
Obs
373
373
Prob > chi2
0.075
0.073
∗∗∗
(0.01);
∗∗
(0.05);
∗∗
(0.1) significance level
offer is − e2.64, and the spread between the two distributions is highly statistically significant (Wilcoxon Signed-Rank Test, p − value < 0.001). The percentage
of participants equating their offers with their MTA is higher in the low-endowment
condition (39.9%); consequently, the spread between the MTA and the offer is lower
than that in the high-endowment condition (mean e0.729). Again, this spread is
statistically significant (Wilcoxon Signed-Rank Test, p − value < 0.001). Therefore,
when setting their offers, most proposers follow a “mark-up strategy” in which they
send responders more than the minimum amount that they would accept themselves
as responders.
When regressing the spread between MTAs and offers on the explanatory factors employed in the previous model estimation (see, for example, Table 3.7), no
significant effect is registered in either of the endowment conditions. Moreover, the
joint hypothesis of null effects is not rejected at conventional significance levels. This
56
Chapter 3. Affective processes in the Ultimatum Game
implies that the experimental manipulations do not affect the distance between participants’ actions as proposers and as responders.
Impact of Incentives
The mean number of digits correctly recalled is 6.41 (standard deviation 4.40). In
qualitative terms, the differences across treatments are modest.19 Evidence in support of the positive impact of incentives on the observed performance comes from
the comparison between treatment cl1.tp1.i0 - number of correct recalls is 4.98 - and
treatment cl1.tp1.i1 - number of correct recalls is 7.15. The difference between the two
distributions is statistically significant (Wilcoxon Rank-Sum Test, p−value = 0.029).
In the other pairwise comparisons no significant effect of the incentives alone is registered. Some statistically significant differences across treatments are registered, but
only in interaction with other experimental factors. This suggests that real incentives
are particularly relevant when the environment calls for more cognitive effort. In fact,
the impact of incentives is statistically significant only when both time pressure is
high and memory task and decision-making are concurrent.
Confidence
Participants were asked to guess their own and their partner’s performance in both the
multiplication and the memory tasks. What emerges from the data is that, on average,
participants tend to overestimate both their own performance and their partner’s
performance in the two tasks. In the multiplication task, participants computed
on average 2.80 multiplication problems correctly, while the mean estimation of the
number of problems correctly solved was 3.26 for participants themselves and 3.24
19
In a post-questionnaire, participants indicated on a 5-point Likert scale (ranging from “not at all”
to “very much”) whether they found the memory task difficult. The average score for participants
in the cognitive-load condition (3.31) did not significantly differ from that for participants in the
no cognitive-load condition (3.18; Wilcoxon Rank-Sum Test, p-value > 0.1). In the cognitive-load
condition, the average score for participants in the high-incentive condition (2.80) was significantly
lower than that for participants in the low-incentive condition (3.56; Wilcoxon Rank-Sum Test,
p-value < 0.01).
3.6 Data Analysis
57
for their partners.
In the memory task, the mean performance20 was 2.21, whereas the mean estimated performance was 2.43 for participants themselves and 2.35 for their partners. The differences between the actual and the estimated performance are statistically significant at least at the 0.05 level for all four comparisons (Wilcoxon
Signed-Rank Test). The estimations of own performances and those of one’s partner
do not statistically differ for the multiplication task (Wilcoxon Signed-Rank Test,
p − value = 0.820), but the two measures differ for the memory task (Wilcoxon
Signed-Rank Test, p − value = 0.035).
To summarize, participants tend to be overconfident about their cognitive skills
and those of their partners and, at the same time, slightly underestimate the relative
skills of their partners in the memory task.
Earnings
Participants’ mean earnings in the experiment were e7.07 (excluding the show-up fee
of e2.50) with high variability (standard deviation 4.18). Some statistically significant
differences are registered across different treatments. Focusing on comparisons where
the level of incentives is fixed and ignoring those differences due to the interaction of
different experimental factors, earnings in treatment cl1.tp0.i0 are higher than those
in treatment cl1.tp1.i0 (Wilcoxon Rank-Sum Test, p − value = 0.032), while earnings
in treatment cl0.tp0.i1 are higher than those in treatment cl0.tp1.i1 (Wilcoxon RankSum Test, p − value = 0.014). Therefore, time pressure has a negative impact on the
overall performance in the experiment.
20
To render the estimation task less penalizing, we did not ask for a point estimate (i.e., the
exact number of digits correctly recalled), but for an interval estimate (i.e., the interval in which the
number of correctly recalled digits falls). We used four intervals: 0-3 digits, 4-7 digits, 8-11 digits,
and 12-15 digits. Performances in the memory task are expressed in terms of intervals, not in terms
of number of correctly recalled digits. To compare actual and estimated performances, the actual
performance was recoded in terms of intervals.
58
Chapter 3. Affective processes in the Ultimatum Game
3.7
Discussion
Proposers’ behaviour
Most proposers split the available resources quite equally: on average, they offered
more than 40% of the available resources in both endowment conditions (41.27% in
the high-endowment condition and 42.29% in the low-endowment one) and even more
under time pressure. Therefore, to the extent that time pressure weakens the deliberative system, the affective system induces proposers to make higher offers. Proposer
behaviour may be either strategic or other-regarding. Regarding the strategic component, we hypothesized that time pressure would increase offers. In a possibly false
consensus (Kuhlman and Wimberley, 1976), the proposer may expect the MTA of the
responder to equal her own MTA when playing the role of responder. Playing strategically, the proposer thus should offer her own MTA. On average, MTAs are higher
under high time pressure than under low time pressure, suggesting that strategic
considerations are important.
Concerning the other-regarding component, we considered two competing hypotheses. According to the intuitionist approach, time pressure would increase offers,
while, according to the rationalist view, time pressure would lower them. The difference between the amount proposers offer and the minimum amount they believe
responders will accept (proxied by their own MTA) provides some insight into the
other-regarding component of ultimatum offers. On average this difference is positive, indicating the presence of other-regarding concerns, but it is the same for both
time-pressure conditions. Thus, the net increase of the offers is mainly caused by
the strategic component, while the other-regarding component remains unchanged.
This finding is corroborated by proposers’ confidence of acceptance. Although offers
increase in the high time-pressure condition, the estimated probability that the offer
is accepted by the responder remains constant. Thus, more generous offers are not
perceived as more likely to be accepted, suggesting that the increase in offers reflects
3.7 Discussion
59
an increase in the expected MTA of the responder.
The positive effect of time pressure on the amount offered is significant only in the
high-endowment condition and this can be attributed to the low variance observed
in the low-endowment condition. Participants may have perceived low-endowment
choices as less relevant. The actual level of endowment (high or low) in each pair was
determined by the number of computation problems correctly solved by the participant in the actual role of proposer. Participants’ estimations of their performance
in the computation task and that of their partner’s performance suggest that, on
average, participants expected to bargain over the high endowment.
In the low-endowment condition, offers are influenced by the level of incentives
and the number of correct recalls in the memory task. The level of incentives in the
memory task has a positive impact on the offer. When the incentives are high, participants have the opportunity to earn a non-negligible amount of money in the memory
task. The opportunity to earn this additional amount seems to render proposers more
generous.
The number of correct recalls can be considered as a measure of actual cognitive
effort exerted by participants in the memory task, and it has a negative impact on
the amount offered. This finding suggests that when the amount of money to be
divided is low (e7), their cognitive investment motivates proposers to demand more
for themselves, possibly to compensate their effort. Participants’ estimations of their
performance in the memory task are significantly higher than those of their partners’
performance. Thus, it could also be that proposers try to retain more for themselves
because they feel they deserve it. Alternatively, the number of correct recalls could
indicate cognitive abilities (Frederick, 2005). According to this interpretation, cognitively more able proposers seem to behave more selfishly. However, previous findings
on the relationship between cognitive abilities and offers are mixed. Benjamin et al.
(2006) found a weak negative relationship between cognitive abilities and giving in a
Dictator Game and Ben-Ner et al. (2004) found the same relationship only for women,
60
Chapter 3. Affective processes in the Ultimatum Game
whereas Brandstätter and Güth (2002) did not find any relationship, neither in the
Dictator nor in the Ultimatum Game.
Contrary to what we expected, cognitive load did not have any effect on the
amount offered. The cognitive load manipulation we used - requiring participants to
remember a string of numbers while performing the task of interest - is common in
psychology studies (e.g., Shiv and Fedorikhin, 1999; Roch et al., 2000; Hinson et al.,
2003; Benjamin et al., 2006). Unlike previous studies, we incentivised the task with
a monetary reward for each digit correctly recalled. This difference in the procedure
may account for the fact that, unlike previous studies, we did not find any effect of
cognitive load. Participants’ judgment about the difficulty of the task may indicate
possible distorting effects of incentives. In the cognitive-load condition, participants
who were given high incentives judged the memory task as significantly less difficult
than participants who were given low incentives. Altogether our data question the
effectiveness of cognitive load manipulation.
Responders’ behaviour
As expected, time pressure has a negative effect on the likelihood of acceptance, both
in the high-endowment and in the low-endowment conditions. Therefore, when deliberation is inhibited by time pressure, affective processes inspire responders to reject.
This result corroborates the findings of Sutter et al. (2003) about higher rejection
rates under time pressure. The result is even more striking when considering that
choices were elicited through the “cold” strategy vector method. Previous studies
investigating the affective aspect of responder behaviour (e.g., Sanfey et al., 2003;
Xiao and Houser, 2005; van´ t Wout et al., 2006) employed the play method, in which
participants know their role when deciding and the responder replies to the actual
offer. Thus, affective responses have been considered only in situations where emotional reactions are expected to be strong. Our findings show that affective reactions
influence responder behaviour also in situations where the uncertainty of the own role
3.7 Discussion
61
as well as the actual offer may mitigate these reactions.
The positive, although weak, effect of the number of correct recalls on the likelihood of acceptance, both in the high-endowment and in the low-endowment conditions, can be explained by the motivation of responders to get at least something
for their effort. Alternatively, if the number of correct recalls is seen as indicating cognitive abilities, our findings suggests that the behaviour of more cognitively
able responders is closer to that predicted by standard economic theory. However,
Brandstätter and Güth (2002) did not find any correlation between cognitive abilities
and responder’s behaviour.
Surprisingly, we did not find a general significant effect of cognitive load on the
likelihood of acceptance.21 However, when letting cognitive load (i.e., the simultaneity
of memory and UG task) interact with the number of correct recalls (i.e., a rough
measure of cognitive effort actually exerted in the memory task), a negative effect on
the likelihood of acceptance is registered.
The greatest heterogeneity in acceptance behaviour across treatments was found
for offers of e5 in the high-endowment condition (i.e., one third of the pie), suggesting
that the tension between accepting and rejecting is possibly greatest for this level of
offers.
Other findings
The impact of incentives on performance in the memory task provides some interesting
methodological insights. Financial incentives improve performance in the memory
task only when the task is extremely demanding, i.e., when memory task and decisionmaking in the UG are concurrent and decision-making takes place under high time
pressure.
Consistent with a large body of previous research (e.g., Larrick et al., 2007),
A weak effect was found only for offers of e5 in the high-endowment condition, but the direction
of the relationship is opposite to that predicted: cognitive load increases the probability that an offer
of e5 is accepted.
21
62
Chapter 3. Affective processes in the Ultimatum Game
participants, on average, overestimated their performance in both the computation
and the memory task. Participants also overestimated their partners’ performance
both in the computation and in the memory task, but perceived themselves as superior
to their partners in the memory task and as equally good in the computation task.
3.8
Conclusions
This study examined the influence of affective processes on proposer and responder
choices in the Ultimatum Game using a dual-system approach. To enhance the influence of affective processes on behaviour, we inhibited deliberative processes by taxing
cognitive resources through time pressure and cognitive load. Our main results show
that time pressure promotes more generous offers by proposers and more rejection
by responders. In contrast, cognitive load does not affect proposer and responder behaviour. The present research suggests that affective processes play a role also when
facing less vivid potential events, which less likely trigger emotional responses. These
findings may be relevant also for other strategic interactions.
About proposers, we cannot disentangle genuine other-regarding behaviour from
strategic behaviour. However, on the basis of acceptance thresholds of proposers who
were asked what they would accept as a responder and proposers’ beliefs about the
acceptance of their offers, we surmise that higher offers under time pressure are chosen
strategically.22
Different from the cognitive load procedure commonly used in psychology, we
introduced financial incentives for the memory task. Future research should investigate whether this methodological difference may account for the lack of effects of the
cognitive load manipulation as reported here. The inconsistency of our results with
previous findings suggests that slightly different cognitive load manipulations may
have different effects on decision processing.
22
This conjecture could be corroborated using a one-person decision task, such as the Dictator
Game.
Appendix
Experimental Instructions – Cognitive Load condition
This appendix contains the instructions (originally in German) we used for the treatment cl1.tp1.i1, that is the treatment with cognitive load, high time pressure, and
high incentives. The instructions for the low time pressure and the low incentives
conditions were adapted accordingly, and the differences among these conditions are
indicated in the text.
Please, read the following instructions carefully. All participants have received the
same instructions. If you have questions, please ask the experimenters privately.
Thank you for taking part in this experiment. You have earned e 2,50 for showing
up on time. You may earn an additional amount of money during the experiment.
The amount you earn depends on what you will do and on what another participant
paired with you will do. All amounts will be paid in cash at the end of the experiment.
The experiment is composed of four stages. In Stage 1, you are asked to solve a
sequence of 5 mental arithmetic multiplication problems. All the problems involve
two-digit numbers (e.g., 12 × 34), and you have 30 seconds to solve each problem.
A countdown timer will be displayed on the screen. You are asked to carry out the
multiplications without any external aid (e.g., calculator, paper and pencil, abacus).
If you solve at least 3 problems correctly, you are provisionally given an amount of
money called X, otherwise you are provisionally given an amount of money called
Y. Note that X is greater than Y. You will be informed about the number of your
correct answers only at the end of the experiment. You have the possibility to earn an
additional amount of money by keeping the results of the 5 multiplications in mind
until you enter Stage 3. You are not allowed to write the results down anywhere. If
63
you are caught using any external aid, you will not be invited to participate in future
experiments anymore.
In Stage 2, you are randomly paired with another participant in the room. You
will never be informed about the identity of your partner, nor will she be informed
of your identity. You and your partner will participate only once in this decision
problem. In each couple, one is in the role of A and the other is in the role of B.
The task is to divide the provisional endowment of A (from the first stage) between
the two of you. Note that only this endowment matters for actual payment. The
provisional endowment of B is cancelled and, thus, will not count for actual payment.
The rules for the division are the following. A makes an offer to B about how to split
the sum. B can accept or reject the offer. If the offer is accepted, the sum of money
is split as agreed. If the offer is rejected, both A and B earn nothing. You will be
randomly assigned to the role of A or B, but you will not be informed of your role
until the end of the experiment. Note also that you do not know your endowment
from Stage 1. This implies that you have to provide a complete strategy profile for
your actions in each of the two roles (A and B) and for each of the two possible
provisional endowments (X and Y) in Stage 1. At this point you are informed about
the value in Euro of X and Y. In total, you have to provide 4 profiles. You have 3023
seconds for each of the two profiles relative to the role of B and 1524 seconds for each
of the two profiles relative to the role of A. A countdown timer will be displayed on
the screen. If you or your partner fail to complete the profile that will be selected for
payment, both of you will gain nothing in this stage. In addition, the show-up fee of
the one of you who failed to submit the profile will be reduced to e 1,50.
In Stage 3, you are asked to answer some questions. Two questions are related
to the performance of your partner. Two questions are related to your performance.
You earn e 0,50 for each correct answer to these 4 questions. Finally, you are asked
23
24
180 seconds in the low time-pressure conditions.
180 seconds in the low time-pressure conditions.
64
to recall the results of the multiplications you carried out in Stage 1. You have to
enter the results in the same order as in Stage 1. The payment rule for this stage
is the following: a) only the results you correctly recall count for payment; b) the
correctly recalled results are compared with the correct results of the multiplications;
c) you earn e 0,3025 for each right digit in the right position. Example: suppose 542
is the result you provided in Stage 1 and correctly recalled in Stage 3. Suppose that
the correct result of the multiplication is 524. You are paid e 0,3026 because only
the first digit of the result you provided is correct.
In Stage 4, you are asked to answer a question related to the behaviour of your
partner in Stage 2. For this question you will earn an amount of money according to
a table which will be displayed on the screen along with the question.
At this point you will be informed about your performance as well as the performance of your partner in Stage 1, your actual role, the actual offer and the actual
response in Stage 2, and your performance in Stage 3. Your final earnings will be
calculated, and you will be paid in cash. The amount of money will be rounded off
to one decimal.
To summarize, your final payoff will be the sum of:
• the show-up fee;
• your earnings in Stage 2;
• your earnings for the questions in Stage 3;
• your earnings for the recollection of results of the multiplications;
• your earnings for the question in Stage 4.
To test your understanding of the instructions, you are asked to answer some
questions before the start of the experiment. You do not get any payment for these
25
26
e 0,03 in the low-incentive conditions.
e 0,03 in the low-incentive conditions.
65
questions. However, the experiment will not start until all participants will have
correctly answered all the questions.
66
Experimental Instructions – No Cognitive Load condition
This appendix contains the instructions (originally in German) we used for the treatment cl0.tp1.i1, that is the treatment with no cognitive load, high time pressure, and
high incentives. The instructions for the low time pressure and the low incentives
conditions were adapted accordingly, and the differences among these conditions are
indicated in the text.
Please, read the following instructions carefully. All participants have received the
same instructions. If you have questions, please ask the experimenters privately.
Thank you for taking part in this experiment. You have earned e 2,50 for showing
up on time. You may earn an additional amount of money during the experiment.
The amount you earn depends on what you will do and on what another participant
paired with you will do. All amounts will be paid in cash at the end of the experiment.
The experiment is composed of four stages. In Stage 1, you are asked to solve a
sequence of 5 mental arithmetic multiplication problems. All the problems involve
two-digit numbers (e.g., 12 × 34), and you have 30 seconds to solve each problem.
A countdown timer will be displayed on the screen. You are asked to carry out the
multiplications without any external aid (e.g., calculator, paper and pencil, abacus).
If you solve at least 3 problems correctly, you are provisionally given an amount of
money called X, otherwise you are provisionally given an amount of money called
Y. Note that X is greater than Y. You will be informed about the number of your
correct answers only at the end of the experiment. You have the possibility to earn an
additional amount of money by keeping the results of the 5 multiplications in mind
until you enter Stage 2. You are not allowed to write the results down anywhere. If
you are caught using any external aid, you will not be invited to participate in future
experiments anymore.
67
In Stage 2, you are asked to answer some questions. Two questions are related to
the performance of another participant, with whom you will be paired in Stage 3. Two
questions are related to your performance. You earn e 0,50 for each correct answer
to these 4 questions. Finally, you are asked to recall the results of the multiplications
you carried out in Stage 1. You have to enter the results in the same order as in Stage
1. The payment rule for this stage is the following: a) only the results you correctly
recall count for payment; b) the correctly recalled results are compared with the
correct results of the multiplications; c) you earn e 0,3027 for each right digit in
the right position. Example: suppose 542 is the result you provided in Stage 1 and
correctly recalled in Stage 2. Suppose that the correct result of the multiplication is
524. You are paid e 0,3028 because only the first digit of the result you provided is
correct.
In Stage 3, you are randomly paired with another participant in the room. You
will never be informed about the identity of your partner, nor will she be informed
of your identity. You and your partner will participate only once in this decision
problem. In each couple, one is in the role of A and the other is in the role of B. The
task is to divide the provisional endowment of A (from Stage 1) between the two of
you. Note that only this endowment matters for actual payment. The provisional
endowment of B is cancelled and, thus, will not count for actual payment. The rules
for the division are the following. A makes an offer to B about how to split the sum.
B can accept or reject the offer. If the offer is accepted, the sum of money is split as
agreed. If the offer is rejected, both A and B earn nothing. You will be randomly
assigned to the role of A or B, but you will not be informed of your role until the end of
the experiment. Note also that you do not know your endowment from Stage 1. This
implies that you have to provide a complete strategy profile for your actions in each
of the two roles (A and B) and for each of the two possible provisional endowments
27
28
e 0,03 in the low-incentive conditions
e 0,03 in the low-incentive conditions
68
(X and Y) in Stage 1. At this point you are informed about the value in Euro of X
and Y. In total, you have to provide 4 profiles. You have 30 seconds29 for each of
the two profiles relative to the role of B and 15 seconds30 for each of the two profiles
relative to the role of A. A countdown timer will be displayed on the screen. If you
or your partner fail to complete the profile that will be selected for payment, both
of you will gain nothing in this stage. In addition, the show-up fee of the one of you
who failed to submit the profile will be reduced to e 1,50.
In Stage 4, you are asked to answer a question related to the behaviour of your
partner in Stage 3. For this question you will earn an amount of money according to
a table which will be displayed on the screen along with the question.
At this point you will be informed about your performance as well as the performance of your partner in Stage 1, your performance in Stage 2, your actual role, the
actual offer, and the actual response in Stage 3. Your final earnings will be calculated, and you will be paid in cash. The amount of money will be rounded off to one
decimal.
To summarize, your final payoff will be the sum of:
• the show-up fee;
• your earnings in Stage 3;
• your earnings for the questions in Stage 2;
• your earnings for the recollection of results of the multiplications;
• your earnings for the question in Stage 4.
To test your understanding of the instructions, you are asked to answer some
questions before the start of the experiment. You do not get any payment for these
29
30
180 seconds in the low time-pressure conditions.
180 seconds in the low time-pressure conditions.
69
questions. However, the experiment will not start until all participants will have
correctly answered all the questions.
70
Chapter 4
Defaults and public goods
provision1
4.1
Introduction
Defaults are predefined choices that become effective when decision makers do not
take an action to change them. We encounter defaults everyday when, for example,
installing software, buying a flight ticket online, or ordering in a fast food. In many
situations marketers, employers, and policymakers set default options, that would not
be problematic in a fully rational world, since people would not stay with defaults
that do not correspond to the best option for them (Thaler and Sunstein, 2003).
However, recent research has shown that defaults have the power to influence individual behaviour in non-strategic situations. Instead, the potential effects of defaults
in strategic situations have not been investigated yet. The present experimental study
is a first step towards redressing this lacuna. Specifically, we focus on default effects
in a social dilemma such as a public goods problem, which is a typical form of strategic interaction. In two experiments we investigate whether the presence of a default
contribution, i.e., a contribution that participants make by default unless they specify
a different contribution, influence contribution choices.
1
This chapter is based on joint work in progress with Matteo Ploner and Luigi Mittone
71
72
Chapter 4. Defaults and public goods provision
The chapter is organized as follows. The next section reviews a series of studies
on default effects in non-strategic situations and the potential explanations for these
effects put forward in the literature. Section 4.3 explains the Public Goods Game,
which we will utilize in Experiment I. In addition, it reviews a series of studies that
explored what factors improves cooperation in the Public Goods Game. We investigate whether the presence of a default contribution is an additional factor. Section
4.4 explains the Threshold Public Goods Game, which we will utilize in Experiment
II, and the equilibria of the game. The default contribution will be set in correspondence to one of these equilibria. Sections 4.5 to 4.7.3 present the experimental design,
data analysis, and discussion of the results of both Experiments I and II, in which
we also test whether restricted cognitive resources of decision-makers can account
for default effects. Section 4.8 offers some concluding remarks and indicates lines of
future research.
4.2
Default effects in non-strategic settings
An important domain in which defaults play a role is retirement savings (e.g., Madrian
and Shea, 2001; Choi et al., 2004; Bütler and Teppa, 2007; Beshears et al., 2008). For
example, Madrian and Shea (2001) investigated participation in 401(k) retirement
plans in a large U.S. company that changed the participation default. Under opt-in
default, employees were not enrolled in the pension plan unless they explicitly elected
to do so, while under opt-out default, employees were automatically enrolled in the
pension plan unless they explicitly requested not to be. The participation rate dramatically increased from 49% for those who were hired before the default change to
86% for those who were hired after it. The authors also found a default effect relative
to both the rate at which employees contributed to the pension fund and how they
allocated their contribution.
Default effects have been found in several other domains. Johnson et al. (1993)
4.2 Default effects in non-strategic settings
73
reported huge differences in car insurance choices between two U.S. states. Car insurance law in one state provides a plan with the full right to sue as the default
option, but motorists can opt for a less expensive policy with limited right to sue. In
contrast, in the other state the default option is the limited-right plan, but motorists
can acquire the full right to sue by paying a higher premium. The vast majority of
motorists stuck with the default plan in both states: about 75% of motorists in the
first retained the full right to sue, while only about 20% of motorists in the second
acquired it.
Park et al. (2000) presented a group of subjects with a fully loaded car from which
they could remove unwanted options and another group of subjects with a base model
car to which they could add desired options. Compared to those in the base-model
default condition, subjects in the fully-loaded default condition chose a car with a
larger and more expensive set of options.
In a Web experiment, Johnson et al. (2002) asked some users about their willingness to be notified about health surveys. Users were randomly assigned to one of the
three default conditions: Participate, Not-Participate, and No-Default. The different
conditions were created using a radio-button input format with one of the two yes-no
responses checked by default or no responses checked for the No-Default condition.
The acceptance in the Participate condition (about 89%) was much higher than that
in the Non-Participate condition (about 60%) and not significantly different from that
in the No-Default condition (about 88%). The same pattern was also found when the
question was framed in a negative way.
Johnson and Goldstein (2003) estimated the rates of consent for organ donation in several European countries with different default policies. In explicit-consent
countries—where people are not organ donors unless they register to be—the consent rate ranged from about 4% to 28%, while in presumed-consent countries—where
people are organ donors unless they specify otherwise—the consent rate ranged from
86% to almost 100%.
74
Chapter 4. Defaults and public goods provision
DellaVigna and Malmendier (2006) analyzed consumer behaviour in three big U.S.
health clubs offering monthly contracts—which are automatically renewed unless the
user explicitly rescinds—and annual contracts—which end after one year unless the
user signs up again (for either of the two contracts). After one year, monthly users
were 17% more likely to remain enrolled than yearly users. In addition, though
the cancellation costs are very small, monthly users tend to remain enrolled for a
considerable amount of time after their final visit before canceling, thus suffering
considerable forgone savings.
4.2.1
Why do default effects occur?
Four basic explanations have been put forward in the literature. The first set of
explanations involves the concept of effort, which is assumed to be cognitive, physical, or emotional. Making decisions requires cognitive resources, which are scarce.
Departing from the default is cognitively effortful since it implies making an active
decision, while keeping the default is effortless. Therefore, people may stick to the
default when they have constrained cognitive resources but also when they have unconstrained cognitive resources in an attempt to economize on cognitive effort. In
some situations changing the default also entails physical effort, such as completing
a form, making phone calls, or going to an office. When these costs are incurred,
the default tends to be chosen. Some decisions involve questions that are likely to
generate negative emotions. Staying with the default may reflect the tendency of
people to avoid unpleasant questions in order to minimize emotional costs.
The second set of explanations focuses on cognitive biases related to the concept
of loss aversion (Kahneman and Tversky, 1979), according to which the disutility
associated with a loss is greater than the utility associated with a gain of the same
magnitude. Evidence of loss aversion can be found in the well-known endowment
effect (Thaler, 1980), i.e., the tendency for people to value an object more when they
possess it than when they do not. In the case of defaults, an endowment effect may
4.2 Default effects in non-strategic settings
75
be at work: people may perceive the default option as something they possess and,
thus, place more value on it. Loss aversion is also responsible for the status-quo
bias (Samuelson and Zeckhauser, 1988), i.e., people’s tendency to prefer the current
state of affairs over a change. The status-quo bias often occurs together with the
omission bias (Ritov and Baron, 1992), i.e., people’s tendency to prefer inaction over
action. Retention of the default option can be explained both in terms of status-quo
bias—since the default can be perceived as the current state of affairs—and in terms
of omission bias—since no action is required to accept the default, while an action is
needed to change it.
The third explanation for default effects is related to the complexity of decisions.
Complexity may stem from a variety of factors, for example the amount of time
available for deciding (Dhar and Nowlis, 1999), the number of options to be evaluated
(Iyengar and Lepper, 2000; Iyengar et al., 2004), or the presence of decisional conflict,
i.e., the lack of compelling reasons to choose one option over another (Shafir et al.,
1993). The complexity of decisions can be exacerbated by the lack of expertise in
those domains where some skills are useful or even necessary for understanding the
information at hand, as in the case of financial decisions or software installation.
Research has shown that, when facing difficult decisions, people tend to defer them
or choose the available default option (Tversky and Shafir, 1992). Staying with the
default may thus represent a simple heuristic that people adopt to deal with the
complexity of a decision (Mitchell and Utkus, 2006).
The fourth explanation proposed in the literature places emphasis on the information conveyed by the default option. Specifically, people may interpret the default
as an implicit recommendation. If there are no conflicts of interest between those
who set the default rule and the recipients, the default option is seen as a reasonable
choice, since it can reflect what most people do or what informed people think is
sensible to do (Sunstein and Thaler, 2003).
These explanations are not necessarily mutually exclusive, and the retention of
76
Chapter 4. Defaults and public goods provision
the default option can be driven by different factors across different domains.
From the brief literature review presented above it is evident that the effects of
defaults have been extensively investigated in various areas of individual decisionmaking. However, to the best of our knowledge, no laboratory, natural nor field
studies have explored the role of defaults in domains of strategic interaction. This is
a promising area of research, and the present study is a first step towards redressing
this lacuna. This study focuses on default effects in social dilemmas such as public
goods problems, which are a typical form of strategic interaction.
4.3
The Public Goods Game
This study aims to investigate whether and to what extent default options influence
choices in strategic settings. Specifically, we adopt a linear Public Goods Game, in
which contributions are made on a voluntary base, and examine whether the presence
of a default option affects cooperative behaviour.
In this game each individual i in a group X composed of N individuals is provided
with an endowment ei . Individual endowments ei are symmetric, i.e., each individual
i ∈ X has an equal endowment. Each individual i ∈ X chooses the amount ci she
wants to contribute to the public good, such that 0 ≤ ci ≤ ei . The amount (ei − ci )
not contributed to the public good is kept in the individual’s private account, which
earns a constant return of 1,2 and paid at the end of the experiment. The sum of
all the contributions in a group is multiplied by a factor α and equally split among
the members of the group. Thus, the payoff function of each individual i ∈ X is
P
1
α
given by the following function πi = e − ci + N1 α
c
i∈X i , where N < N < 1.
Given the structure of the payoff function, the social optimum is reached when all the
2
In our experiments, the return on the private account is equal to 1 and the same for all group
members, as commonly used in the literature. However, it can be different than 1 and different
within a group. For an example, see Fisher et al. (1995).
4.3 The Public Goods Game
77
individuals i ∈ X contribute all their endowment, i.e.,
P
i∈X
ci = N e. However, a self-
interested individual has no incentive to contribute to the public good, since for each
unit contributed to the public good she earns only
α
.
N
Thus, following conventional
economic reasoning, the public good will not be produced and individuals will retain
all their endowment in their private account. However, empirical evidence shows that
in one-shot Public Goods Games and in the first round of finitely repeated Public
Goods Games individuals generally contribute about half of their endowment to the
public good (Marwell and Ames, 1979; Ledyard, 1995).
What improves cooperation
Extensive research has been conducted to understand what factors, specifically what
characteristics of the design, improve cooperation in a Public Goods Game. As it can
be seen from the brief overview reported below, the presence of a default contribution
has never been investigated. The present experimental study is a first step towards
filling this gap.
1. Marginal per capita return (MPCR) from the public good. The MPCR can be
defined as the marginal gain, relative to the costs, for moving an additional unit of
the endowment from the private account to the public good.3 When MPCR is higher,
more is at stake, and the cost of contributing to the public good decreases; as a result,
contributions may increase. It must be noted that there is a close relationship between
MPCR and the size of the group (i.e., N ). An increase in N may be accompanied by
either a decrease or an increase in contributions. A decrease in contributions can be
explained by the fact that the coordination of contributions and the detection, and
thus the punishment, of free-riding behaviour are more difficult in larger groups; an
increase in contributions can derive from altruistic motives, since the social benefits
3
If we define the payoff function as πi = p(e − ci ) +
α
N
ci +
P
c
, where p is the return
j
j6=i
α
from the private account, the MPCR is equal to ( p1 )( N
). In our experiments p = 1 and, therefore,
α
M P CR = N .
78
Chapter 4. Defaults and public goods provision
of contributing to the public good increases with N (Davis and Holt, 1993; Ledyard,
1995). Isaac and Walker (1988b) designed an experiment to isolate the effect of
MPCR from that of N and found that an increase in MPCR (from 0.3 to 0.75) is
accompanied by a significant decrease in free riding and a significant increase in the
contribution rates independent of N (which was equal to 4 or 10). In contrast, they
found a weak, if any, size effect going in the direction of the large group being relatively
more cooperative. A reanalysis of Isaac & Walker’s results conducted by Ashley et al.
(2008) shows that the higher MPCR increases contributions by 57 percentage points,
while increasing the group size from 4 to 10 decreases contributions by 6.6 percentage
points.4 In her meta-analysis of linear public goods game, Zelmer (2003) found a
significant positive effect (of about 40 percentage points) of MPCR on contributions
and no significant effect of group size on contributions.
2. Partners versus strangers. In a partners condition, the same composition of
the group is maintained throughout the experiment, while in a strangers condition
subjects are rematched in each repetition of the game. If reputation has a positive impact on contributions, one should expect partners to contribute more than strangers.
However, experimental studies comparing subjects’ behaviour in the partners and
the strangers condition report mixed results: while some studies found that strangers
contribute more than partners (Andreoni, 1988; Palfrey and Prisbrey, 1996), other
studies found the opposite (Croson, 1996b; Keser and van Winden, 2000) or no difference at all between the two conditions (Weimann, 1994). In her meta-analysis,
Zelmer (2003) found that partners contribute nearly 16 percentage points more than
strangers.
3. Framing. Andreoni (1995) showed that using a positive frame – i.e., saying that
contributing to the public good generates a positive externality for the other group
4
Ashley et al. (2008) also report a negative interaction effect, showing that the effect of the higher
MPCR is about 4 percentage points smaller for large groups.
4.3 The Public Goods Game
79
members – significantly increases contributions relative to using a negative frame –
i.e., saying that investing in the private account generates a negative externality for
the other group members. Park et al. (2000) confirmed this result overall, but also
found that the framing effect is significant for subjects who have individualistic value
orientation, while it is not significant for those who have cooperative value orientation.
In her meta-analysis, Zelmer (2003) found that subjects in the positive-frame condition contribute about 19 percentage points more than those in the negative-frame
condition.
4. Information and identification. Evidence on the effect of information about the
actions of the other group members is mixed. While Sell and Wilson (1991) found
a positive effect, other authors did not find any significant effect, neither when the
information was available every period (Weimann, 1994; Croson, 1998; Andreoni and
Petrie, 2004) nor when it was available every few periods (Cason and Khan, 1999).
The combination of information and identification, so that each contribution is associated with the person who made it, has been found to positively influence cooperation when the individualized information was available in each period (Andreoni and
Petrie, 2004; Rege and Telle, 2004), but not when it was available only at the end of
the experiment (Gächter et al., 1996).
5. Repetition. One of the most robust findings in public goods experiments is that
contributions decay over time, with a dramatic decrease in the last period (Isaac et al.,
1985; Isaac and Walker, 1988b). Exceptions can be found when factors that promote
cooperation, such as communication (Isaac and Walker, 1988a) and punishment (Fehr
and Gächter, 2000), are introduced in the design. Another exception is reported by
Noussair and Soo (2008), who showed that decay over time does not occur in a
dynamic setting where the MPCR is an increasing function of previous total group
contributions.
80
Chapter 4. Defaults and public goods provision
6. Communication. Another robust findings in public goods experiments is that
allowing communication among group members has a positive effect on the level
of cooperation (Isaac and Walker, 1988a; Palfrey and Rosenthal, 1991), even though
communication is cheap talk, in the sense that it is not binding.5 In her meta-analysis,
Zelmer (2003) found that when communication is allowed, contributions increase of
about 40 percentage points.
7. Heterogeneity. Heterogeneity can be present in MPCR and in endowments.6 Experimental evidence on the effect of heterogeneous MPCR is quite limited. Fisher
et al. (1995) found weak evidence that, when participating in mixed groups, subjects with low [high] MPCR tend to contribute more [less] then those with the same
MPCR participating in homogeneous groups. However, this effect is not statistically
significant. In her meta-analysis, Zelmer (2003) found no significant effect of heterogeneous MPCR on contributions. Regarding heterogeneity of endowments, Isaac and
Walker (1988a) found that, when communication is allowed, contributions are higher
under conditions of homogeneous endowments.7 Cherry et al. (2005) also found that
heterogeneous endowments lower average contributions. In particular, subjects with
high endowments contribute significantly less than their counterparts under conditions of homogeneous endowments.8 Buckley and Croson (2006) found that under
heterogeneity conditions low-income subjects contribute a higher percantage of their
income than high-income subjects. In her meta-analysis, Zelmer (2003) found that
5
Cason and Khan (1999) showed that communication is effective not only when subjects receive
feedback on total group contributions at the end of each round, but also when subjects receive this
information every few periods.
6
Anderson et al. (2008) introduced heterogeneity by giving group members different show-up fees
for the experiment and found that group contributions decrease when show-up fees are heterogeneous
and known.
7
A re-analysis of Isaac & Walker’s data conducted by Chan et al. (1999) shows that heterogeneity
of endowments reduces aggregate contributions by about 24%, but the effect is significant only at
0.1 level.
8
Experimental evidence also showed that the origin of the endowment does not have any effect on
contributions: participants contribute about the same level regardless of whether their endowment
is windfall, earned (Cherry et al., 2005) or their own money (Clark, 2002).
4.3 The Public Goods Game
81
heterogeneous endowments lower contributions of about 14.5%.
8. Threshold. In a threshold public good, a minimum-aggregate-contribution level
must be reached in order for the public good to be provided. The presence of this
requirement, called provision point or threshold, seems to have some positive effects
on contributions,9 in both one-shot (Dawes et al., 1986) and repeated interactions
(Isaac et al., 1989; Suleiman and Rapoport, 1992). Specifically, Isaac et al. (1989)
and Suleiman and Rapoport (1992) found that an increase in the threshold increases
contributions, but decreases the probability of the public good being provided. However, the positive effect on contributions seems to disappear over time. In addition,
Rapoport and Suleiman (1993) showed that the threshold seems to be ineffective in
the presence of heterogeneous endowments.
9. Punishment. Experimental research showed that the existence of instruments for
costly punishing free-riding, such as pecuniary sanctions (Fehr and Gächter, 2000) or
expulsion from the group (Cinyabuguma et al., 2005), causes a dramatic increase in
contribution levels and the maintenance of high levels over time.10 This is observed
both in Partners and Strangers conditions (Fehr and Gächter, 2000). Sefton et al.
(2007) found that, unlike the opportunity to punish free-riders, the opportunity to
reward cooperators does not increase contributions, while the combination of punishment and rewards is even more effective than punishment alone. However, Walker
and Halloran (2004) demonstrated that the opportunity to reward or punish is quite
ineffective in one-shot interactions.
10. Beliefs. Evidence about the effect of elicitation of subjects’ beliefs about what
the other group members will do is mixed. Croson (2000) showed that eliciting
9
This has also been confirmed in a field experiment conducted by Rondeau et al. (2005).
Nikiforakis (2008) showed that, when the opportunity for counter-punishment is present, subjects are less willing to punish free-riding and, thus, the positive effect of one-sided punishment on
contributions is mitigated.
10
82
Chapter 4. Defaults and public goods provision
subjects’ beliefs before contribution decisions significantly lowers contributions and
increases the number of zero contributions compared to the case in which beliefs are
not elicited. However, in a similar experiment, Wilcox and Feltovich (2000) failed to
find any significant impact on contributions,11 and Gächter and Renner (2006) found
the opposite of Croson’s result.
4.4
The Threshold Public Goods Game
In this game each individual i in a group X composed of N individuals is provided
with an endowment ei . Individual endowments ei are symmetric, i.e., each individual
i ∈ X has an equal endowment. Each individual i ∈ X chooses the amount ci she
wants to contribute to the public good, such that 0 ≤ ci ≤ ei . The amount (ei −ci ) not
contributed to the public good is kept in the individual’s private account, which earns
a constant return of 1, and paid at the end of the experiment. If provided, the public
good gives each individual i an identical reward R. The public good is provided only
if a minimum-aggregate-contribution requirement, called threshold or provision point,
P
is met, i.e., if N
i=1 ci ≥ T . For there to be a social dilemma, ei < T < N ·ei or R < T ,
so that no subject is either able or willing to provide the public good individually. In
addition, T < N · R in order to have social efficiency. In case the threshold is not
reached, we employ a money-back guarantee, i.e., contributions are fully refunded if
PN
12
Thus,
i=1 ci < T . The money-back guarantee decreases the risk of contributions.
the payoff function of each individual i ∈ X is given by the following function
πi =


e i


11
if
PN
ei − ci + R if
PN
i=1 ci
i=1 ci
<T
(4.1)
≥T
Although there are small differences between their experiment and that of Croson, the authors
convincingly argue that these differences are unlikely to explain the different results obtained.
12
Dawes et al. (1986) reported that in one-shot binary decisions the money-back guarantee does
not significantly improve contributions, while other authors (e.g., Isaac et al., 1989; Cadsby and
Maynes, 1999) found that the money-back guarantee is associated with higher contributions and
more provision in repeated interactions with continuous contributions.
4.4 The Threshold Public Goods Game
83
In this case excess contributions are wasted. However, other rebate rules have been
used in the literature (see Marks and Croson, 1998, for a comparison of these rules),
such as the proportional and the utilization rebate rules. According to the proportional
rebate rule, excess contributions are returned to contributors proportionally to their
individual contributions. Thus, under this rule the payoff function of each individual
i ∈ X is
πi =




if
PN
− T ) if
PN
ei


e i − c i + R +
ci
PN
i=1 ci
·(
PN
i=1 ci
i=1 ci
i=1 ci
<T
(4.2)
≥T
Under the utilization rebate rule, excess contributions are used to increase the
production of the public good in a continuous manner. Thus, beyond the threshold,
the game becomes as the one described in Section 4.3, and the payoff function of each
individual i ∈ X is
πi =


e i


4.4.1
e i − ci + R +
if
α
N
PN
i=1 ci
<T
(4.3)
PN
P
·( N
i=1 ci ≥ T
i=1 ci − T ) if
Equilibria of the game
The Nash equilibria of this game are the same independently of the rebate rule considered. There is a symmetric pure strategy equilibrium in which the threshold is
not reached and each subject contributes zero, i.e., ci = 0 for all i. However, the
presence of the money-back guarantee makes this contribution pattern a weak Nash
equilibrium, in the sense that if subject i believes that all the other subjects in her
group will contribute zero, contributing zero is not the unique best response (since
contributions are fully refunded if the threshold is not reached). This symmetric pure
strategy equilibrium is an element of a large set of pure strategy Nash equilibria in
P
which the threshold is not reached, i.e., N
i=1 ci < T and, thus, the public good is
84
Chapter 4. Defaults and public goods provision
not provided. In these equilibria, each subject i contributes any amount ci ≥ 0 and
P
believes that j6=i cj + min(ei , R) < T .13 However, these equilibria are weak, since,
given subject i’s beliefs, there is more than one best response. In addition, these
equilibria would not exist in the absence of a money-back guarantee.
In addition, there is a set of pure strategy Nash equilibria in which subjects
P
contribute exactly enough to reach the threshold, i.e., N
i=1 ci = T . These equilibria
are not Pareto-rankable and are different only in the way subjects share the costs of
providing the public goods among themselves. One of these equilibria is symmetric,
in the sense that all subjects contribute an equal amount, i.e., ci =
T
N
for all i. Given
its symmetric nature, this contribution pattern may act as a focal point on which
groups may coordinate (Cadsby and Maynes, 1999; Marks and Croson, 1998).
P
It must be noticed that the contribution patterns in which N
i=1 ci = T are different in terms of efficiency under the three rebate rules considered above. Under no
rebate rule, these equilibria provide an efficient outcome, in the sense that aggregate
group earnings are maximum. Under proportional rebate, these equilibria are still
P
efficient, but in a weak way since also contribution patterns in which N
i=1 ci > T
lead to the maximum aggregate group earnings. In contrast, under utilization rebate,
P
these equilibria are not efficient, since contribution patterns in which N
i=1 ci > T
lead to higher aggregate group earnings as long as α > 1. In this case, the socially
efficient outcome is obtained when all subjects i contribute all their endowment, i.e.,
PN
i=1 ci = N · ei .
4.5
The present study
As seen in the brief review reported in Section 4.3, experimental research has extensively investigated what factors promote cooperation in Public Goods Games. Among
these factors, the presence of a default option has never been considered. We saw that
13
The lowest value of the endowment ei and the individual reward R from the public good is the
binding constraint on rational contributions.
4.6 Experiment I
85
default options can strongly influence individuals’ behaviour in non-strategic situations and it could be that they have the same strong effect also in strategic situations.
In this study we investigate whether default options can influence contributions to
public goods. In addition, we test whether there is a type of carryover effect, i.e.,
whether default effects persist even after the default option has been removed.
We report on two experiments. In Experiment I, we adopt a standard linear Public
Goods Game and establish a default contribution by using the same method adopted
by Johnson et al. (2002) in Internet privacy policy adoption. Among the proposed
explanations for default effects, we examine whether having constrained cognitive
resources lead people to stick with the default. In terms of dual-system models, we
want to investigate whether following the default is an instinctive response of System
1, which is not overridden by a deliberative response of System 2.
In Experiment II, we adopt a Threshold Public Goods Game and set the default
contribution equal to the amount necessary to exactly reach the threshold when all
subjects make an equal contribution. It has been argued that this contribution pattern is characterized by focality given its symmetric nature (Cadsby and Maynes,
1999; Marks et al., 1999). Thus, in this experiment default contributions assume
“meaningful” values. A default option is established using the same method of Johnson et al. (2002). As in Experiment I, we want to investigate whether constrained
cognitive resources are responsible for default effects.
4.6
Experiment I
In this experiment, participants played a linear Public Goods Game for two periods,
and no feedback about the choices made in the first period was provided until the end
of the experiment. Participants were assigned to 4-person groups and were informed
that the composition of each group was to be kept constant across both periods. At
the end of the experiment, one of the two periods was randomly selected for payment.
86
Chapter 4. Defaults and public goods provision
In each period, each participant was endowed with 10 ECU (Experimental Currency Unit) and asked to decide how much to contribute to a public good that yields
a marginal percapita return of 0.4. In the first period, we set a default contribution
by using the radio-button input format adopted by Johnson et al. (2002): all the
possible options were displayed and one of them was already selected. Participants
could accept the default by clicking the OK button or change the default by selecting
a different option. Figures 4.4(a) and 4.4(c) in the Appendix display what participants saw on their screen in the first period.14 We say that a default effect occurs if
there is a tendency to choose the default itself or to anchor on it. To check for the
presence of a carryover effect, we removed the default in the second period. Participants decided how much to contribute to the public good by typing the amount in
a free-text box. Figures 4.4(b) and 4.4(d) in the Appendix display what participants
saw on their screen in the second period. We say that a carryover effect occurs if in
the second period there is a tendency to choose the default amount of the first period
or to anchor on it.
In addition to default and carryover effects, we investigated whether having constrained cognitive resources can account for default effects. In particular, we taxed
cognitive resources through cognitive load, which was manipulated through a dualtask procedure in which participants had to complete a memory task while making
decisions in the Public Goods Game.
Cognitive load can contribute to default effects for several reasons. Under cognitive load, processing resources are restricted since they are scarce and must be
allocated to different simultaneous tasks. This means that under cognitive load less
cognitive resources are available for decision-making. Sticking with the default may
be a conscious or an unconscious way to economize on cognitive resources. Retaining
the default option can be considered a simple heuristic (Mitchell and Utkus, 2006),
14
To avoid subtle presentation effects, the options were displayed, with equal likelihood, either in
ascending or in descending order.
4.6 Experiment I
87
and past research has shown that under conditions of limited cognitive resources
people rely more on heuristics (Roch et al., 2000; Cornelissen et al., 2007). Past research has also shown that people under cognitive load are more influenced by salient
situational cues (Mann and Ward, 2007). A preselected option can be considered
as a framing cue that possibly focuses the attention of the decision-maker on that
particular option.
4.6.1
Treatments
The experiment was run under 10 different between-subjects treatments. Table 4.1
summarizes the features of these treatments.
Table 4.1: Summary of Treatments in the experiment
Treatment
Options(o)
Default(d)
Cognitive Load(cl)
Code Time (t)
A)
d0.cl0.t30
11 [0,10]
0
0
30
B)
d10.cl0.t30
11 [0,10]
10
0
30
C)
d0.cl1.t30
11 [0,10]
0
1
30
D)
d10.cl1.t30
11 [0,10]
10
1
30
E)
d0.cl1.t7
11 [0,10]
0
1
7
F)
d10.cl1.t7
11 [0,10]
10
1
7
G)
d2.cl1.t7
2 {2,8}
2
1
7
H)
d8.cl1.t7
2 {2,8}
8
1
7
L)
d2.cl0.t7
2 {2,8}
2
0
7
M)
d8.cl0.t7
2 {2,8}
8
0
7
In each treatment, 4 factors were manipulated. The first factor is the number of
options (Options) available in the first period of the Public Goods Game. In 4 of the
10 treatments participants were given a binary choice, i.e., contribute 2 ECU or 8
ECU. This opt-in/opt-out structure of the decision is the same adopted by Johnson
88
Chapter 4. Defaults and public goods provision
et al. (2002), whose way of presenting the default option is replicated in the present
experiment. Past research has shown that increasing the number of options available
inhibits choices, since the decision becomes more complex, thus favoring the retention
of the default. In the remaining 6 treatments, we increased the number of options from
2 to 11, and participants were given the opportunity to choose their contribution in
the set {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10}. If the strength of default effects is affected by the
number of options available, we should observe a stronger default effect in treatments
where the number of options available is higher.
The second factor is cognitive load (Cognitive Load ), which assumes, in Table
4.1, value 0 in treatments where cognitive load is absent and value 1 in treatments
where cognitive load is present. In treatments where cognitive load was present, at the
beginning of the experiment participants received a 7-digit number that was displayed
on their screen for a certain amount of time (see Time Code below). Participants
were asked to memorize this number and recall it at the end of the experiment. In
this way, the memory task and the decision task in the Public Goods Game were
performed simultaneously. The memory task was incentivised: the correct recall of
the number was rewarded with e4. To control for possible wealth effects, before
receiving feedback on the recall performance, participants were asked to guess the
correctness of the recall. A correct forecast was rewarded with e1.
In treatments were cognitive load was absent, participants received a 7-digit number and asked to keep it in mind while reading a short text within 240 seconds.
Participants were then asked to answer a question about the content of the text. A
correct answer to the question warranted a reward of e1. Before starting the Public
Goods Game, participants were asked to recall the number they had been asked to
memorize and to guess the correctness of the recall. Feedback was given only at the
end of the experiment. A correct recall was rewarded with e3, and a correct guess
was rewarded with e1. This allowed us to keep the gain from the memory task constant across treatments. Since participants recalled the number before starting the
4.6 Experiment I
89
Public Goods Game, the memory task and the decisional task in the Public Goods
Game were not performed simultaneously. If cognitive load contributes to the retention of default options, we should observe stronger default effects in treatments where
cognitive load is present.
The third factor is the default option (Default). In half of the treatments the
default option was set equal to the maximum possible contribution level, i.e., 10 ECU
in treatments with 11 available options and 8 ECU in treatments with 2 available
options. In the remaining half of the treatments the default option was set equal to
the minimum possible contribution level, i.e., 0 ECU in treatments with 11 available
options and 2 ECU in treatments with 2 available options.
The fourth factor is the amount of time given to memorize the number (Code
Time). Commonly in studies adopting cognitive load manipulations, the stimulus is
presented for a few seconds. As a check, we set the amount of time given to memorize
the number at two levels. In 4 of the 10 treatments the number to memorize was
displayed on the screen of each participants for 30 seconds, while in the remaining 6
treatments it was displayed for 7 seconds.
4.6.2
Participants and Procedures
200 students at the University of Trento (Italy) participated in the experiment. They
were randomly assigned to one of the 10 treatments described in the previous section.
Participants were recruited by poster advertising placed at the University. Upon their
arrival at the laboratory, participants were seated in visually isolated computer terminals. In order to prevent the use of external aids (e.g., paper and pencil, cellphone)
during the memory task, participants were asked to leave their personal belongings
at the entrance.
The experiment was programmed and conducted using the Z-tree software (Fischbacher, 2007) at the Computable and Experimental Economics Laboratory of the
90
Chapter 4. Defaults and public goods provision
University of Trento.15 Participants received on-screen written instructions,16 which
were first read individually by the participants and then aloud by the experimenter
to establish common knowledge. The instructions emphasized that the identity of
interacting partners would have never been revealed to the participants. Understanding of the instructions was tested through an on-screen questionnaire that subjects
were asked to answer before the experiment. The experiment started only after all
the participants had answered all the questions correctly. Sessions lasted for about
45 minutes, and participants earned, on average, about e9.17
4.6.3
Results
Figure 4.1 displays, for treatments from A to F, i.e., treatments with 11 options
available (see Table 4.1), a summary of the distribution of contribution choices in the
first and in the second period. The boxplots provide us with the usual information
on relevant percentiles of the distribution, the X dots represent the average value of
the distribution, and the horizontal lines are drawn in correspondence to the default
value provided to participants. The squared dots in the figure represent individual
observations, with dashed lines connecting the choice a participant made in the first
period (group on the left) to the choice she made in the second period (group on the
right).
The average contribution in the first period, i.e., when the default is provided, lies
in the median range 4–6 in each treatment. A few extreme values (i.e., 0 or 10) are
observed in each treatment, and the standard deviation is quite large overall, spanning
from 2.48 in d0.cl1.t30 to 3.94 in d10.cl1.t30. The number of choices equating the
15
The staff of CEEL is acknowledged for support in the recruiting of participants and in the
management of the experimental sessions.
16
Experimental instructions are provided in the Appendix.
17
Each ECU was converted in e0.35.
4.6 Experiment I
91
Figure 4.1: Contributions (11 options)
8
6
X
4
0
0
radio button (t1)
form (t2)
radio button (t1)
10
8
6
4
X
0
0
radio button (t1)
form (t2)
radio button (t1)
form (t2)
6
8
form (t2)
X
radio button (t1)
form (t2)
4
radio button (t1)
X
0
0
2
X
2
X
Contribution
6
8
10
F) d10.cl1.t7
10
E) d0.cl1.t7
4
X
2
4
X
Contribution
8
6
X
2
Contribution
form (t2)
D) d10.cl1.t30
10
C) d0.cl1.t30
Contribution
X
2
X
Contribution
6
4
X
2
Contribution
8
10
B) d10.cl0.t30
10
A) d0.cl0.t30
default are overall quite low, with the highest percentage (i.e., 25%) registered in
d10.cl1.t30. The absence of substantial differences across treatments emerging from
the qualitative assessment of Figure 4.1 is confirmed also by a set of non-parametric
tests (i.e., Wilcoxon Rank-Sum Test). Indeed, in all the 15 pairwise comparisons
performed, a p-value greater than 0.1 was observed.
92
Chapter 4. Defaults and public goods provision
When evaluating period to period choices at the individual level, a quite erratic
picture emerges. Nevertheless, the distributions of choices at the population level
seem not to differ substantially in the two periods. This is confirmed also by a set of
non parametric tests testing the difference between choices in the same treatment in
the two periods of contribution (Wilcoxon Signed-Rank Test, p-values> 0.1 in each
of the 6 comparisons performed).
Figure 4.2 presents summary information about the distribution of choices in
treatments G–L, i.e., treatments with 2 options available (see Table 4.1). Unlike in
Figure 4.1, no boxplot for first-period choices is reported here. The dichotomous
nature of the choice in the first period suggested to report in the figure the relative
frequency of the two options available.
In each of the four treatments, the majority of observations are registered in
correspondence with the lower contribution level (i.e., 2). This implies that the average contribution level is below the median value of 5 in each treatment. A set of
Fisher’s Exact Tests does not detect any difference in the contribution choices across
treatments (p-value> 0.1, in each of the 6 pairwise comparisons). The analysis of
individual choices from period to period highlights an overall convergence towards
median contribution levels. It can also be noticed that, coherently, extreme values
of the distribution are chosen according to their proximity to the choice in the first
period. A series of Wilcoxon Rank-Sum Tests detects a weakly significant difference
only between d2.cl1.t7 and d8.cl1.t7 (p-value= 0.088). However, the average contribution is, contrary to what expected in a situation of anchoring to the default, higher
in the former condition than in the latter. This suggests the presence of a negative anchoring effect of the default in the dichotomous choice of participants under
cognitive load.
4.6 Experiment I
93
Figure 4.2: Contributions (2 options)
X
radio (t1)
55%
form (t2)
radio (t1)
10
8
6
X
75%
X
0
2
75%
25%
4
4
X
X
Contribution
6
8
25%
0
2
form (t2)
J) d8.cl1.t7
10
I) d2.cl1.t7
Contribution
X
0
2
70%
6
8
X
45%
4
X
Contribution
4
6
8
30%
0
2
Contribution
10
H) d8.cl0.t7
10
G) d2.cl0.t7
radio (t1)
form (t2)
radio (t1)
form (t2)
To complement the analysis of the impact of defaults on cooperative choices, the
contributions were classified as high if being greater than the median value of 5 and
low if being smaller than this threshold. Data from treatments A–F and G–L were
separately pooled to compute distinct contingency distributions of low/high contributions and low/high defaults in the first and in the second period. This produced
94
Chapter 4. Defaults and public goods provision
four distinct contingency tables. A series of Fisher’s Exact Tests applied to the tables
did not detect any difference in the distribution of low/high contributions according
to the default provided, neither in the first nor in the second period (p-values> 0.1).
4.6.4
Discussion
In Experiment 1 we adopted the design that Johnson et al. (2002) applied to Internet privacy policy decisions. Participants were asked to make their contribution
to a public good by using a radio-button input format. One of the responses was
preselected, so that participants contributed the preselected amount if no action was
taken to change it. Overall, the percentage of people who stuck with the default
was very low, both when the number of responses was kept at its minimum (i.e.,
dichotomous choices) and when the number of responses was increased. We failed to
detect any significant differences across treatments. No default effect was found, either when cognitive resources were unrestricted or when they were restricted through
a cognitive-load manipulation. In addition, no carryover effect was found.
While Johnson et al. (2002) found a sizable effect of defaults in the domain of
Internet privacy policy adoption, we failed to find this effect in strategic decisions
such as a public goods problem. We are fairly confident in excluding that the lack
of effect is due to the fact that participants had such strong and stable preferences
that could not be influenced by minimal framing manipulations such as preselecting
an option. In fact, individual contributions across periods are quite erratic: very few
people made the same contribution in both periods. It could be that the strategic
nature of the decision served to focus participants’ attention on the decision itself, so
that a minimal manipulation was ineffective. Whether subtle manipulations are less
effective in strategic contexts is a promising question for future research. It could also
be that the lack of a default effect was due to the weakness of the design adopted.
Specifically, participants could have interpreted the preselection as a technical need
(i.e., the bullet needed to be positioned somewhere), so that participants did not
4.7 Experiment II
95
even realize that a default was present. The strategic nature of the decision and the
presence of immediate and tangible economic consequences, which were lacking in
Johnson et al.’s (2002) study, could have encouraged participants to invest cognitive
effort. It is also true that when people are in the laboratory, they do not have
any other activity to perform than making decisions in the experiment, so that the
opportunity cost of using cognitive resources may be smaller than when people are
in their natural environment.
4.7
Experiment II
Results of Experiment I showed that the presence of a default contribution established through a “soft” method like the one used by Johnson et al. (2002) does not
produce any significant default effect. However, it could be that defaults have an
effect when the default option is reinforced by some kind of prominence. In Experiment I, the default contribution was set alternatively equal to the minimum and
the maximum possible contribution. The extremeness characteristic of these default
contributions might lead to the opposite effect, i.e., escape from the default, if subjects are characterized by a sort of aversion to extreme values. In Experiment II we
adopt a Threshold Public Good Game and make the default contribution assume a
prominent value. Specifically, the default is set equal to the amount necessary to
exactly reach the threshold when all subjects make an equal contribution. Cadsby
and Maynes (1999) and Marks and Croson (1998) argued that the symmetric nature
of this contribution pattern might make it a behavioural focal point (Schelling, 1960).
Experiment I and Experiment II are similar, but differ in three important aspects.
First, instead of a linear Public Good Game we use a Threshold Public Good Game
and make the default contribution assume a salient value. Second, we add a control
period with no default contribution in order to gather information about what preferences exist in the absence of a default. Finally, in each period we elicit subjects’
96
Chapter 4. Defaults and public goods provision
beliefs about the contributions made by the other group members. Eliciting beliefs
may allow us to shed light on the processes underlying default effects, i.e., to understand whether defaults influence people’s contribution preferences and/or people’s
expectations about others’ contribution behaviour. It has been demonstrated that
people’s behaviour and their beliefs about others’ behaviour in social dilemma situations are positively correlated (e.g., Croson, 2007). However, there is a debate about
the causal direction of the relationship: some propose that people choose their behaviour on the basis of their beliefs, while others suggest that beliefs are projections
(e.g., false consensus effect) or ex-post rationalization of one’s own behaviour (see
Croson and Miller, 2005, for a discussion on these theories). Although Croson and
Miller (2005) found support for the former causal explanation, the debate remains
open. Knowing whether or not the presence of a default contribution affects people’ beliefs about others’ behaviour may be informative per se, regardless of whether
beliefs cause behaviour or vice versa.
In Experiment II, subjects played a Threshold Public Goods Game for three periods. Groups of four subjects were formed. Subjects knew that the group composition
would remain the same throughout the experiment and that the members’ identity
would never be revealed. The three periods were independent, in the sense that no
information about their payoffs and the individual choices of the other members of
their group was given to subjects between one period and the following. At the beginning of each period each subject was endowed with 10 Experimental Currency
Units (ECU) which could be either allocated to a private account or contributed to a
public good. The private account paid 1 ECU per ECU to the subjects only, while the
public good provided an equal reward to each of the group members, independently
of their individual contributions. However, the public good was produced only if aggregate contributions reached a minimum level (threshold). A money-back guarantee
was present, so that contributions were fully refunded in case the threshold was not
reached. For contributions exceeding the required threshold a utilization rebate rule
4.7 Experiment II
97
was applied: excess contributions were used to increase the production of the public
good, which yielded, beyond the threshold contribution, a marginal percapita return
of 0.4.
Each period was composed of two stages: the contribution and the guessing stages.
In the contribution stage, subjects were asked to make their contribution to the public
good, while in the guessing stage subjects were asked to guess the contributions to
the public good made by the other three members of their group. The accuracy of
estimations was rewarded with modest payoffs: if subjects estimated the sum of the
contributions made by the other three members of their group precisely, then each of
them earned 3 ECU; if they guessed wrongly, each of them earned 1.5 ECU divided
by the absolute difference between the estimated sum of contributions and the sum
of contributions actually made.18
In the first period, subjects played a Threshold Public Good Game without any
default contribution. Decisions in this period allow us to control what preferences for
contribution would be independent of default effects. In the second period, subjects
played a Threshold Public Good Game with a default contribution, which was set
using the radio-button input format of Johnson et al. (2002), also used in Experiment
I. In this format all the possible contributions, i.e., all the integers included in the
range 0 − 10, were displayed and one of them was already selected. Subjects could
accept the default contribution by clicking the OK button or change the default by
selecting a different contribution. The default contribution was set equal to
T
,
N
where
T is the threshold and N is the number of subjects in a group (i.e., 4 in our setting).
In other words, the default is equal to the amount necessary to exactly reach the
threshold when all subjects make an equal contribution. As seen in Section 4.4.1,
this contribution pattern is an equilibrium that lead to a socially inefficient outcome:
if contributions exceed the required threshold, the social gains are higher. In this
sense, the default
18
T
N
might turn out to be a “two-edged sword”: on one hand it may
This payment scheme is the same used by Croson (2007).
98
Chapter 4. Defaults and public goods provision
have a positive effect on the probability of provision, but, on the other hand, it may
have a negative effect on the total contributions above the threshold. Comparisons
between choices in the first and in the second period allow us to assess the impact
of potential default effects. If total contributions are below the threshold in absence
of a default, a default effect would increase total contributions up to the threshold,
leading to the provision of the public good. If total contributions are equal to the
threshold in absence of a default, a default effect would not make any change in terms
of total contributions, although it might change the way subjects share the cost of the
public good. If total contributions are above the threshold in absence of a default, a
default effect would shift total contributions down to the threshold.
In the third period, subjects played a Threshold Public Good Game without any
default contribution. Decisions in this period allow us to check for the presence of a
carryover effect, that is whether subjects continue to choose the default contribution,
or at least anchor on it, once the default has been removed.
In addition, we investigated whether weakening System 2 through cognitive load
is correlated with the occurrence of default effects. Specifically, the presence of a
default option, signalled in this case by a tick on one of the options available, is
more likely to stimulate System 1. If System 2 has difficulties to intervene, due to
cognitive load, System 1 is more likely to take control over the final decision. As in
Experiment I, cognitive load was manipulated through a dual-task procedure in which
subjects had to complete a memory task while making their contribution decisions
in the Threshold Public Goods Game. Treatments and parameters are described in
detail in the next section.
4.7.1
Treatments
The experiment has a 2 (threshold: high vs. low) X 2 (cognitive load: load vs.
no load) between-subject design, which originates the four experimental treatments
summarized in Table 4.2.
4.7 Experiment II
99
Table 4.2: Experimental treatments
Cognitive Load
No Cognitive Load
Low Threshold
LT.CL
LT.noCL
High Threshold
HT.CL
HT.noCL
Threshold. In the low threshold condition, the threshold number of ECU necessary
for the provision of the public good was set at 12. Thus, the default contribution,
being equal to the individual share resulting from an equal share rule, was set at 3
ECU (i.e., 30% of the individual endowment). In the high threshold condition, the
threshold number of ECU was set at 28, implying a default contribution of 7 ECU
(i.e., 70% of the individual endowment).
Cognitive Load. In the cognitive load condition, participants were asked to memorize a
7-digit number and recall it at the end of the experiment. The number appeared on the
screen for 7 seconds at the beginning of the experiment, so that participants performed
the memory task and the decision task in the Public Goods Game simultaneously.
Correctness of the recalled number was rewarded with 8 ECU. To control for possible
wealth effects, before receiving feedback on the recall performance, participants were
asked to guess the correctness of the recall. A correct forecast was rewarded with 1
ECU.
In the no cognitive load condition, the recall of the 7-digit number occurred before
starting the Public Goods Game, so that the cognitive resources of participants were
unrestricted during the decisional task in the Public Goods Game. Correctness of
the recalled number was rewarded with 8 ECU, while a correct guess about the
correctness of the recall was rewarded with 1 ECU. To keep the exertion of effort
during the memory task similar to that in the cognitive load condition, participants
were asked to read a short text about memory within 4 minutes, while keeping the
100
Chapter 4. Defaults and public goods provision
number in mind, and to answer a question about the content of the text. A correct
answer to the question was rewarded with 1 ECU. Any feedback was given only at
the end of the experiment.
4.7.2
Participants and Procedures
76 students from various disciplines at the University of Trento took part in the
experiment. They were randomly assigned to one of the 4 treatments described in
the previous section. The procedures were the same as those followed in Experiment
I and described in Section 4.6.2.
Unlike Experiment I, at the end of this experiment, two periods were randomly
selected for payment. For the first period selected, subjects were paid their payoffs
from the contribution decision, while, for the second period selected, subjects were
paid their payoffs from the guessing task. Each ECU was converted in e0.50. Paying
subjects for either their contribution decisions or their guesses in a single period rules
out any hedging strategy that subjects might attempt to apply.
4.7.3
Results and Discussion
The analysis of experimental data will be structured as follows: first, an overview
of some descriptive statistics about contributions and beliefs in distinct experimental
treatments will be provided. A series of non-parametric tests will help assess the
existence of differences in contributions and beliefs across experimental treatments.
Second, a regression analysis will integrate the evidence emerging from the descriptive
analysis and provide further evidence about the impact of experimental factors.
Descriptive Statistics
Figure 4.3 provides us with an immediate overview of choices in the four experimental
treatments given by the two experimental factors (i.e., cognitive load and level of
threshold). In more details, the two threshold levels are represented on the vertical
4.7 Experiment II
101
dimension, while the two cognitive load conditions are represented on the horizontal
dimension.
In Figure 4.3 both summary and individual level pieces of evidence are reported. In
each of the 4 quadrants, 3 boxplots, one for each period of the game, are reported. The
X dots represent the average values of the distribution of choices and the horizontal
lines are drawn in correspondence with the threshold level. Individual choices are
represented by the empty squared dots in the graph and the shaded lines connect
choices made by the same individual.
As Figure 4.3 shows, the central tendency measures are above the individual
T
threshold ( N
) in each of the four experimental treatments. The highest mean contri-
bution is registered in the second choice of the HT.CL treatment (i.e., 8.25).
The central tendencies of the contribution distributions, together with the distribution of individual choices, suggest that the threshold level affects the contribution
level in the PGG. This is confirmed also by a series of non-parametric tests between
experimental treatments. Indeed, when keeping fixed the cognitive load condition
and comparing treatments differing only for the threshold level, all the p-values associated with a Wilcoxon Rank-Sum Test are < 0.01. In contrast, when comparing
treatments characterized by the same threshold but different cognitive load, no statistically significant difference is found (Wilcoxon Rank-Sum Test, all p-values> 0.1).
The existence of differences across experimental treatments is corroborated also by
three Kruskal-Wallis Rank-Sum Tests comparing choices across experimental treatments within the same contribution period (all p-values < 0.001).
A measure of the impact of the threshold level on contribution choices comes
also from a test about the central tendency of the distribution against the threshold
level. A series of Wilcoxon Signed-Rank Tests provide the following picture. In
the first period, a significant difference (i.e., p-values< 0.05) is observed only in the
LT.noCL treatment (p-value= 0.002); in the second period, significant differences are
observed in the HT.CL and LT.noCL treatments (both p-values= 0.001); in the third
102
Chapter 4. Defaults and public goods provision
Figure 4.3: Contributions to the PG
8
6
6
X
X
X
X
t2
t3
t2
t3
8
X
X
4
6
X
0
2
4
6
8
t1
X
10
t3
2
Contribution
X
0
0
t2
10
t1
0
High Threshold
X
2
4
X
4
Contribution
X
2
Low Threshold
8
10
Load
10
No Load
t1
Timing
t1
t2
t3
Timing
period, significant differences are observed in the HT.CL and LT.noCL treatments
(p-value= 0.032 and p-value= 0.031, respectively). Thus, the introduction of the
default contribution seems to push contributions above the threshold level in 2 out
of 4 treatments in the period of introduction (i.e., period 2) and in the following one
(i.e., period 3).
4.7 Experiment II
103
At the individual level, it can be noticed that subjects tend not to confirm their
choices from period to period. However, changes in contributions seem not to follow a
systematic pattern. This is confirmed by a series of non-parametric tests that do not
signal any statistical differences between periods in the same experimental treatment
(Wilcoxon Signed-Rank Tests, all p-values> 0.1)
A complex picture emerges from the descriptive analysis. The threshold level
seems to affect contributions: an increase in the threshold increases contributions
made by the participants. This is in line with the findings of previous experimental
studies (Ledyard, 1995). In contrast, cognitive load does not systematically impact
on contribution choices. The presence of a default seems to have an impact on contributions in period 2 and 3, but this impact follows an unexpected pattern: the
presence of a default seems to push choices above the default level.
Distribution of Choices against the Threshold
Table 4.3 reports the percentage frequency of choices that are equal, bigger, or smaller
T
than the individual threshold ( N
), for each period of the game and each experimental
treatment.
The table shows that the large majority of choices are either equal or bigger than
the threshold, in each period of the game and for each experimental treatment. In
qualitative terms, it can be noticed that in 3 out of 4 treatments (i.e., LT.CL, HT.CL,
and LT.noCL) the highest frequency for contribution higher than the threshold is
observed in the second period of the game (i.e., when the default is introduced).
To understand whether choices systematically differ across periods of the game, we
perform a series of McNemar’s chi-squared tests. Basically, the test informs us about
the statistical difference between the frequencies of those conforming to the threshold
in period t and not in period t+1, and those not conforming in period t and conforming
in period t+1. This amounts to say that the test shows whether the period has
104
Chapter 4. Defaults and public goods provision
Table 4.3: Contributions against the Threshold
Cognitive Load
I
II
III
%
=t
>t
<t
=t
>t
<t
=t
>t
<t
low threshold
50.00
43.75
6.25
18.75
62.50
18.75
50.00
31.25
18.75
high threshold
30.00
55.00
15.00
5.00
85.00
10.00
15.00
65.00
20.00
No Cognitive Load
I
II
III
%
=t
>t
<t
=t
>t
<t
=t
>t
<t
low threshold
30.00
65.00
5.00
30.00
70.00
0.00
35.00
50.00
15.00
high threshold
25.00
60.00
15.00
20.00
55.00
25.00
20.00
55.00
25.00
an impact on the frequency of conformity to the threshold. The tests show that
a marginally significant difference is registered when comparing periods 1 and 2 in
the HT.CL treatment, where the percentage of contributions equal to the threshold
decreases from 30% to 5%, and periods 2 and 3 in the LT.CL treatment, where the
percentage of contributions equal to the threshold increases from 18.75% to 50%
(p-value=0.074, in both cases; p-value> 0.1, for all the other comparisons). Thus,
the introduction of the default seems not to positively impact on the likelihood of
conforming to it, overall. On the contrary, in 2 out of 4 treatments it causes a decrease
in the likelihood of conforming to the suggested value (i.e., the threshold level), in
comparison to either the first or the last period.
Beliefs
Beliefs elicitation provided us with additional information about behavior in the experimental setting under investigation. Table 4.4 presents some summary statistics
4.7 Experiment II
105
about the expected total contributions of the other group members, in distinct periods
and distinct experimental treatments.
Table 4.4: Total Contribution Expected by the other Members of the Group
LT.noCL
LT.CL
HT.noCL
HT.CL
Period 1
N
20.000
16.000
20.000
20.000
Mean
12.600
11.812
21.050
22.250
Median
12.000
9.000
21.000
22.500
StdDev
4.160
6.379
4.019
3.024
Min
7.000
7.000
12.000
15.000
Max
21.000
30.000
28.000
30.000
Period 2
N
20.000
16.000
20.000
20.000
Mean
12.800
11.875
20.900
23.700
Median
12.000
10.000
21.000
23.500
StdDev
4.479
5.841
4.822
3.011
Min
8.000
7.000
9.000
20.000
Max
23.000
30.000
30.000
30.000
Period 3
N
20.000
16.000
20.000
20.000
Mean
12.400
12.125
19.015
23.010
Median
10.500
9.500
20.500
22.350
StdDev
5.185
6.592
6.668
3.205
Min
4.000
5.000
0.300
17.000
Max
24.000
30.000
29.000
30.000
106
Chapter 4. Defaults and public goods provision
Table 4.4 highlights the fact that beliefs are higher in the presence of the high
threshold than in the presence of the low one. A series of Wilcoxon Rank-Sum Tests
confirms that all the comparisons between the two threshold levels within the same
period are characterized by a statistically significant difference (all p-values< 0.01).
Concerning comparisons between the same threshold level and distinct cognitive load
conditions, a significant difference is observed in periods 2 and 3 for the high threshold
level (p-value=0.013 and p-value=0.029, respectively). This evidence suggests that
the introduction of the default is likely to raise expectations in the presence of the
interaction between the high threshold and the cognitive load.
For what concerns the main effect of the presence of a default contribution, i.e.,
the comparisons between periods in the same treatment, no statistically significant
differences are observed for all the comparisons performed (Wilcoxon Signed-Rank
Test, all p-values> 0.1). Thus, the introduction of the default seems not to alter
expectations about others’ behavior.
Regression Analysis
The dependent variable of the regression analysis performed is dichotomous and is
equal to 1 when the contribution is equal to the threshold, and equal to 0 otherwise. The following fixed explanatory factors, all having dichotomous structure, were
considered: cognitive load, high threshold, period 2, and period 3. In addition, the interactions between cognitive load and period 2 and between cognitive load and period
3 were included. To control for potential biases in the estimation due to repeated
choices and unobservable characteristics of the decision-makers, a random explanatory
factor for each participant in the experiment was also introduced in the estimation.
Given the nature of the dependent variable and of the explanatory factors, a
Generalized Linear Mixed Model with a logistic link function is estimated (i.e., Logit
model).19 The results of the estimation, expressed as odds ratios, are resumed in
19
The glmmML function from the R statistical package (R Development Core Team, 2005) was
4.7 Experiment II
107
Table 4.5.
Table 4.5: Choices equal to the threshold - Generalized Linear Mixed Model (Logit,
Odds Ratio)
Odds Ratio (Std. Err.)
Cogn.load
2.339 (0.726)
high threshold
0.302 (0.553)**
Period 2
0.826 (0.619)
Period 3
1.000 (0.610)
Cogn.load×Period 2
0.117 (1.000)**
Cogn.load×Period 3
0.571 (0.868)
Cons
0.412 (0.580)
Obs
228 (76)
∗∗∗
(1%);
∗∗
(5%); ∗ (10%) significance level
Table 4.5 highlights some interesting patterns in the choice to contribute at the
threshold level. Facing an high threshold negatively impacts on the likelihood of
setting the contribution equal to the threshold. Similarly, the interaction between
cognitive load and period 2 reduces the likelihood of a matching between the threshold
– which in period 2 is the default – and the contribution. The regression analysis does
not inform us about the direction of the deviation from the threshold. However, the
descriptive analysis of Section 4.7.3 shows that these deviations are mainly upwards.
The other experimental factors do not significantly impact on the dependent variable.
In sum, the presence of a high threshold seems to encourage higher level of contributions than the minimum individual requirement for the provision of the public
employed in the estimation.
108
Chapter 4. Defaults and public goods provision
good, and this result is in line with the findings of previous studies (Ledyard, 1995).
In addition, in terms of default effect, the presence of the cognitive load in correspondence with the introduction of the default tends to move contributions away from
the threshold, specifically in an upward manner. Thus, the effect of defaults found
here goes in the opposite direction to that predicted by the default bias, i.e., that
people stay with the default. Finally, the impact of the default contribution is transient, since it disappears in period 3, i.e., once the default is removed. Thus, we can
exclude the presence of a carryover effect.
4.8
Concluding remarks
In this study we investigated whether the provision of a default option has an impact
on contributions to a public good and whether this impact persists even after the
default option has been removed. We also tested whether default effects can be
accounted for by a restriction of cognitive resources.
Our results show that there is no default effect when people have unconstrained
cognitive resources. In contrast, default contributions have an impact when people
have constrained cognitive resources, but only when the default contribution assumes
a focal value like the equal individual share of the threshold contribution. However,
this impact goes in the opposite direction than the one in which a default bias would
point: people moved away from the default contribution. Movements were mainly
upwards, and, given the utilization rebate rule used for exceeding contributions, the
presence of a default contribution turned out to be beneficial. However, the effect
was transient, since it disappeared once the default had been removed.
The experiments presented here represent a first step towards the identification of
potential default biases in public goods provision, and much remains to be understood.
4.8 Concluding remarks
109
In this study the default option was provided through minimal framing manipulations, i.e., by preselecting one of the options available. Future research may investigate the effect of default contributions utilizing different mechanisms that make the
presence of a default more evident.
Future work may also explore the effect of defaults in Public Goods Games with
different features, such as the absence of a money-back guarantee, the use of different
rebate rules for exceeding contributions, or the presence of heterogeneity in the return
on the public good. Unlike in this experiment, it might be that, since the probability
of provision is much lower in the absence of a money-back guarantee (Davis and Holt,
1993), a default contribution could be beneficial in terms of probability of provision.
It could also be probable that using a no-rebate rule, in which contributions exceeding
the threshold are wasted, the upward flight from the default contribution disappears.
Finally, one might suspect that default contributions are more effective in an heterogeneous environment, since coordination is more difficult. The results of Croson and
Marks (2001) give support to this view. Exploring the effect of recommendations in
the provision of public goods, they found that recommended contributions have no
effect in homogeneous environments, while they are effective in heterogeneous ones.
Another line of research may be the study of response times. Choices made instinctively, i.e., with greater reliance on System 1, requires less time, while choices
made through the deliberative processes of System 2 require more time (Rubinstein,
2007). The analysis of response times of those who stick with the default contribution
and those who do not may give some insights into the underlying processes of the
default bias.
Appendix
Figures — Experiment I
Figure 4.4: Screenshots of the Contribution Stage
(a) Multiple options: period 1
(b) Multiple options: period 2
(c) Two options: period 1
(d) Two options: period 2
110
Experimental instructions — Experiment I
This appendix contains the experimental instructions translated from Italian. Differences between the cognitive load condition (cl1) and the no cognitive load condition
(cl0) are indicated in the text.
Please, read the following instructions carefully. During the experiment all communication is prohibited. If you have questions, please ask the experimenters privately.
Dear Participant,
during the experiment you will be randomly matched with other three participants
and, thus, you will form a 4-person group. You will never know the identity of the
other three members of your group. Similarly, your identity will never be revealed to
the others.
The experiment consists of two periods and you will be matched with the same
three participants in both periods.
At the beginning of each period you will receive 10 Experimental Currency Units
(ECU). In each period you have to decide, individually and autonomously, how many
of the 10 ECU you want to contribute to a project. The other three group members
will be asked to make the same decision. Whatever you do not contribute will be
deposited on your personal account. The project produces an income that is equally
divided among the four group members. Your earnings are calculated as the sum of
two parts:
- the ECU deposited on your personal account;
- the income from the project.
For example, defining you as A and the other group members as B, C, and D, your
earnings will be equal to
111
EarningsA = 10ECU − ContributionA + 0.4 ∗ (ContributionA + ContributionB +
ContributionC + ContributionD)
You will be informed about the income from the project of period 1 only at the
end of period 2. One of the two periods will be randomly selected and your earnings
in that period will be converted into EUROS and paid out to you at the end of the
experiment. Each ECU will be converted into 0.35 EURO.
At the beginning of the experiment you will have to answer some questions testing
understanding of the instructions. The answers given will not affect your earnings
from the experiment. However, the experiment will start only when all participants
will have correctly answered all the questions.
In addition to the earnings from the period randomly selected, you have the chance
to earn 5 EUROS for your participation in the experiment, depending on your ability
to memorize a 7-digit number (for example, 1234567). The number will be displayed
on your screen for 30 seconds20 at the beginning of the experiment.
[Participants in the cognitive load condition read:]
At the end of the experiment, you will have to answer two questions:
Question 1) Do you think you remember the number you received? (YES/NO)
Question 2) What is the number you received?
If you answer Question 1) correctly, you will earn 1 EURO. If you answer Question
2) correctly, you will earn 4 EUROS.
[Participants in the no cognitive load condition read:]
At the end of the 30 seconds21 , you will be asked to read a short text within 4 minutes,
20
21
7 seconds in treatments E to M.
7 seconds in treatments E to M.
112
after which you will have to answer 3 questions:
Question 1) It is a question relative to the content of the text
Question 2) Do you think you remember the number that you received?
(YES/NO)
Question 3) What is the number you received?
If you answer Question 1) correctly, you will earn 1 EURO. If you answer Question
2) correctly, you will earn 1 EURO. If you answer Question 3) correctly, you will earn
3 EURO.
113
Experimental instructions – Experiment II
This appendix contains the instructions (originally in Italian) we used for the low
threshold condition (LT). Differences between the cognitive load condition (CL) and
the no cognitive load condition (noCL) are indicated in the text.
Dear Participant,
thank you for taking part in this experiment. During the experiment all communication is prohibited. If you have a question please raise your hand, and an experimenter
will answer your questions individually.
The experiment
The experiment consists of three rounds. The three round are independent, in
the sense that the decisions taken in a specific round influence only earnings in that
round and not earnings in the other two rounds.
In each round you will be randomly matched with other three participants and,
thus, you will form a 4-person group. You will never know the identity of the other
three members of your group. Similarly, your identity will never be revealed to
the others. The composition of your group will be kept constant throughout the
experiment, that is the other three participants of your group will be the same in
every round.
Your tasks
In each round you will be asked to complete two tasks: decide the amount you
want to contribute to a project and make some estimates.
1) The contribution to a project
114
At the beginning of each round you will receive 10 Experimental Currency Units
(ECU). You have to decide, individually and autonomously, how many of the 10
ECU you want to contribute to a project. The other three group members will be
asked to make the same decision. Let assume, for simplicity, that you are called A
and the other three group members are called B, C, and D respectively. Let define
your contribution to the project as CA and the contributions of the other three group
members as CB , CC e CD .
Whatever you do not contribute (that is, 10 ECU − CA ) will be deposited on your
personal account.
In each round your earnings are calculated as the sum of two parts:
- ECU deposited on your personal account;
- the income from the project.
The project produces an income only if the sum of the contributions of the participants
in your group equals or exceeds 12 ECU,22 that is if
(CA + CB + CC + CD ) ≥ 12ECU
• If the sum of the contributions is less than 12 ECU, the contribution you made
will be refunded; the project will not be realized and your earnings will be equal
to your original endowment (that is, 10 ECU)
Earnings A = 10 ECU
• If the sum of the contributions is at least equal to 12 ECU, the income from
the project is calculated as follows:
o If the sum of the contributions is equal to 12 ECU, the project produces an
income of 24 ECU that are equally divided among the four group members.
22
In the high threshold condition, the threshold was set at 28 ECU.
115
Therefore, each group member receives 6 ECU.23 Thus, your earnings will
be equal to:
Earnings A = 10 ECU − CA + 6 ECU
o If the sum of the contributions is greater than 12 ECU, the project produces an income of 24 ECU, which are equally divided among the four
group members, plus an income, divided among the four group members,
calculated by multiplying contributions exceeding 12 ECU by 1.6. Thus,
your earnings will be equal to:
Earnings A = 10 ECU − CA + 6 ECU +
1.6·(Ctot −12)
4
2) Your estimates
In addition to make a contribution decision, you will be asked to make estimates
about the other three group members’ contributions. Specifically, you will be asked
to type in three values relative to the contributions of the other three group members.
Your earnings from this task are calculated as follows:
- If the sum of contributions you estimated is equal to the actual sum of contributions of the other three group members, you will earn 3 ECU;
- If the sum of contributions you estimated is different from the actual sum of
contributions of the other three group members, you will earn an amount of
ECU calculated by the following formula:
1.5ECU
|estimated sum of contributions−actual sum of contributions|
Your final earnings
23
In the high threshold condition, the threshold was set at 28 ECU and the income from the
project was 56 ECU. The instructions were modified accordingly.
116
At the end of the experiment two rounds will be randomly selected. You will be paid
for task 1), that is the contribution decision, of the first round selected and for task
2), that is the estimation task, of the second round selected.
Each ECU will be converted into 0.40 EURO.
Your information
You will receive information about the contributions of the other three group
members, and thus information about your earnings, only at the end of the experiment
and not at the end of each round. Similarly, you will be informed about the two rounds
selected for payment only at the end of the experiment, since the random selection
will be made at the end of the experiment.
Additional earnings
[Participants in the cognitive load condition (CL) read]
In addition to the earnings from the two rounds randomly selected, you have the
chance to earn 9 ECU for your participation in the experiment, depending on your
ability to memorize a 7-digit number (for example, 1234567). The number will be
displayed on your screen for 7 seconds at the beginning of the experiment. At the
end of the experiment, you will have to answer two questions:
Question 1) Do you think you remember the number that you received?
(YES/NO)
Question 2) What is the number you received?
If you answer Question 1) correctly, you will earn 1 ECU. If you answer Question 2)
correctly, you will earn 8 ECU.
The use of any external aids (for example paper and pencil or cellphones) is strictly
forbidden. Those who will use external aid to memorize the number will be excluded
from all payments at the end of the experiment and will not be allowed to participate
117
in future experiments.
[Participants in the no-cognitive load condition (noCL) read]
In addition to the earnings from the two rounds randomly selected, you have the
chance to earn 10 ECU for your participation in the experiment, depending on your
ability to memorize a 7-digit number (for example, 1234567). The number will be
displayed on your screen for 7 seconds at the beginning of the experiment. At the
end of the 7 seconds, you will be asked to read a short text within 4 minutes, after
which you will have to answer 3 questions:
Question 1) It is a question relative to the content of the text
Question 2) Do you think you remember the number that you received?
(YES/NO)
Question 3) What is the number you received?
If you answer Question 1) correctly, you will earn 1 ECU. If you answer Question 2)
correctly, you will earn 1 ECU. If you answer Question 3) correctly, you will earn 8
ECU.
The use of any external aids (for example paper and pencil or cellphones) is strictly
forbidden. Those who will use external aid to memorize the number will be excluded
from all payments at the end of the experiment and will not be allowed to participate
in future experiments.
Questionnaire
At the beginning of the experiment you will have to answer some questions testing
understanding of the instructions. The answers given will not affect your earnings
from the experiment. However, the experiment will start only when all participants
will have correctly answered all the questions.
118
Chapter 5
Defaults as recommendations in
public goods provision1
5.1
Introduction
The results of the two experiments reported in the previous chapter show that, unlike
in non-strategic settings where they appear to be robust, default effects in strategic
settings seem to be fragile. In Experiment I we did not find any significant effect
of the presence of a default contribution, while in Experiment II we did find an effect, but it was in the opposite direction than that expected. It is thus evident that
more research is needed to shed light on the mechanisms underlying default effects in
strategic settings. This chapter reports on a novel experiment investigating default
effects in public goods provision. In Experiments I and II, default contributions were
provided through minimal framing manipulations. In contrast, in the present experiment the presence of a default is made evident to participants. In addition, while the
two previous experiments investigated the hypothesis that default effects are related
to restricted cognitive resources, this experiment is designed to examine whether information conveyed by default options is responsible for default effects. Specifically,
it has been proposed that people may interpret default options as a suggestion by
1
This chapter is based on joint work in progress with Matteo Ploner and Luigi Mittone
119
120
Chapter 5. Defaults as recommendations in public goods provision
those who set the default (Madrian and Shea, 2001; Johnson and Goldstein, 2004).
If there are no conflicts of interest between those who set the default rule and the
recipients, the default option is seen as a reasonable choice, since it can reflect what
most people do or what informed people think is sensible to do (Sunstein and Thaler,
2003).
Previous research has investigated the role of suggestions in public goods experiments. An example is given by the studies examining the effect of leadership on
contributions in public goods (or public bad) provision (Güth et al., 2007). In these
studies a group leader decides first how much to contribute to a public good. This
contribution is common knowledge: before simultaneously deciding their contributions, the other group members are informed about the leader’s contribution, and
the leader knows, when deciding, that her contribution will be communicated to the
other group members. The leader’s contribution can be interpreted as a suggestion by
the leader. Do the other group members follow this suggestion? Güth et al. (2007)
found that group members condition their contribution decisions on their leader’s
contribution, leading contributions to higher levels compared to the condition without a leader. However, although being highly correlated, leaders’ contributions are
systematically higher than the other group members’ contributions.
The present experiment differs from leadership studies in two major aspects. First,
unlike in leadership studies, in this experiment suggestions take the form of default
options, i.e., participants contribute by default what is suggested, unless they choose
a different amount. However, as a control, there is also a treatment in which suggested
contributions remain a simple suggestion without being reinforced by the default form.
Second, in leadership studies the participant who sends the suggestion (i.e., the leader)
is a member of the group and, as such, is called to make a contribution. Thus, the
suggestion is somehow reinforced by the fact that, while sending the suggestion, the
leader actually contributes the suggested amount. In contrast, in this experiment the
participant who set the default is associated with a group, but is not a member of
5.1 Introduction
121
that group and, thus, she is not called to make a contribution to the public good.
However, she has an interest in the performance of the group, since her own payoff is
equal to the mean payoff of the group members.
Another study closely related to the present one was conducted by Croson and
Marks (2001). The authors investigated the role of suggestions in a Threshold Public
Goods Game with a money-back guarantee and no-rebate for contributions exceeding
the threshold. Suggestions were posted in the instructions and were based on a
symmetric cost-sharing rule, i.e., the cost of the public good is equally divided among
the group members.2 As also underlined by the authors, this symmetric option is
already a strong focal point, and suggesting it may promote common knowledge that
this is a reasonable pattern of contributions.
There are three major differences between the present experiment and that reported by Croson and Marks (2001). First, in our experiment suggestions are default
contributions, i.e., what participants actually contribute if they do not take an action
to change it. Second, in Croson & Marks’ experiment suggestions are exogenously
made by the experimenter, while in our experiment suggestions are endogenously
made by some participants. Suggestions made by the experimenter have the potential to create a strong demand effect or to make participants perceive a conflict of
interest (e.g., “Why should the experimenter suggest what is in my best interest and
spend more money?”). Third, our suggestions are not strengthened by focal contribution patterns. We adopt a standard Public Goods Game in which, unlike in the
Threshold Public Goods Game, there seems not to be strong focal points.
Unlike the previous studies mentioned above, we elicit participants’ beliefs about
the contributions of the other members of their group. Elicitation of beliefs may help
shed light on processes underlying default effects. Specifically, default options can
2
In the condition with heterogeneous preferences, i.e., group members receive different payoffs
from the public good, a suggestion based on equal period payoffs was also implemented, but did not
produce significantly different results from those obtained with the symmetric cost-sharing suggestion.
122
Chapter 5. Defaults as recommendations in public goods provision
have an impact on both participants’ preferences and on participants’ expectations
about others’ contributions.
5.2
The experimental design
In this experiment, participants played a linear Public Goods Game for three periods. As in Experiments I and II, participants were assigned to 4-person groups and
informed that the group composition would not change across periods. Participants
did not receive any feedback about the choices made by other members of their group
until the end of the experiment. One of the three periods was randomly selected for
payment of the decision at the end of the experiment.
In each period of the game, each participant was endowed with 10 ECU and
asked to decide how much to contribute to a public good that yields a marginal
percapita return of 0.4. Each period was composed of two stages: in the decision stage,
participants made their contribution to the public good, while in the estimation stage
participants estimated the contributions to the public good made by the other three
members of their group.3 The accuracy of estimations was rewarded with modest
payoffs using the same payment scheme adopted in Experiment II: if the participants
estimated the sum of the contributions made by the other three members of their
group precisely, then each of them earned 3 ECU; if they guessed wrongly, each of
them earned 1.5 ECU divided by the absolute difference between the estimated sum
of contributions and the sum of contributions actually made.4 To avoid any hedging
strategy, beliefs were paid in all periods except the one selected for payment of the
decision.
In the first period, participants played a standard linear Public Goods Game. This
allowed us to gather information about the dispositional attitude of participants, i.e.,
3
Participants were also asked to indicate, on a 5-point Likert scale ranging from definitely unsure
to definitely sure, how sure they were about their estimates. However, this task was not incentivised.
4
This payment scheme is the same used by Croson (2007).
5.2 The experimental design
123
what preferences exist in the absence of a default.
In the second period, participants played one of two different randomly assigned
roles, referred to as role A and role B. In each session, there were five 4-person groups
present in the laboratory. The members of four groups played role B, while the
members of one group played role A. Each of the four participants A was associated
with one of the groups composed of participants B. Participants B were asked to
make a contribution to a public good and to estimate the contributions made by
the other three members of their group. Participants A were asked to establish a
contribution that was communicated to the four participants B of the group associated
with them. To collect data for control purposes, all the participants were asked to
establish a contribution before knowing their role. The contribution was effective only
in case a participant was actually assigned the role A. The payoff of a participant A
was calculated as the mean of the payoffs of participants B associated with her. In
addition, participants A were asked to estimate the contributions made by the four
members of the group associated with them.
To check for the presence of default effects, we compared behaviour in two different conditions: the default condition, in which the contribution suggested by the
participant A becomes the default contribution, and the advice condition, in which
the contribution suggested by the participant A remains a suggestion. A default effect occurs if there is a tendency to stick with the default contribution or at least to
anchor on it.
In addition to default effects, we investigated whether these effects occur because
of the informational content of a default option. It has been proposed that the default
option can be interpreted as a suggestion from those who set the default (e.g., Sunstein and Thaler, 2003). Of course, there must be no room for conflicts of interest
between those who set the default rule and the recipients of the rule. Given the
structure of participants A’s payoffs, in this experiment there is no conflict of interest
between the two parties. To check for the importance of the informational content
124
Chapter 5. Defaults as recommendations in public goods provision
of default options, we compared behaviour in two different conditions, in which we
manipulated who set the suggested contribution, alternatively the participant A and
the computer. If a default effect occurs because of the informational content of the
default option, we should observe a tendency to stick with the default when it is set
by the participant A and not when it could be set by the computer.
In addition to default effects and the importance of the informational value of
default options, we checked for the presence of carryover effects. In the third period,
participants played a standard linear Public Goods Game. A carryover effect occurs
if there is a tendency to choose the default contribution of the second period or at
least to anchor on it.
5.2.1
Treatments
This experiment was run under three different between-subjects treatments: advice 100,
def ault 100, and def ault 50-50. These treatments differed only in the second period
of the Public Goods Game. Table 5.1 summarizes the features of these treatments.
Table 5.1: Experimental treatments
Treatment
Form of suggested contribution
Source of suggested contribution
A)
def ault 50-50
Default
50-50
B)
def ault 100
Default
100
C)
advice 100
Advice
100
In each treatment, two factors were manipulated: the form of the suggested contribution (form conditions) and the source of the suggested contribution (source conditions). The contribution suggested by the participant A associated with a group
alternatively assumed the form of a default option and the form of simple advice.
When it assumed the form of a default option (default condition), participants B
5.2 The experimental design
125
in each group were informed about the contribution suggested by the participant A
associated with their group and were told that they would automatically contribute
this amount unless they specified a different amount.
When the suggested contribution assumed the form of simple advice (advice condition), participants B in a group were informed about the contribution suggested by
the participant A associated with their group and were asked to actively make their
contribution, without any kind of automaticity.
The source of the suggested contribution, i.e., who set it, could be the participant
A associated with a group or the computer. In the 100 condition, participants B in
a group were informed about the suggested contribution and were told that it was
established with 100% probability by the participant A associated with their group.
In the 50-50 condition, participants B in a group were informed about a contribution
and were told that it was established with 50% probability by the participant A
associated with their group and with 50% probability by the computer randomly.
5.2.2
Participants and Procedures
120 students at the University of Trento (Italy) participated in the experiment. They
were randomly assigned to one of the three treatments described in the previous section. Participants were recruited by poster advertising placed at the University. The
experiment was conducted at the Computable and Experimental Economics Laboratory of the University of Trento.5
Upon entering the laboratory, participants were randomly assigned to computerequipped cubicles that do not allow visual interaction among the participants. Participants received on-screen written instructions,6 which emphasized that the identity
of interacting partners would have never been revealed to the participants. Questions
were answered individually by the experimenter at the participants’ seats. Sessions
5
Marco Tecilla is acknowledged for developing the software and for support in the recruiting of
participants and in the management of the experimental sessions.
6
Experimental instructions are provided in the Appendix.
126
Chapter 5. Defaults as recommendations in public goods provision
lasted for about 45 minutes, and participants earned, on average, about e9.74 (including a show-up fee of e3).7
5.3
Results
Contribution choices
In both treatments default 50-50 and default 100, 43.75 % of the subjects choose to
follow the suggested contribution in the second period of the game. Thus, the source
of the suggested contribution seems not to affect the likelihood of it being chosen.
However, it must be considered that the suggested contributions are overall higher
in treatment default 50-50 (5.63 vs. 4.00). The regression analysis presented below
will provide a better assessment of the participants behavior by taking into account
also the level of the suggested contribution observed. In contrast, a startling difference in the percentage of participants following the suggested contribution received
is registered when comparing experimental treatments with and without default. In
the treatment without default only 3 out of 32 participants (i.e., 9.38%) follow the
suggested contribution received. This results is corroborated by the fact that the distributions of suggested contributions provided by other participants in default 100
and advice 100 do not statistically differ. A Fisher Exact Test confirms that there is
a statistical difference between the treatment with the default and the treatment without the default, both when keeping the source condition constant (i.e., default 100
vs. advice 100 ; p-value=0.004) and when pooling the data collected in the presence
of a default irrespectively of the source condition (i.e., default 50-50 and default 100
vs. advice 100 ; p-value=0.001).
Figure 5.1 presents a summary description, for each experimental treatment, of
choices in the three independent periods of the PGG, distinguishing between those
who accept and those who reject the suggested contribution in the second period.
7
Each ECU was converted in e0.50.
5.3 Results
127
Specifically, the boxplots provide us with usual information on relevant percentiles of
the distribution of choices, the squared dots identify individual observations, which
are connected across periods by a dashed line. Finally, the X dots identify the average
contribution level in each period of the game.
Figure 5.1: Choices
t3
0
0
10
t3
t3
X
X
6
8
t2
X
4
X
X
0
0
t3
10
t2
6
4
X
X
X
0
0
2
Contribution
8
10
t1
8
6
X
2
4
t2
2
X
Contribution
8
4
6
X
2
Contribution
Default_100
X
t1
Contribution
X
t1
10
t1
Advice_100
6
8
t2
X
4
X
X
2
X
Contribution
6
4
X
2
Contribution
Default_50−50
8
10
Sugg. Contrib. ACCEPTED
10
Sugg. Contrib. REJECTED
t1
t2
t3
t1
t2
t3
Across periods, contributions to the public good tend to concentrate on values
128
Chapter 5. Defaults as recommendations in public goods provision
comprised between 2 and 6.
At the individual level, choices tend to be quite erratic across periods. This
is particularly true for what concerns those who follow the suggested contribution.
Smaller “jumps” are registered for those refusing the suggested contribution, even if
very few participants stick to the same contribution value across all periods. Finally,
among those accepting the suggested contribution no carryover of the contribution in
the second period to that in the third one is observed.
Overall, alternative source conditions (i.e., 50-50 and 100 ) seem not to significantly affect the contribution levels. A series of non-parametric tests detects a
marginally significant difference between the two source conditions (i.e., default 5050 vs. default 100 ) only among choices in the second period of those accepting
the suggested contribution (Wilcoxon Rank-Sum Test, p-value=0.076). For all the
other comparisons no significant differences are detected (Wilcoxon Rank-Sum Test,
p-value> 0.1).
When pooling data irrespectively of the source condition (i.e., default 50-50 and
default 100 ), a statistically significant difference between choices of those accepting
the suggested contribution and those refusing it is observed for the second period
of the game (Wilcoxon Rank-Sum Test, p-value=0.024). In particular, the former
contribute on average 4.68, while the latter only 3.39. The same comparison for the
first and third period does not evidence any statistical difference (Wilcoxon RankSum Test, p-value> 0.1 for both comparisons). This joint evidence seems to suggest
that those choosing the suggested contribution in the second period of the game do
not a priori differ from those not choosing it and that it is the acceptance of the
suggested contribution that raises their choices in the second period of the game.
Concerning the impact of the form conditions (i.e., default and advice) on choices,
it is worthwhile to notice that those refusing the suggested contribution when it takes
the form of simple advice tend to contribute statistically more than those refusing
the suggested contribution when it assumes the form of a default option. This holds
5.3 Results
129
both when considering only the same source condition across form conditions (i.e.,
default 100 vs. advice 100 ) and when pooling data of the conditions with default
(i.e., default 50-50 and default 100 vs. advice 100 ; Wilcoxon Rank-Sum Test, pvalues< 0.001 for both comparisons). This, together with the absence of difference
in the amount given in the first period across the form conditions, seems to suggest
a “crowding-out” effect of the default among those refusing it. In other terms, the
presence of a binding device is likely to induce a more opportunistic behavior when
deviating from the suggested contribution. Unfortunately, the very few acceptances of
the suggested contribution observed in the absence of default inhibit us from performing any meaningful assessment of the impact of the default among these self-selected
sample of participants.
Beliefs
In each period of the game, beliefs about others’ contributions were elicited. This
allows us to assess the impact of the suggested contribution on the expectations about
partners’ contributions to the public good. Table 5.2 reports the average beliefs about
others’ contributions of subjects who contributed to the public good in the second
period (i.e., participants B).
Concerning the impact of the source of the suggested contribution, no significant
differences between beliefs in different source conditions (i.e., 100 and 50−50) are
detected (Wilcoxon Rank-Sum Tests, all p-values> 0.1). The only exception is represented by a significant difference between the third period beliefs of those refusing
the suggested contribution in default 50-50 and default 100 treatments (Wilcoxon
Rank-Sum Test, p-value=0.027). Overall, the source of the default seems not to
substantially affect the expectations about others’ behavior.
Concerning the impact of the form of the suggested contribution on beliefs, it
can be noticed that those refusing the suggested contribution maintain higher expectations about second period contributions when the suggested contribution takes
130
Chapter 5. Defaults as recommendations in public goods provision
Table 5.2: Average beliefs about others’ contributions
Treatment
def ault 100
def ault 50-50
advice 100
suggested contribution accepted
Period
I
II
III
I
II
III
I
II
III
N
14
14
14
14
14
14
3
3
3
Mean
4.548
3.619
3.214
4.405 4.381
3.333
3.556
4.222
3.333
Std.Dev.
1.683
1.307
1.448
2.117
2.249
2.051
1.388
1.388
1.667
Median
4.667
3.833
2.500
4.000 4.667
3.000
4.00
4.667
3.333
Min
1.333
1.333
1.667
0.667
0.333
0.000
2.00
2.667
1.667
Max
6.667
6.667
6.000
8.000
9.000
6.667
4.667
5.333
5.000
suggested contribution rejected
Period
I
II
III
I
II
III
I
II
III
N
18
18
18
18
18
18
29
29
29
Mean
3.500
3.204
2.815
3.907 3.889
4.204
4.552
5.402
5.230
Std.Dev.
2.011
1.893
1.961
1.771
1.455
1.630
2.736
2.168
2.537
Median
3.500
4.000
2.667
4.667 4.000
4.500
5.000
5.000
5.333
Min
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0
Max
7.000
6.667
7.000
6.333
6.333
6.333
10.000
10.000
10.000
the form of simple advice than when it takes the form of a default option. Indeed,
when comparing beliefs in the advice 100 and default 100 treatments, a statistically significant difference is detected (Wilcoxon Rank-Sum Test, p-value=0.001). A
significant difference of the same nature is registered also in the third period, but
not in the first (p-value=0.001 and p-value=0.192, respectively). Thus, refusing a
suggested contribution that is not strengthened by a default form seems to induce
more optimistic expectations about others’ contributions and this seems to extend
5.3 Results
131
also to the period following the presentation of the suggested contribution.
Concerning the comparison between contribution choices and beliefs about others’
contributions, Wilcoxon Signed-Rank Tests show differences between the two distributions, for those who accepted the suggested contribution, in treatment default 100
in the first and the third periods (p-value = 0.091 and p-value = 0.054, respectively)
and in treatment default 50-50 in the second period (p-value = 0.045). This suggests that, when it is certainly set by another participant, the default contribution
promotes the alignment between contributions and beliefs of those who stick with the
default. In contrast, when it is set by another participant only with 50% probability, the default contribution creates a divergence between contributions and beliefs of
those who stick with the default.
Regression analysis
To assess the impact of treatments and individual variables on the likelihood of accepting the suggested contribution, a Logistic regression was performed (see Table
5.3). As independent variables the two experimental conditions capturing the source
of the suggested contribution (Source) and the form of the suggested contribution
(Default) were considered. The variable Source is set equal to 1 when the suggested
contribution comes with certainty from participants A, and equal to 0 when it is with
50% probability set by participants A and with 50% probability a randomly drawn
number over the support of choices in the game. The variable Default is set equal to 1
when the suggested contribution assumes the form of a default option, and equal to 0
otherwise. The other two variables introduced, i.e., SC.observed and SC.obs-ChoiceI,
capture respectively the level of the suggested contribution that is observed by the
participants and the absolute distance between the suggested contribution observed
and the contribution made by the participants in the first period.
The regression analysis clearly shows that the presence of a default has a strong
132
Chapter 5. Defaults as recommendations in public goods provision
Table 5.3: Logistic regression
Accepts
Coeff (Std. Err.)
Source
0.107 (0.567)
Default
2.488 (0.748)***
SC.observed
-0.003 (0.106)
SC.obs-ChoiceI
-0.320 (0.128)**
intercept
-1.737 (1.001)*
Obs
96
Log-likelihood test (p-value)
∗∗∗
(1%);
0.0002
∗∗
(5%); ∗ (10%) significance level
positive impact on the decision to accept the suggested contribution (Odds Ratio=12.037, p-value< 0.001). In contrast, the distance between the suggested contribution observed and the contribution made by the participants in the first period
has a negative impact on the decision to accept the suggested contribution (Odds
Ratio=0.726, p-value=0.012). The remaining explanatory variables do not have a
statistically significant impact on the dependent variable: neither the level nor the
source of the suggested contribution seem to impact on the likelihood to accept the
suggested contribution.
Suggested contributions
Before knowing their actual role in the second period, all the participants were asked
to establish a contribution to suggest to a group in case they were assigned the role
A.
Table 5.4 reports some summary statistics of the differences at the individual level
between the suggested contributions and the contributions made in the first period.
5.4 Discussion and Concluding remarks
133
Table 5.4: Differences between suggested contributions and contributions in period 1
advice 100
default 50-50
default 100
N
40.00
40.00
40.00
Mean
0.05
0.57
0.93
Median
0.00
0.00
0.00
StdDev
3.00
2.87
2.99
Min
-10.00
-10.00
-5.00
Max
9.00
8.00
10.00
Wilcoxon Signed-Rank Test (p-value)
0.669
0.118
0.070
Wilcoxon Signed-Rank Tests (see p-values in Table 5.4) show that there are no significant differences between what participants suggested and what they contributed in
the previous period in treatments advice 100 and default 50-50. A marginal difference is detected in treatment default 100 : in the “strongest” situation, i.e., when the
suggested contribution is the default contribution and is established with certainty
by a participant, people tended to suggest a higher contribution than that they made
in the previous period.
5.4
Discussion and Concluding remarks
In this study we investigated whether the provision of a default option has an impact
on contributions to a public good and whether this impact persists even after the
default option has been removed. We also tested whether default effects can be
accounted for by the information conveyance of default options.
Unlike in the experiments presented in the previous chapter, it was clear that a
134
Chapter 5. Defaults as recommendations in public goods provision
default was present. For each group, the default option was established by a participant who was associated with that group, but did not make any contribution to the
public good; nevertheless, this participant had an interest in the public good since
her payoff was the average payoff of the members of the group.
We found a number of interesting results. First, we found a sizable default effect:
almost 44% of participants stuck with the suggested contribution when it assumed
the form of a default option, while only 9% of participants chose the suggested contribution when it assumed the form of a piece of advice. Interestingly, we found that
default options can have an impact on choices, i.e., influence people’s preferences, but
leave people’s expectations about others’ behaviour unaffected.
Second, when the suggested contribution assumed the form of a default option,
those who retained the default contributed, on average, more than those who did
not retain the default. By looking at preferences in the absence of a default (i.e.,
contributions in the first period), we can exclude that those who retained the default
were already more cooperative than those who did not retain the default. However,
we cannot say that those who retained the default became more cooperative, because
their contributions in the first period (i.e., in the absence of a default) and in the
second period (i.e., in presence of a default) did not differ. What we can say is that
the default increased the difference in the contributions of those who retained and
those who did not retain the default.
Third, those who did not choose the suggested contribution when it assumed the
form of a piece of advice contributed, on average, more than those who did not choose
the suggested contribution when it assumed the form of a default option. Considering
that contributions in the first period did not differ between the two groups, it seems
that the default created a type of “crowding-out” effect among those who did not
retain the default.
Fourth, no carryover effect was detected. Default effects did not result in a permanent change in contributions: for those who were affected by the default option,
5.4 Discussion and Concluding remarks
135
the effect disappeared as soon as the default was removed.
Fifth, the nature of the suggested contribution — i.e., established with certainty by
the participant associated with the group or established with 50% probability by this
participant and with 50% probability by the computer randomly — did not impact on
contribution choices: when the suggested contribution assumed the form of a default
option, the same fraction of participants retained the default regardless of who set the
default. However, the source of the default has an impact on the beliefs about others’
contributions of those who retained the default. Specifically, when the source of the
default had certain nature, contributions and beliefs were in line; in contrast, when
the source of the default had uncertain nature, contributions and beliefs diverged. A
potential interpretation of this pattern is that the decision-maker, although sticking
with the default, interprets the default with uncertain nature as a “noisy” signal that
does not systematically impact on the others’ choices. The expected impact of the
default on others’ choices, thus, seems to depend on the “reliability” of the signal.
Finally, participants coherently set the suggested contribution equal to their contribution in the first period, except in the “strongest” situation, i.e., when the suggested contribution was the default and was established with certainty by a participant: in this case people tended to suggest more than what they contributed in the
first period.
Our findings suggest that to be effective in strategic settings such as a Public
Goods Game, defaults must be evident. When they are provided through minimal
framing manipulations, as in the experiments reported in the previous chapter, defaults do not have any impact on strategic choices or, if they have it, the impact goes
in the opposite direction than that predicted by a default bias. In contrast, when
their presence is clearly evident to participants, defaults can cause a default bias, i.e.,
an exaggerated preference for the default option. However, it seems that this effect
is overall negative for cooperation; specifically, the default option does not function
as a mechanism to coordinate on a contribution. On the contrary, it enlarges the
136
Chapter 5. Defaults as recommendations in public goods provision
difference between the levels of contributions of those who retain the default and of
those who do not. In addition, it has a crowding-out effect on those who do not
retain it, since these respondents contribute less than those who do not choose the
suggested contribution when it assumes the form of simple advice.
The results reported here are encouraging. Nevertheless, much remains to be understood. Future research may investigate the effect of different sources of the default
contribution, such as an experienced participant that performed well in previous public goods experiments, a participant that performed better than the others in a task,
or a participant elected by the others. One might expect to find an even stronger
default effect, since these sources have a certain degree of “legitimacy” relative to a
simple random appointment as in the present experiment.
Another line of research may also investigate the impact of default contributions
in Public Goods Games with different characteristics, such as repetition, the presence
of a provision point, or the presence of heterogeneity in the initial endowments or in
the return on the public good. In a Threshold Public Goods Game, Croson and Marks
(2001) found that recommendations are effective in an environment with heterogeneous returns on the public good, possibly because heterogeneity makes coordination
more complex. If this is the case, one might expect to find an even stronger default
effect in such an environment.
Interesting would also be to check whether there is any relationship between the
occurrence of a default bias and the cognitive abilities of those who fall prey to the
default bias. A recent and fast growing line of research investigates how cognitive
skills are related to preferences and behavioural biases (e.g., Dohmen et al., 2007;
Oechssler et al., 2008; Burks et al., 2008; Jones, 2008). For example, Oechssler et al.
(2008) found lower incidences of certain biases, such as the conjunction fallacy and
overconfidence, in individuals with greater cognitive skills. Knowing whether the
incidence of the default bias is related with cognitive abilities may help deep our
understanding of the mechanisms underlying the default bias.
5.4 Discussion and Concluding remarks
137
Further research is needed to understand why default effects occur in public goods
provision. The experiments reported in the present work suggest that information
conveyed by the default option and cognitive load, under which decision-makers may
be while deciding, do not account for default effects. In fact, these effects occurred also
when the default was set randomly by the computer and no restrictions of cognitive
resources were made.
Appendix
Experimental instructions
This appendix contains the instructions (originally in Italian) we used for treatment
advice 100 of Experiment II. The three experimental treatments differ only in Round
2. Differences in the def ault 100 and def ault 50-50 treatments are indicated in the
text. Participants received the instructions for a round at the beginning of that
specific round.
Dear Participant,
thank you for taking part in this experiment. During the experiment all communication is prohibited. If you have a question please raise your hand, and an experimenter
will answer your questions individually.
The experiment
The experiment consists of three rounds. At the beginning of each round the instructions relative to that round will be displayed on your screen. During the experiment
you will be randomly matched with other participants so that you will form a group.
You will never know the identity of the other three members of your group. Similarly,
your identity will never be revealed to the others.
Your earnings
Your earnings depend both on the decisions you have individually and autonomously
taken and on the decisions taken by your group members. In the instructions for each
round you will find detailed information about how your earnings will be calculated.
The three round are independent, in the sense that the decisions taken in a specific
round influence only earnings in that round and not earnings in the other two rounds.
138
At the end of the experiment one of the three rounds will be randomly selected. This
selection will be relevant for the computation of your final earnings, in a way that will
be explained in details at the beginning of each round. In addition to your earnings
from the experiment, you will receive 3 Euro for having shown up on time.
Click on the OK button only when you have finished to read the instructions
carefully, since it will not be possible to go back.
Round 1 [and Round 3]
In this round you will be randomly matched with other three participants and, thus,
you will form a 4-person group.
[In Round 3 participants read: In this round you will be matched with the same three
participants of Round 1 and Round 2 and, thus, you will form a 4-person group. ]
You will never know the identity of the other three members of your group. Similarly,
your identity will never be revealed to the others.
Your tasks
1) The contribution to a project
You will receive 10 Experimental Currency Units (ECU). You have to decide, individually and autonomously, how many of the 10 ECU you want to contribute to a
project. The other three group members will be asked to make the same decision.
Whatever you do not contribute will be deposited on your personal account.
The project produces an income that is equally divided among the four group members.
Your earnings for this task are calculated as the sum of two parts:
- the ECU deposited on your personal account;
139
- the income from the project.
For example, defining you as X and the other group members as Y, Z, and W, your
earnings will be equal to
Earnings X = 10 ECU −Contribution X+0.4∗(Contribution X+Contribution Y +
Contribution Z + Contribution W )
2) Your estimates
In addition to make a contribution decision, you will be asked to make estimates
about the other three group members’ contributions. Specifically, you will be asked
to type in three values relative to the contributions of the other three group members.
Your earnings from this task are calculated as follows:
- If the sum of contributions you estimated is equal to the actual sum of contributions of the other three group members, you will earn 3 ECU;
- If the sum of contributions you estimated is different from the actual sum of
contributions of the other three group members, you will earn an amount of
ECU calculated by the following formula:
1.5ECU
|estimated sum of contributions−actual sum of contributions|
Your final earnings from this round
If this round will be randomly selected, you will be paid for task 1), that is the
contribution decision. If this round will not be randomly selected, you will be paid
for task 2), that is the estimation task. Each ECU will be converted into 0.50 EURO.
For example, if you earn 2 ECU you will receive 1 Euro.
140
Your information
You will receive information about the contributions of the other three group members, and thus information about your earnings, only at the end of the experiment
and not at the end of each round. Similarly, you will be informed about the two
rounds selected for payment only at the end of the experiment, since the random
selection will be made at the end of the experiment.
Click on the OK button only when you have finished to read the instructions
carefully, since it will not be possible to go back.
Round 2
In this round you will be matched with the same three participants of Round 1 and,
thus, you will form a 4-person group. You will never know the identity of the other
three members of your group. Similarly, your identity will never be revealed to the
others.
Roles
In this round there are two different roles: role A and role B. Of the five groups
present in the room, one group (group A) will be randomly selected and the members
of that group will be assigned the role A, while all the members of the four groups
not selected (groups B) will be assigned the role B.
Each of the four members of group A will be associated with one of the four groups B.
ROLE B
In case you are a member of groups B, here below you find detailed information about
your tasks, the payment of your earnings and the information you will receive.
141
Your tasks
1) The contribution to a project
You will receive 10 Experimental Currency Units (ECU). You have to decide, individually and autonomously, how many of the 10 ECU you want to contribute to a
project. The other three group members will be asked to make the same decision.
The participant in the role A associated with your group proposes a contribution,
that is a value included in the range 0-10, that will be communicated to you and
the other three group members. You and the other three group members can decide,
individually and autonomously, how much to contribute to the project.
[Participants in the def ault 100 treatment read:
The participant in the role A associated with your group set an automatic contribution, that is a value included in the range 0-10, that each member of your group
will automatically contribute, unless she decides to change her own contribution.
Therefore, you and the other three group members can decide, individually and autonomously, whether to confirm the automatic contribution or not to confirm it and
choose a different contribution. ]
[Participants in the def ault 50-50 treatment read:
The participant in the role A associated with your group set an automatic contribution, that is a value included in the range 0-10, that each member of your group will
automatically contribute, unless she decides to change her own contribution. The
same value of the automatic contribution will be display on your screen and those of
the other three group members. This value will be with 50% probability the value set
by the participant in the role A associated with your group and with 50% probability
a value randomly established by the computer. You and the other three group members can decide, individually and autonomously, whether to confirm the automatic
142
contribution or not to confirm it and choose a different contribution.]
Whatever you do not contribute will be deposited on your personal account.
The project produces an income that is equally divided among the four group members.
Your earnings for this task are calculated as the sum of two parts:
- the ECU deposited on your personal account;
- the income from the project.
For example, defining you as X and the other group members as Y, Z, and W, your
earnings will be equal to
Earnings X = 10 ECU −Contribution X+0.4∗(Contribution X+Contribution Y +
Contribution Z + Contribution W )
2) Your estimates
In addition to make a contribution decision, you will be asked to make estimates
about the other three group members’ contributions. Specifically, you will be asked
to type in three values relative to the contributions of the other three group members.
Your earnings from this task are calculated as follows:
- If the sum of contributions you estimated is equal to the actual sum of contributions of the other three group members, you will earn 3 ECU;
- If the sum of contributions you estimated is different from the actual sum of
contributions of the other three group members, you will earn an amount of
ECU calculated by the following formula:
1.5ECU
|estimated sum of contributions−actual sum of contributions|
143
Your final earnings from this round
If this round will be randomly selected, you will be paid for task 1), that is the
contribution decision. If this round will not be randomly selected, you will be paid
for task 2), that is the estimation task. Each ECU will be converted into 0.50 EURO.
For example, if you earn 2 ECU you will receive 1 Euro.
Your information
You will receive information about the contributions of the other three group members, and thus information about your earnings, only at the end of the experiment
and not at the end of each round. Similarly, you will be informed about the two
rounds selected for payment only at the end of the experiment, since the random
selection will be made at the end of the experiment.
ROLE A
Each of the four members of group A will be associated with one of the four groups B.
In case you are a member of groups A, here below you find detailed information about
your tasks, the payment of your earnings and the information you will receive.
Your tasks
1) Propose a contribution
[Participants in the def ault 100 and in the def ault 50-50 treatments read : 1) Set
an automatic contribution]
The four members of the group associated with you are asked how many of the 10
ECU at their disposal they want to contribute to a project, as described above in
task 1) of role B.
You can propose a contribution, that is a value included in the range 0-10, that will
144
be communicated to the four members of the group associated with you.
[Participants in the def ault 100 treatment read : You can set an automatic contribution, that is a value included in the range 0-10, that each member of the group
associated with you will automatically contribute, unless she decides to change her
own contribution. ]
[Participants in the def ault 50-50 treatment read: You can set an automatic contribution, that is a value included in the range 0-10, that each member of the group
associated with you will automatically contribute, unless she decides to change her
own contribution. The same value of the automatic contribution will be display on
the screen of the four members of the group associated with you. This value will
be with 50% probability the value that you set and with 50% probability a value
randomly established by the computer. ]
Your earnings from this task will be equal to the average earnings of the four members
of the group B associated with you. For example, defining you as A and the four
members of the group associated with you as X, Y, Z e W, your earnings will be
equal to
Earnings A =
(Earnings X+Earnings Y +Earnings Z+Earnings W )
4
2) Your estimates
In addition to propose a contribution,
[Participants in the def ault 100 and in the def ault 50-50 treatments read : In addition to set an automatic contribution,]
you will be asked to make estimates about the contributions made by the four members of the group associated with you. Specifically, you will be asked to type in four
145
values relative to the contributions of the four members of the group associated with
you.
Your earnings from this task are calculated as follows:
- If the sum of contributions you estimated is equal to the actual sum of contributions made by the four members of the group associated with you, you will
earn 3 ECU;
- If the sum of contributions you estimated is different from the actual sum of
contributions made by the four members of the group associated with you, you
will earn an amount of ECU calculated by the following formula:
1.5ECU
|estimated sum of contributions−actual sum of contributions|
Your final earnings from this round
If this round will be randomly selected, you will be paid for task 1).
If this round will not be randomly selected, you will be paid for task 2), that is the
estimation task.
Each ECU will be converted into 0.50 EURO.
Your information
You will receive information about the contributions of the four members of the group
associated with you, and thus information about your earnings, only at the end of
the experiment and not at the end of each round. Similarly, you will be informed
about the two rounds selected for payment only at the end of the experiment, since
the random selection will be made at the end of the experiment.
Note that before knowing your role, you will be asked to propose a contribution that
will be communicated only in case you are assigned the role A.
146
[Participants in the def ault 100 and in the def ault 50-50 treatments read: Note
that before knowing your role, you will be asked to set an automatic contribution
that will be effective only in case you are assigned the role A.]
Click on the OK button only when you have finished to read the instructions carefully,
since it will not be possible to go back.
147
148
Chapter 6
Summary and Concluding Remarks
6.1
Summary and future research
The present work provides an experimental analysis of non-normative behaviour in
two typical examples of strategic interaction: ultimatum bargaining and public goods
provision. Most part of the study adopted a dual-system perspective, which has
recently received great attention in economics. According to this perspective, individual behaviour results from the interaction of two different systems, which operate
according to different rules and have potentially conflicting motivations. Behaviour
driven by the deliberative system derives from deliberation and is more goal-oriented,
forward-looking, and largely unaffected by affective states. It is often thought to be
close to that postulated by standard economic theory (Lobel and Loewenstein, 2005).
In contrast, behaviour driven by the affective system is more instinctive, myopic,
based on intuition, and influenced by affective states and environmental stimuli. The
final behaviour depends on which of the two system prevails.
Chapter 3 presented an experiment designed to investigate the behaviour of proposers and responders in the Ultimatum Game when the affective system is likely
to prevail on the deliberative one. The deliberative system was weakened through
time pressure and cognitive load, and the latter was manipulated through a dual-task
149
150
Chapter 6. Summary and Concluding Remarks
procedure, commonly used in psychology, in which decision-making and a memory
task are performed simultaneously. Decisions were elicited through the strategy vector method, which allowed us to collect information about decisions as both proposer
and responder for each individual. On the responder side, our results show that, when
deciding under time pressure, responders are more likely to reject. Thus, when the
deliberative system is less capable to intervene into the process to modify or override
the responses of the affective system, responders demand more from the proposers,
suggesting that greater reliance on the affective system strengthens fairness concerns.
This is in line with previous research investigating affective responses as a “hot” reaction to a specific proposal (e.g., Pillutla and Murnighan, 1996; Sanfey et al., 2003;
Sutter et al., 2003; van´ t Wout et al., 2006; Koenigs and Tranel, 2007). The results
presented here add to this body of research by showing that affective processes matter
not only in “hot” environments, but also in “cold” environments where decisions have
a hypothetical character.
On the proposer side, our results show that, when deciding under time pressure,
proposers make higher offers. Ultimatum offers may have two components, one driven
by fear of rejection and the other driven by other-regarding concerns. When strategic
thinking is disrupted by the taxation of cognitive resources, the affective system,
which relies more on heuristics, may prompt proposers to apply a equality heuristic.
About the other-regarding component, greater reliance on the affective system may
either promote other-regarding concerns – as postulated by the intuitionist approach
(“other-regardings are instinctive”) – or disrupt other-regardings – as postulated by
the rationalist approach (“self-interest is instinctive”). One might expect the former
approach to find support: since the affective system strengthens fairness concerns
in responders, it is plausible that it has the same effect on proposers’ behaviour.
However, it has to be noticed that the decision faced by proposers and that faced
by responders are inherently different: the responder’s decision is simply a choice
between accepting and rejecting an offer and misses the strategic component that
6.1 Summary and future research
151
characterizes the proposer’s decision. Therefore, it might be that the additional
strategic component generates different affective processes and decision outcomes.
Having information about decisions in both roles for each individual helps us shed
light on this point. Considering the false consensus effect, i.e., the tendency of people
to think that others behave as they do, we may expect that the proposer thinks
the minimum threshold of acceptance (MTA) of the responder to equal her own
MTA when playing the role of responder. All that is offered above the MTA can
be considered as other-regarding. Our results show that under time pressure MTAs
are higher, but on average the difference between the offer and the MTA, although
positive, remains constant across different time-pressure conditions. This suggests
that the increase in offers registered under time pressure is driven by the strategic
component, while the other-regarding does not vary. Proposers’ beliefs of acceptance
of the offer made provide support to this finding: although offers are higher under
time pressure, proposers do not perceive them as more likely to be accepted, reflecting
an expected increase in the MTA of the responder.
These considerations are drawn considering potential false consensus effects. To
shed more light on the weight of the other-regarding and the strategic components
when the affective system prevails, it would be helpful to compare ultimatum offers
with allocations made in a Dictator Game – in which the strategic component of
the decision is removed – when decision-makers are under time pressure. This is a
natural extension of the experiment presented here. Further research is also needed
to understand the lack of effect of the cognitive load manipulation that we reported
in this experiment, both on the proposers’ and on the responders’ behaviour. Unlike
psychology studies, we introduced financial incentives for the second task. Further
research is needed to test the robustness of the cognitive load manipulation with
respect to factors such as the provision of incentives or the characteristics of the
environment in which decision-makers operate.
152
Chapter 6. Summary and Concluding Remarks
Chapter 4 presented two exploratory experiments designed to investigate the presence of default effects in the provision of public goods. Defaults have been found to
influence individual behaviour in many non strategic situations, since people show
a tendency to exaggeratedly prefer the default option. However, the potential effects of defaults in strategic situations have not been investigated yet. We focused
on a broadly investigated social interaction, namely the public goods provision, and
explored, among other explanations put forward in the literature, whether having
constrained cognitive resources is relevant for default effects in this kind of contexts.
In the two experiments presented in this chapter we constrained cognitive resources
through cognitive load, manipulated using the dual-task procedure commonly adopted
in psychology. This manipulation should make behaviour more influenced by the affective system, which relies more on heuristics and is susceptible to salient environmental
characteristics, such as the presence of a default contribution. In both experiments
the default contribution was established using the same method adopted by Johnson
et al. (2002) in Internet privacy policy adoption: all the possible options were displayed and one of them was already selected. In the first experiment we used a linear
Public Goods Game and set the default contribution alternatively at the minimum
and the maximum possible contribution. In the second experiment we used a Threshold Public Goods Game with a utilization rebate rule for exceeding contributions and
set the default at a focal value, i.e., the individual contribution necessary to exactly
reach the threshold when all subjects make an equal contribution.
In the first experiment we did not detect any default effect, neither when cognitive resources were unconstrained nor when they were constrained. In the second
experiment we found that, when cognitive resources were constrained, the presence
of a default contribution had an effect on choices, but this effect is opposite to that
predicted by the default bias. In fact, people tended to contribute the focal value
when it was not the default contribution and to contribute a higher value when the
6.1 Summary and future research
153
focal value was the default contribution. Thus, we found that the presence of a default contribution increased cooperation not because people stick with the default
contribution, but rather because people do not stick with it. However, this effect
was transient, since it was not able to change preferences in a permanent way: it
disappeared once the default had been removed.
These experiments have exploratory nature, and the results do not provide a
satisfactory answer to whether default biases are present also in strategic interaction.
Further research is needed to understand whether and under what circumstances
default options affect contribution behaviour in the public goods provision. Future
lines of research may investigate default effects under other circumstances – such as
the absence of a money-back guarantee, the use of different rebate rules for exceeding
contributions, or the presence of heterogeneity in the return on the public good –
and with different mechanisms than the simple minimal framing manipulation used
here. In addition, the study of response time (Rubinstein, 2007) could help better
understand the processes underlying potential default biases in public goods provision.
Chapter 5 extended the exploratory analysis of default effects in the provision of
public goods by testing whether information conveyed by the default option, i.e., the
implicit recommendation given by those who set the default, can account for default
effects. In the experiment presented, a participant could set a contribution for a
group of contributors associated with her. This participants did not have to make
any contribution, but had an interest in the average performance of the group. The
contribution set assumed either the form of the default contribution, i.e., what each
group member automatically contributed unless she specified a different contribution,
or the form of a simple suggestion. To check for the relevance of the source of the
default contribution, i.e., who set the default, the default contribution communicated
to the group was either surely that established by the participant associated with that
group or only with 50% probability that established by this participant (and with the
154
Chapter 6. Summary and Concluding Remarks
50% probability set by the computer randomly).
We found a sizable default effect: a significantly higher number of people stuck
with the established contribution when it assumed the form of default than when it
assumed the form of simple suggestion, independently of the level of the established
contribution. In addition, the default effect was independent of the source of the
default, i.e., who set the default: when the established contribution assumed the form
of default, the same fraction of people stuck with the default contribution regardless of
who set the default. The default contribution did not change the expectations about
others’ contributions. The presence of a default contribution was able to change
people’s preferences, but not in a permanent way: the default effect disappeared once
the default was removed. The data present an intriguing pattern. Future research
may investigate the effect of defaults in other public goods circumstances – such as
with repetition, with the presence of a provision point or with heterogeneity in the
initial endowments or in the return on the public good – and with different sources of
the default contribution – such as a participant that performed better than the others
in a task, a participant that got a high payoff in previous public goods experiments, or
a participant that is chosen by the others. Interesting would also be to check whether
default biases are related with cognitive skills, since other kinds of biases have been
found to be related to them (e.g., Oechssler et al., 2008). In general, the effect of
default options in other kinds of strategic interaction seems to be a promising area of
research.
6.2
Concluding remarks
To conclude, from what emerged from the present work, two main general considerations deserve to be remarked.
First, it is important to consider affective processes in attempt to better understand economic behaviour. As Angner and Loewenstein (200x) stress, “large domains
6.2 Concluding remarks
155
of economic behavior will remain outside of the range of economic models unless
economists begin to get a grip on the role of emotions in behavior” (pp. 51-52). The
effect of affective states on economic behaviour has been demonstrated in a wide range
of situations, and the results obtained in our experiment show that they are able to
influence behaviour also in “cold” situations, where their effects are expected to be
greatly reduced. Frantz (2006) noticed that “economists [...] understood that much
is left out of economists’ theory of human behavior, but that this is inescapable in
our attempt to be scientific” (p. 54). Progresses in the field of neuroscience have led
to the development of an array of tools that allow objective measurements of affective
states. The opportunity to identify and measure the physiological changes and the
activity in the neural areas associated with affective processes may represents a way
for economists to deal with affective states in a more scientific manner.
If affective processes appear to matter in an aseptic environment such as the
laboratory, even more so should they matter in the real world. Apparently irrelevant
factors have, through their impact on affective states, an influence on behaviour that is
large enough to be observable in real and important decisions. For example, analyses
made on large amounts of data showed that sunshine, which has a positive impact
on mood, is strongly significantly correlated with daily returns in the stock-market
(Hirshleifer and Shumway, 2003) and the probability of certain types of candidates
to be admitted at the university (“clouds make nerds look good”, Simonsohn, 2007).
Second, the dual-system approach used in most part of the present work, although
inevitably approximative, appears to be a useful way of investigating economic behaviour. Since the affective and the deliberative systems have potentially conflicting
motivations, understanding the conditions under which one of the two systems is
more likely to prevail can be useful for marketers in defining selling strategies, for
policy-makers in defining better policy interventions, and for people themselves in
defining precautionary and commitment strategies. For example, greater reliance on
the affective system leads people to spend money impulsively, to be more susceptible
156
Chapter 6. Summary and Concluding Remarks
to environmental stimuli, and more generally to be less able to resist temptations.
Retailers can exploit these observations when defining their selling strategies.
Bibliography
Akerlof, G. A. (1982). Labor contracts as partial gift exchange. Quarterly Journal of
Economics, 97, 543–569.
Anderson, L. R., Mellor, J. M., and Milyo, J. (2008). Inequality and public good
provision: An experimental analysis. Journal of Socio-Economics, 37, 1010 – 1028.
Andreoni, J. (1988). Why free ride? Journal of Public Economics, 37, 291 – 304.
Andreoni, J. (1995). Warm-glow versus cold-prickle: The effects of positive and
negative framing on cooperation in experiments. Quarterly Journal of Economics,
110(1), 1 – 21.
Andreoni, J. and Petrie, R. (2004). Public goods experiments without confidentiality:
a glimpse into fund-raising. Journal of Public Economics, 88, 1605 – 1623.
Angner, E. and Loewenstein, G. (200x). Behavioral economics. In U. Mäki, editor,
Handbook of the Philosophy of Science, volume 13 of Philosophy of Economics.
Amsterdam: Elsevier. Available at http://papers.ssrn.com/sol3/papers.cfm?
abstract_id=957148 (Accessed January 23, 2009).
Ashley, R., Ball, S., and Eckel, C. (2008). Motives for giving: A reanalysis of
two classic public goods experiments.
Center for Behavioral and Experimen-
tal Economic Science (CBEES) Working Paper , No. 08 - 01.
Available at
http://cbees.utdallas.edu/papers.php (Accessed December 9, 2008).
157
158
BIBLIOGRAPHY
Ashraf, N., Camerer, C. F., and Loewenstein, G. (2005). Adam smith, behavioral
economist. Journal of Economic Perspectives, 19(3), 131–145.
Bechara, A. and Damasio, A. R. (2005). The somatic marker hypothesis: A neural
theory of economic decision. Games and Economic Behavior , 52, 336–372.
Becker, G. S. (1974). A theory of social interactions. The Journal of Political Economy, 82(6), 1063–1093.
Bell, D. E. (1982). Regret in decision making under uncertainty. Operations Research,
30, 961 – 981.
Bell, D. E. (1985). Disappointment in decision making under uncertainty. Operations
Research, 33, 1 – 27.
Ben-Ner, A., Kong, F., and Putterman, L. (2004). Share and share alike? genderpairing, personality, and cognitive ability as determinants of giving. Journal of
Economic Psychology, 25, 581–589.
TM
Benartzi, S. and Thaler, R. (2004). Save more tomorrow
: Using behavioral eco-
nomics to increase employee saving. Journal of Political Economy, 112, 164–187.
(supplement).
Benhabib, J. and Bisin, A. (2005). Modeling internal commitment mechanisms and
self-control: A neuroeconomics approach to consumption-saving decisions. Games
and Economic Behavior , 52(2), 460–492.
Benjamin, D. J., Brown, S. A., and Shapiro, J. M. (2006). Who is behavioral?
cognitive ability and anomalous preferences. Harvard University Working Paper .
Available at http://www.dklevine.com/archive/refs4122247000000001334.
pdf (accessed March, 09, 2007).
Bernheim, B. D. and Rangel, A. (2004). Addiction and cue-triggered decision processes. American Economic Review , 94(5), 1558–1590.
BIBLIOGRAPHY
159
Beshears, J., Choi, J. J., Laibson, D., and Madrian, B. C. (2008). The importance of
default options for retirement savings outcomes: Evidence from the united states.
In S. J. Kay and T. Sinha, editors, Lessons from Pension Reform in the Americas,
pages 59 – 87. Oxford: Oxford University Press.
Blaney, P. L. (1986). Affect and memory: A review. Psychological Bulletin, 99,
229–246.
Bodenhausen, G. V. (1990). Stereotypes as judgmental heuristics: Evidence of circadian variations in discrimination. Psychological Science, 1(5), 319 – 322.
Bohnet, I. and Zeckhauser, R. (2004). Social comparisons in ultimatum bargaining.
Scandinavian Journal of Economics, 106(3), 495–510.
Bolton, G. E. and Ockenfels, A. (2000). Erc: A theory of equity, reciprocity and
competition. The American Economic Review , 90(1), 166–193.
Bornstein, G. and Yaniv, I. (1998). Individual and group behavior in the ultimatum
game: Are groups more ”rational” players? Experimental Economics, 1, 101–108.
Bosman, R., Sonnemans, J., and Zeelenberg, M. (2001). Emotions, rejections, and
cooling off in the ultimatum game. Working paper . CREED Department of Economics University of Amsterdam. Available at: http://www1.fee.uva.nl/CREED/
pdffiles/coolingoff.pdf (accessed July 29, 2007).
Bower, G. H. and Cohen, P. R. (1982). Emotion influences in memory and thinking:
Data and theory. In M. Clark and S. Fiske, editors, Affect and cognition. Hillsdale,
NJ: Erlbaum.
Brandstätter, H. and Güth, W. (2002). Personality in dictator and ultimatum games.
Central European Journal of Operations Research, 10(3), 191–215.
160
BIBLIOGRAPHY
Bruyneel, S., Dewitte, S., Vohs, K. D., and Warlop, L. (2006). Repeated choosing increases susceptibility to affective product features. International Journal of
Research in Marketing, 23, 215–225.
Buckley, E. and Croson, R. (2006). Income and wealth heterogeneity in the voluntary
provision of linear public goods. Journal of Public Economics, 90, 935 – 955.
Burks, S. V., Carpenter, J. P., Götte, L., and Rustichini, A. (2008). Cognitive skills
explain economic preferences, strategic behavior, and job attachment. IZA Discussion Paper No. 3609 . Available at http://ftp.iza.org/dp3609.pdf (Accessed
September 18, 2008).
Bütler, M. and Teppa, F. (2007). The choice between an annuity and a lump sum:
Results from swiss pension funds. Journal of Public Economics, 91, 1944 – 1966.
Cadsby, C. B. and Maynes, E. (1999). Voluntary provision of threshold public goods
with continuous contributions: experimental evidence. Journal of Public Economics, 71, 53 – 73.
Camerer, C., Loewenstein, G., and Prelec, D. (2005). Neuroeconomics: How neuroscience can inform economics. Journal of Economic Literature, 43, 9–64.
Camerer, C. F. (2000). Prospect theory in the wild: Evidence from the field. In
D. Kahneman and A. Tversky, editors, Choices, Values and Frames, pages 288–
300. Cambridge: Cambridge University Press.
Camerer, C. F. (2003). Behavioral Game Theory: Experiments in Strategic Interaction. New York: Russell Sage Foundation.
Camerer, C. F. and Loewenstein, G. (2004). Behavioral economics: Past, present,
future. In C. F.Camerer, G. Loewenstein, and M. Rabin, editors, Advances in
Behavioral Economics, chapter 1, pages 3–51. Princeton University Press.
BIBLIOGRAPHY
161
Cameron, L. A. (1999). Raising the stakes in the ultimatum game: Experimental
evidence from indonesia. Economic Inquiry, 37(1), 47 – 59.
Caplin, A. and Leahy, J. (2001). Psychological expected utility theory and anticipatory feelings. Quarterly Journal of Economics, 116(1), 55 – 79.
Carpenter, J., Verhoogen, E., and Burks, S. (2005). The effect of stakes in distribution
experiments. Economics Letters, 86, 393 – 398.
Carter, J. R. and Irons, M. D. (1991). Are economists different, and if so, why?
Journal of Economic Perspectives, 5(2), 171–177.
Cason, T. N. and Khan, F. U. (1999). A laboratory study of voluntary public goods
provision with imperfect monitoring and communication. Journal of Development
Economics, 58, 533 – 552.
Chan, K., Mestelman, S., Moir, R., and Muller, R. (1999). Heterogeneity and the
voluntary provision of public goods. Experimental Economics, 2, 5 – 30.
Charness, G. and Dufwenberg, M. (2006). Promises and partnership. Econometrica,
74(6), 1579 – 1601.
Charness, G. and Gneezy, U. (2008). Whats in a name? anonymity and social distance
in dictator and ultimatum games. Journal of Economic Behavior & Organization,
68, 29–35.
Cherry, T. L., Frykblom, P., and Shogren, J. F. (2002). Hardnose the dictator.
American Economic Review , 92(4), 1218–1221.
Cherry, T. L., Kroll, S., and Shogren, J. F. (2005). The impact of endowment heterogeneity and origin on public good contributions: evidence from the lab. Journal
of Economic Behavior & Organization, 57, 357 – 365.
162
BIBLIOGRAPHY
Choi, J. J., Laibson, D., Madrian, B. C., and Metrick, A. (2004). For better or
for worse: Default effects and 401(k) savings behavior. In D. A. Wise, editor,
Perspectives on the Economics of Aging, pages 81 – 121. Chicago: University of
Chicago Press.
Cinyabuguma, M., Page, T., and Putterman, L. (2005). Cooperation under the threat
of expulsion in a public goods experiment. Journal of Public Economics, 89, 1421
– 1435.
Clark, J. (2002). House money effects in public good experiments. Experimental
Economics, 5, 223 – 231.
Cohen, J. D. (2005). The vulcanization of the human brain: a neural perspective
on interactions between cognition and emotion. Journal of Economic Perspectives,
19(4), 3–24.
Cornelissen, G., Dewitte, S., and Warlop, L. (2007). Social value orientation as a
moral intuition: Decision-making in the dictator game. Economics Working Papers
series. no. 1028, Department of Economics and Business, Universitat Pompeu
Fabra. Available at http://www.econ.upf.edu/docs/papers/downloads/1028.
pdf (accessed July 28, 2007).
Croson, R. (1998). Feedback in voluntary contribution mechanisms: an experiment
in team production. In R. Isaac, editor, Research in Experimental Economics. JAI
Press, Greenwich, CT.
Croson, R. and Marks, M. (2001). The effect of recommended contributions in the
voluntary provision of public goods. Economic Inquiry, 39(2), 238–249.
Croson, R. and Miller, M. (2005). Exploring the relationship between actions and beliefs: Projection vs. reaction. Working Paper . Available at http://opim.wharton.
upenn.edu/~crosonr/research/%5B58%5D.pdf (accessed July 30, 2008).
BIBLIOGRAPHY
163
Croson, R. T. (1996a). Information in ultimatum games: An experimental study.
Journal of Economic Behavior & Organisation, 30, 197 – 212.
Croson, R. T. (1996b). Partners and strangers revisited. Economics Letters, 53, 25
– 32.
Croson, R. T. (2000). Thinking like a game theorist: factors affecting the frequency
of equilibrium play. Journal of Economic Behavior & Organization, 41, 299 – 314.
Croson, R. T. A. (2007). Theories of commitment, altruism and reciprocity: Evidence
from linear public goods games. Economic Inquiry, 45(2), 199 – 216.
Damasio, A. R. (1994). Descartes error: emotion, reason and the human brain.
Papermac, London.
Davis, D. D. and Holt, C. A. (1993). Experimental Economics. Princeton University
Press.
Dawes, R., Orbell, J., Simmons, R., and van de Kragt, A. (1986). Organizing groups
for collective action. American Political Science Review , 80, 1171 – 1185.
DellaVigna, S. and Malmendier, U. (2006). Paying not to go to the gym. American
Economic Review , 96, 694 – 719.
Dewitte, S., Verguts, T., and Lens, W. (2003). Thinking hard can break your will:
The role of working memory in ego depletion. running head: ego depletion and
working memory. DTEW Research Report 0338, K.U.Leuven.
Dhar, R. and Nowlis, S. M. (1999). The effect of time pressure on consumer choice
deferral. Journal of Consumer Research, 25, 369 – 384.
Dohmen, T., Falk, A., Huffman, D., and Sunde, U. (2007). Are risk aversion and impatience related to cognitive ability? Available at http://ftp.iza.org/dp2735.pdf
(Accessed September 6, 2008).
164
BIBLIOGRAPHY
Dolan, R. (2002). Emotion, cognition, and behavior. Science, 298, 1191–1194.
Drolet, A. and Luce, M. F. (2004). The rationalizing effects of cognitive load on
emotion-based trade-off avoidance. Journal of Consumer Research, 31, 63 – 77.
Duffy, J. and Feltovich, N. (1999). Does observation of others aect learning in strategic
environments? an experimental study. International Journal of Game Theory, 28,
131–152.
Eckel, C. C. and Grossman, P. J. (2001). Chivalry and solidarity in ultimatum games.
Economic Inquiry, 39(2), 171 – 188.
Edland, A. and Svenson, O. (1993). Judgement and decision making under time
pressure. In O. Svenson and A. J. Maule, editors, Time Pressure and Stress in
Human Judgment and Decision Making, chapter 2, pages 27 – 40. Plenum Press,
New York.
Elster, J. (1998). Emotions and economic theory. Journal of Economic Literature,
36(1), 47 – 74.
Epstein, S. (1994). Integration of the cognitive and the psychodynamic unconscious.
American Psychologist, 49(8), 709–724.
Evans, J. S. (1998). Matching bias in conditional reasoning: Do we understand it
after 25 years? Thinking and Reasoning, 4(1), 45 – 82.
Evans, J. S. (2003). In two minds: dual-process accounts of reasoning. TRENDS in
Cognitive Sciences, 7(10), 454–459.
Evans, J. S. and Over, D. (1996). Rationality and Reasoning (Essays in Cognitive
Psychology). Psychology Press.
Evans, J. S. B. T. (1972). Interpretation and matching bias in a reasoning task.
Quarterly Journal of Experimental Psychology, 24(2), 193 – 199.
BIBLIOGRAPHY
165
Fehr, E. and Gächter, S. (2000). Cooperation and punishment in public goods experiments. American Economic Review , 90, 980 – 994.
Fehr, E. and Schmidt, K. M. (1999). A theory of fairness, competition and cooperation. The Quarterly Journal of Economics, 114, 817–868.
Fehr, E., Kirchler, E., Weichbold, A., and Gächter, S. (1998). When social norms
overpower competition: Gift exchange in experimental labor markets. Journal of
Labor Economics, 16(2), 324–351.
Finucane, M. L., Alhakami, A., Slovic, P., and Johnson, S. M. (2000). The affect
heuristic in judgments of risks and benefits. Journal of Behavioral Decision Making,
13, 1 – 17.
Fischbacher, U. (2007). Z-tree: Zurich toolbox for ready-made economic experiments.
Experimental Economics, 10(2), 171–178.
Fisher, J., Isaac, R. M., Schatzberg, J. W., and Walker, J. M. (1995). Heterogenous
demand for public goods: Behavior in the voluntary contributions mechanism.
Public Choice, 85, 249 – 266.
Frantz, R. (2006). Intuition in behavioral economics. In M. Altman, editor, Handbook
of Contemporary Behavioral Economics. M.E. Sharpe.
Frederick, S. (2005). Cognitive reflection and decision making. Journal of Economic
Perspectives, 19(4), 25–42.
Frederick, S., Loewenstein, G., and O’Donoghue, T. (2002). Time discounting and
time preference: A critical review. Journal of Economics Literature, XL, 351–401.
Frijda, N. H. (1986). The emotions. Cambridge University Press.
Fudenberg, D. and Levine, D. K. (2006). A dual-self model of impulse control. American Economic Review , 96(5), 1449–1476.
166
BIBLIOGRAPHY
Gächter, S. and Renner, E. (2006). The effects of (incentivized) belief elicitation in
public good experiments. CeDEx Discussion Paper , No. 2006-16.
Gächter, S., Fehr, E., and Kment, C. (1996). Does social exchange increase voluntary
cooperation? Kyklos, 49(4), 541 – 554.
Geanakoplos, J., Pearce, D., and Stacchetti, E. (1989). Psychological games and
sequential rationality. Games and Economic Behavior , 1, 60–79.
Gilbert, D., Pelham, B. W., and Krull, D. S. (1988). On cognitive business: When
person perceivers meet persons perceived. Journal of Personality and Social Psychology, 54, 733 – 740.
Glimcher, P. W., Dorris, M. C., and Bayer, H. M. (2005). Physiological utility theory
and the neuroeconomics of choice. Games and Economic Behavior , 52, 213–256.
Goette, L. and Huffman, D. (2007). Affect and cognition as a source of motivation: A
new model and evidence from natural experiments. In K. D. Vohs, R. F. Baumeister, and G. Loewenstein, editors, Do emotions help or hurt decision making? A
Hedgefoxian Perspective, pages 267–294. New York: Russell Sage Foundation Press.
Gonzalez, R. M. and Loewenstein, G. (2004). Effects of circadian rhythm on cooperation in an experimental game. SSRN Working Paper Series. Available at SSRN:
http://ssrn.com/abstract=485442.
Greiner, B. (2004). An online recruitment system for economic experiments. In
K. Kremer and V. Macho, editors, Forschung und wissenschaftliches Rechnen,
pages 79–93. GWDG Bericht 63. Ges. fr Wiss. Datenverarbeitung, Gttingen.
Griggs, R. and Cox, J. (1982). The elusive thematic-materials effect in wasons selection task. British Journal of Psychology, 73(3), 407 – 420.
Güth, W. (1995). On ultimatum bargaining experiments – A personal review. Journal
of Economic Behavior and Organization, 27, 329 – 344.
BIBLIOGRAPHY
167
Güth, W. (2008). (Non-)behavioral economics - A programmatic assessment. Journal
of Psychology, 216(4), 245 – 254.
Güth, W. and van Damme, E. (1998). Information, strategic behavior, and fairness in ultimatum bargaining: An experimental study. Journal of Mathematical
Psychology, 42, 227–247.
Güth, W., Schmittberger, R., and Schwarze, B. (1982). An experimental analysis of
ultimatum bargaining. Journal of Economic Behavior & Organization, 3, 367–388.
Güth, W., Levati, M. V., Sutter, M., and van der Heijden, E. (2007). Leading by
example with and without exclusion power in voluntary contribution experiments.
Journal of Public Economics, 91, 10231042.
Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach to moral judgment. Psychological Review , 108, 814 – 834.
Hanoch, Y. (2002). neither an angel nor an ant: Emotion as an aid to bounded
rationality. Journal ofEconomic Psychology, 23, 1–25.
Harbaugh, W. T., Krause, K., and Liday, S. J. (2003).
children.
Bargaining by
University of Oregon Economics Working Paper , No. 2002-
4. Available at http://economics.uoregon.edu/papers/UO-2002-4_Harbaugh_
Kid_Bargaining.pdf (Accessed January 28, 2009).
Hertwig, R. and Ortmann, A. (2001).
Experimental practices in economics: A
methodological challenge for psychologists? Behavioral and Brain Sciences, 24,
383–451.
Hinson, J. M., Jameson, T. L., and Whitney, P. (2003). Impulsive decision making
and working memory. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 29(2), 298–306.
168
BIBLIOGRAPHY
Hirshleifer, D. and Shumway, T. (2003). Good day sunshine: Stock returns and the
weather. Journal of Finance, 58(3), 1009–1032.
Hoffman, E., McCabe, K., Shachat, K., and Smith, V. (1994). Preferences, property
rights, and anonymity in bargaining games. Games and Economic Behavior , 7,
346–380.
Hoffman, E., McCabe, K. A., and Smith, V. L. (1996). On expectations and the
monetary stakes in ultimatum games. International Journal of Game Theory, 25,
289 – 301.
Isaac, R. and Walker, J. (1988a). Communication and free-riding behavior: the
voluntary contributions mechanism. Economic Inquiry, 26, 585 – 608.
Isaac, R. M. and Walker, J. M. (1988b). Group size effects in public goods provision:
The voluntary contributions mechanism. Quarterly Journal of Economics, 103(1),
179 – 199.
Isaac, R. M., McCue, K. F., and Plott, C. R. (1985). Public goods provision in an
experimental environment. Journal of Public Economics, 26, 51 – 74.
Isaac, R. M., Schmidtz, D., and Walker, J. M. (1989). The assurance problem in a
laboratory market. Public Choice, 62, 217 – 236.
Isen, A. M. and Means, B. (1983). The influence of positive affect on decision-making
strategy. Social Cognition, 2, 18–31.
Isen, A. M. and Patrick, R. (1983). The effect of positive feelings on risk-taking:
when the chips are down. Organizational Behavior and Human Performance, 31,
194–202.
Isen, A. M. and Shalker, T. E. (1982). Do you ”accentuate the positive, eliminate the
negative” when you are in a good mood? Social Psychology Quarterly, 45, 58–63.
BIBLIOGRAPHY
169
Isen, A. M., Shalker, T. E., Clark, M. S., and Karp, L. (1978). Positive affect,
accessibility of material in memory, and behaviour: A cognitive loop? Journal of
Personality and Social Psychology, 36, 1–12.
Iyengar, S. S. and Lepper, M. R. (2000). When choice is demotivating: Can one
desire too much of a good thing? Journal of Personality and Social Psychology,
79(6), 995 – 1006.
Iyengar, S. S., Huberman, G., and Jiang, W. (2004). How much choice is too much?
contributions to 401(k) retirement plans. In O. S. Mitchell and S. P. Utkus, editors,
Pension Design and Structure: New Lessons from Behavioral Finance, pages 83 –
95. Oxford University Press.
Johnson, E. J. and Goldstein, D. (2003). Do defaults save lives?
Science, 302,
1338–1339.
Johnson, E. J. and Goldstein, D. G. (2004). Defaults and donation decisions. Transplantation, 78(12), 1713 – 1716.
Johnson, E. J., Hershey, J., Meszaros, J., and Kunreuther, H. (1993). Framing,
probability distortions, and insurance decisions. Journal of Risk and Uncertainty,
7, 35 – 51.
Johnson, E. J., Bellman, S., and Lohse, G. L. (2002). Defaults, framing and privacy:
Why opting in-opting out. Marketing Letters, 13(1), 515.
Jones, G. (2008). Are smarter groups more cooperative? evidence from prisoner’s
dilemma experiments, 1959-2003. Journal of Economic Behavior and Organization,
68(3-4), 489–497.
Kagel, J., Kim, C., and Moser, D. (1996). Fairness in ultimatum games with asymmetric information and asymmetric payoffs. Games & Economic Behavior , 13,
100–110.
170
BIBLIOGRAPHY
Kahneman, D. (2003a). Maps of bounded rationality: Psychology for behavioral
economics. American Economic Review , 93(5), 1449 – 1475.
Kahneman, D. (2003b). A perspective on judgment and choice: Mapping bounded
rationality. American Psychologist, 58(9), 697 – 720.
Kahneman, D. and Frederick, S. (2002). Representativeness revisited: Attribute
substitution in intuitive judgment. In T. Gilovich, D. Griffin, and D. Kahneman,
editors, Heuristics & Biases: The Psychology of Intuitive Judgment, chapter 2,
pages 49–81. New York: Cambridge University Press.
Kahneman, D. and Tversky, A. (1979). Prospect theory: An analysis of decision
under risk. Econometrica, 47(2), 263 – 292.
Kahneman, D., Knetsch, J. L., and Thaler, R. H. (1986). Fairness and the assumptions
of economics. The Journal of Business, 59(4), 285–300.
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.
Keser, C. and van Winden, F. (2000). Conditional cooperation and voluntary contributions to public goods. Scandinavian Journal of Economics, 102(1), 23 – 39.
Ketelaar, T. and Todd, P. M. (2000). Framing our thoughts: ecological rationality as
evolutionary psychologys answer to the frame problem. In S. Davies and H. R. Holcomb, editors, The evolution of minds: psychological and philosophical perspective.
Kluvier, Dordrecht.
Kirchsteiger, G. (1994). The role of envy in ultimatum games. Journal of Economic
Behavior and Organization, 25, 373 – 389.
Kirchsteiger, G., Rigotti, L., and Rustichini, A. (2006). Your morals might be your
moods. Journal of Economic Behavior & Organization, 59, 155–172.
BIBLIOGRAPHY
171
Knez, M. J. and Camerer, C. F. (1995). Outside options and social comparison in
three-player ultimatum game experiments. Games and Economic Behavior , 10,
65–94.
Knoch, D., Pascual-Leone, A., Meyer, K., Treyer, V., and Fehr, E. (2006). Diminishing reciprocal fairness by disrupting the right prefrontal cortex. Science, 314,
829–832.
Knutson, B., Taylor, J., Kaufman, M., Peterson, R., and Glover, G. (2005). Distributed neural representation of expected value. The Journal of Neuroscience,
25(19), 4806 – 4812.
Koenigs, M. and Tranel, D. (2007). Irrational economic decision-making after ventromedial prefrontal damage: Evidence from the ultimatum game. The Journal of
Neuroscience, 27(4), 951–956.
Kogler, C. and Kühberger, A. (2007). Dual process theories: A key for understanding
the diversification bias? Journal of Risk and Uncertainty, 34, 145 – 154.
Kuhlman, D. M. and Wimberley, D. L. (1976). Expectations of choice behavior held
by cooperators, competitors, and individualists across four classes of experimental
games. Journal of Personality and Social Psychology, 34, 69–81.
Laibson, D. (1997). Golden eggs and hyperbolic discounting. The Quarterly Journal
of Economics, 112, 443–477.
Larrick, R. P. and Blount, S. (1997). The claiming effect: Why players are more
generous in social dilemmas than in ultimatum games. Journal of Personality and
Social Psychology, 72, 810–825.
Larrick, R. P., Burson, K. A., and Soll, J. B. (2007). Social comparison and confidence:
When thinking youre better than average predicts overconfidence (and when it does
not). Organizational Behavior and Human Decision Processes, 102, 76–94.
172
BIBLIOGRAPHY
Ledyard, J. (1995). Public goods: A survey of experimental research. In J. Kagel
and A. Roth, editors, The Handbook of Experimental Economics, pages 111 – 194.
Princeton: Princeton University Press.
Lee, L., Amir, O., and Ariely, D. (2006). In search of homo economicus: Preference
consistency, emotions, and cognition. SSRN Working Paper Series. Available at:
http://ssrn.com/paper=925978 (accessed June 5, 2007).
Lerner, J. S., Small, D. A., and Loewenstein, G. (2004). Heart strings and purse
strings. carryover effects of emotions on economic decisions. Psychological Science,
15(5), 337–341.
Lieberman, M. D. (2003). Reflective and reflexive judgment processes: a social cognitive neuroscience approach. In J. P. Forgas, K. D. Williams, and W. von Hippel,
editors, Social Judgments: Explicit and Implicit Processes, chapter 3, pages 44–67.
New York: Cambridge University Press.
List, J. A. and Cherry, T. L. (2000). Learning to accept in ultimatum games: Evidence
from an experimental design that generates low offers. Experimental Economics,
3, 11–29.
Lobel, J. and Loewenstein, G. (2005). Emote control: the substitution of symbol
for substance in foreign policy and international law. Chicago Kent Law Review ,
80(3), 1045–1090.
Loewenstein, G. (1996). Out of control: Visceral influences on behavior. Organizational Behavior and Human Decision Processes, 65(3), 272–292.
Loewenstein, G. and O’Donoghue, T. (2005). Animal spirits: Affective and deliberative processes in economic behavior. Working Paper . Carnegie Mellon Univeristy.
Available at: http://www.people.cornell.edu/pages/edo1/will.pdf (accessed
June 25, 2007).
BIBLIOGRAPHY
173
Loewenstein, G. and Schkade, D. (1999). Wouldnt it be nice? predicting future
feelings. In D. Kahneman, E. Diener, and N. Schwarz, editors, Well-Being: The
Foundations of Hedonic Psychology, pages 85 – 105. New York: Russell Sage Foundation Press.
Loewenstein, G. F. and Lerner, J. S. (2003). The role of affect in decision making. In
R. J. Davidson, K. R. Scherer, and H. H. Goldsmith, editors, Handbook of Affective
Sciences, chapter 31, pages 619–642. Oxford: Oxford University Press.
Loomes, G. and Sugden, R. (1982). Regret theory: an alternative theory of rational
choice under uncertainty. Economic Journal , 92, 805 – 824.
Loomes, G. and Sugden, R. (1986). Disappointment and dynamic inconsistency in
choice under uncertainty. Review of Economic Studies, 53, 271 – 282.
Madrian, B. C. and Shea, D. F. (2001). The power of suggestion: Inertia in 401(k)
participation and savings behavior. Quarterly Journal of Economics, 116(4), 1149
– 1187.
Mann, T. and Ward, A. (2007). Attention, self-control, and health behaviors. Current
Directions in Psychological Science, 16(5), 280 – 283.
Mano, H. (1994). Risk-taking, framing effects, and affect. Organizational Behavior
and Human Decision Processes, 57, 38 – 58.
Marks, M. and Croson, R. (1998). Alternative rebate rules in the provision of a threshold public good: An experimental investigation. Journal of Public Economics, 67,
195 – 220.
Marks, M. B., Schansberg, D. E., and Croson, R. T. A. (1999). Using suggested
contributions in fundraising for public good: an experimental investigation of the
provision point mechanism. Nonprofit Management & Leadership, 9(4), 369 – 384.
174
BIBLIOGRAPHY
Marwell, G. and Ames, R. (1979). Experiments on the provision of public goods i:
Resources, interest, group size and the free-rider problem. American Journal of
Sociology, 84, 1335 – 1360.
Maule, A. J., Hockey, G. R. J., and Bdzola, L. (2000). Effects of time-pressure
on decision-making under uncertainty: changes in affective state and information
processing strategy. Acta Psychologica, 104(3), 283–301.
McClure, S. M., Laibson, D. I., Loewenstein, G., and Cohen, J. D. (2004). Separate
neural systems value immediate and delayed monetary rewards. Science, 306,
503–507.
Metcalfe, J. and Mischel, W. (1999). A hot/cool-system analysis of delay of gratification: Dynamics of willpower. Psychological Review , 106(1), 3–19.
Mischel, W., Shoda, Y., and Rodriguez, M. L. (1989). Delay of gratification in
children. Science, 244(4907), 933 – 938.
Mitchell, O. S. and Utkus, S. P. (2006). How behavioral finance can inform retirement
plan design. Journal of Applied Corporate Finance, 18(1), 82 – 94.
Moore, D. A. and Loewenstein, G. (2004). Self-interest, automaticity, and the psychology of conflict of interest. Social Justice Research, 17(2), 189–202.
Mullainathan, S. and Thaler, R. H. (2000).
Behavioral economics.
ternational Encyclopedia of the Social and Behavioral Sciences.
InAvail-
able at http://papers.ssrn.com/sol3/papers.cfm?cfid=371831\&cftoken=
32179229\&abstract_id=245828 (Accessed January 23, 2009).
Munier, B. and Zaharia, C. (2003). High stakes and acceptance behavior in ultimatum
bargaining: A contribution from an international experiment. Theory and Decision,
53, 187207.
BIBLIOGRAPHY
175
Murnighan, J. and Saxon, M. S. (1998). Ultimatum bargaining by children and adults.
Journal of Economic Psychology, 19, 415–445.
Nikiforakis, N. (2008). Punishment and counter-punishment in public good games:
Can we really govern ourselves? Journal of Public Economics, 92, 91 – 112.
Noussair, C. and Soo, C. (2008). Voluntary contributions to a dynamic public good:
Experimental evidence. Economics Letters, 98, 71 – 77.
Ochsner, K. N. and Gross, J. J. (2005). The cognitive control of emotion. TRENDS
in Cognitive Sciences, 9(5), 242–249.
O’Donoghue, T. and Rabin, M. (2000). The economics of immediate gratification.
Journal of Behavioral Decision Making, 13, 233–250.
Oechssler, J., Roider, A., and Schmitz, P. W. (2008). Cognitive abilities and behavioral biases. Available at http://ftp.iza.org/dp3481.pdf (Accessed September
5, 2008).
Oosterbeek, H., Sloof, R., and van de Kuilen, G. (2004). Cultural differences in
ultimatum game experiments: Evidence from a meta-analysis. Experimental Economics, 7, 171–188.
Palfrey, T. R. and Prisbrey, J. E. (1996). Altruism, reputation and noise in linear
public goods experiments. Journal of Public Economics, 61, 4094 – 27.
Palfrey, T. R. and Rosenthal, H. (1991). Testing for effects of cheap talk in a public
goods game with private information. Games and Economic Behavior , 3(2), 183 –
220.
Park, C. W., Jun, S. Y., and MacInnis, D. J. (2000). Choosing what i want versus
rejecting what i do not want: An application of decision framing to product option
choice decisions. Journal of Marketing Research, 37(2), 187 – 202.
176
BIBLIOGRAPHY
Pillutla, M. M. and Murnighan, J. K. (1996). Unfairness, anger, and spite: Emotional
rejections of ultimatum offers. Organizational Behavior and Human Decision Processes, 68(3), 208–224.
R Development Core Team (2005). R: A language and environment for statistical
computin. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3900051-07-0.
Rabin, M. (1993). Incorporting fairness into game theory and economics. American
Economic Review , 83(6), 1281 – 1302.
Rankin, F. W. (2003). Communication in ultimatum games. Economics Letters, 81,
267–271.
Rapoport, A. and Suleiman, R. (1993). Incremental contribution in step-level public goods games with asymmetric players. Organizational Behavior and Human
Decision Processes, 55(2), 171 – 194.
Rege, M. and Telle, K. (2004). The impact of social approval and framing on cooperation in public good situations. Journal of Public Economics, 88, 1625 – 1644.
Rick, S. and Loewenstein, G. F. (2007). The role of emotion in economic behavior.
SSRN Working Paper Series. Available at: http://ssrn.com/abstract=954862
(accessed June 6, 2007).
Ritov, I. and Baron, J. (1992). Status-quo and omission biases. Journal of Risk and
Uncertainty, 5, 49 61.
Roch, S. G., Lane, J. A. S., Samuelson, C. D., Allison, S. T., and Dent, J. L. (2000).
Cognitive load and the equality heuristic: A two-stage model of resource overconsumption in small groups. Organizational Behavior and Human Decision Processes,
83(2), 185–212.
BIBLIOGRAPHY
177
Rondeau, D., Poe, G. L., and Schulze, W. D. (2005). Vcm or ppm? a comparison
of the performance of two voluntary public goods mechanisms. Journal of Public
Economics, 89, 1581 – 1592.
Rosenthal, R. W. (1981). Games of perfect information, predatory pricing and the
chain-store paradox. Journal of Economic Theory, 25(1), 92 – 100.
Rozin, P., Haidt, J., and McCauley, C. (2000). Disgust. In M. L. . J. Haviland, editor,
Handbook of emotions, chapter 40, pages 637 – 653. New York: Guilford Press.
Rubinstein, A. (2002). Irrational diversification in multiple decision problems. European Economic Review , 46, 1369 – 1378.
Rubinstein, A. (2007). Instinctive and cognitive reasoning: a study of response times.
Economic Journal , 117, 1243–1259.
Rustichini, A. (2008). Dual or unitary system? two alternative models of decision
making. Cognitive, Affective, & Behavioral Neuroscience, 8(4), 355–362.
Samuelson, W. and Zeckhauser, R. (1988). Status quo bias in decision making. Journal of Risk and Uncertainty, 1, 7 59.
Sanfey, A. G., Rilling, J. K., Aronson, J. A., Nystrom, L. E., and Cohen, J. D. (2003).
The neural basis of economic decision-making in the ultimatum game. Science, 300,
1755–1758.
Sanfey, A. G., Loewenstein, G., McClure, S. M., and Cohen, J. D. (2006). Neuroeconomics: cross-currents in research on decision-making. TRENDS in Cognitive
Sciences, 10(3), 108–116.
Schelling, T. (1960). The strategy of conflict. Harvard University Press.
178
BIBLIOGRAPHY
Schmeichel, B. J., Vohs, K. D., and Baumeister, R. F. (2003). Intellectual performance and ego depletion: Role of the self in logical reasoning and other information
processing. Journal of Personality and Social Psychology, 85(1), 33–46.
Schmitt, P. M. (2004). On perceptions of fairness: The role of valuations, outside
options, and information in ultimatum bargaining games. Experimental Economics,
7, 49–73.
Schotter, A. and Sopher, B. (2007). Advice and behavior in intergenerational ultimatum games: An experimental approach. Games and Economic Behavior , 58,
365–393.
Schwarz, N. and Clore, G. L. (1983). Mood, misattribution, and judgments of wellbeing: Informative and directive functions of affective states. Journal of Personality
and Social Psychology, 45, 513 – 523.
Schwarz, N. and Clore, G. L. (2003). Mood as information: 20 years later. Psychological Inquiry, 14(3&4), 296 – 303.
Sefton, M., Shupp, R., and Walker, J. M. (2007). The effect of rewards and sanctions
in provision of public goods. Economic Inquiry, 45(4), 671 – 690.
Sell, J. and Wilson, R. (1991). Levels of information and contributions to public
goods. Social Forces, 70, 107 – 124.
Sent, E.-M. (2004). Behavioral economics: How psychology made its (limited) way
back into economics. History of Political Economy, 36(4), 735 – 760.
Shafir, E., Simonson, I., and Tversky, A. (1993). Reason-based choice. Cognition,
49(1 - 2), 11 – 36.
Shiv, B. and Fedorikhin, A. (1999). Heart and mind in conflict: The interplay of
affect and cognition in consumer decision making. Journal of Consumer Research,
26, 278–292.
BIBLIOGRAPHY
179
Shupp, R., Schmitt, P., and Swope, K. (2006). On the role of the hostage in ultimatum
bargaining games. The Journal of Socio-Economics, 35, 399–411.
Simon, H. A. (1955). A behavioral model of rational choice. The Quarterly Journal
of Economics, 69(1), 99–118.
Simon, H. A. (2008). Bounded rationality. In S. N. Durlauf and L. E. Blume, editors,
The New Palgrave Dictionary of Economics. Palgrave Macmillan, second edition.
Simonsohn, U. (2007). Clouds make nerds look good: Field evidence of the impact
of incidental factors on decision making. Journal of Behavioral Decision Making,
20, 143–152.
Skitka, L. J., Mullen, E., Griffin, T., Hutchinson, S., and Chamberlin, B. (2002).
Dispositions, scripts, or motivated correction? understanding ideological differences
in explanations for social problems. Journal of Personality and Social Psychology,
83, 470 – 487.
Sloman, S. A. (1996). The empirical case for two systems of reasoning. Psychological
Bulletin, 119(1), 3–22.
Slonim, R. and Roth, A. E. (1998). Learning in high stakes ultimatum games: An
experiment in the slovak republic. Econometrica, 66(3), 569–596.
Smith, A. (1759/1982). The Theory of Moral Sentiments. Liberty Fund, Indianapolis.
Solnick, S. J. (2001). Gender differences in the ultimatum game. Economic Inquiry,
39(2), 189–200.
Solnick, S. J. and Schweitzer, M. E. (1999). The influence of physical attractiveness
and gender on ultimatum game decisions. Organizational Behavior and Human
Decision Processes, 79(3), 199–215.
180
BIBLIOGRAPHY
Stanley, T. D. and Tran, U. (1998). Economics students need not be greedy: Fairness
and the ultimatum game. Journal of Socio-Economics, 27(6), 657–664.
Stanovich, K. E. (1999). Who is rational? Studies of individual differences in reasoning. Hillsdale, NJ: Erlbaum.
Stanovich, K. E. and West, R. F. (1998). Cognitive ability and variation in selection
task performance. Thinking and Reasoning, 4(3), 193 – 230.
Starmer, C. (2000). Developments in Non-Expected Utility Theory: The Hunt for a
Descriptive Theory of Choice under Risk. Journal of Economic Literature, 38(2),
332–382.
Strack, F. and Deutsch, R. (2004). Reflective and impulsive determinants of social
behavior. Personality and Social Psychology Review , 8(3), 220 – 247.
Strack, F., Werth, L., and Deutsch, R. (2006). Reflective and impulsive determinants
of consumer behavior. Journal of Consumer Psychology, 16(3), 205 – 216.
Straub, P. G. and Murnighan, J. K. (1995). An experimental investigation of ultimatum games: information, fairness, expectations, and lowest acceptable offers.
Journal of Economic Behavior and Organization, 27, 345 –364.
Suleiman, R. and Rapoport, A. (1992). Provision of step-level public goods with
continuous contribution. Journal of Behavioral Decision Making, 5, 133 – 153.
Sunstein, C. R. and Thaler, R. H. (2003). Libertarian paternalism is not an oxymoron.
The University of Chicago Law Review , 70(4), 1159 – 1202.
Sutter, M., Kocher, M., and Strau, S. (2003). Bargaining under time pressure in an
experimental ultimatum game. Economics Letters, 81, 341 – 347.
Thaler, R. (1980). Toward a positive theory of consumer choice. Journal of Economic
Behavior and Organization, 1, 39 – 60.
BIBLIOGRAPHY
181
Thaler, R. H. and Shefrin, H. M. (1981). An economic theory of self-control. Journal
of Political Economy, 89(2), 392–406.
Thaler, R. H. and Sunstein, C. R. (2003). Libertarian paternalism. The American
Economic Review , 93(2), 175 – 179.
Tversky, A. and Kahneman, D. (1973). Availability: a heuristic for judging frequency
and probability. Cognitive Psychology, 5, 207–232.
Tversky, A. and Kahneman, D. (1974). Judgment under uncertainty: Heuristics and
biases. Science, 185, 1124–1131.
Tversky, A. and Kahneman, D. (1983). Extensional vs. intuitive reasoning: The
conjunction fallacy in probability judgment. Psychological Review , 90(4), 293 –
315.
Tversky, A. and Shafir, E. (1992). Choice under conflict: The dynamics of deferred
decision. Psychological Science, 3(6), 358 – 361.
van den Bos, K., Peters, S. L., Bobocel, D. R., and Ybema, J. F. (2006). On preferences and doing the right thing: Satisfaction with advantageous inequity when
cognitive processing is limited. Journal of Experimental Social Psychology, 42,
273289.
van´ t Wout, M., Kahn, R. S., Sanfey, A. G., and Aleman, A. (2006). Affective state
and decision-making in the ultimatum game. Experimental Brain Research, 169,
564–568.
van Winden, F. (2007). Affect and fairness in economics. Social Justice Research,
20(1), 35–52.
Vohs, K. D., Baumeister, R. F., Schmeichel, B. J., Twenge, J. M., Nelson, N. M.,
182
BIBLIOGRAPHY
and Tice, D. (2008). Making choices impairs subsequent self-control: A limitedresource account of decision making, self-regulation, and active initiative. Journal
of Personality and Social Psychology, 94(5), 883 – 898.
Walker, J. M. and Halloran, M. A. (2004). Rewards and sanctions and the provision
of public goods in one-shot settings. Experimental Economics, 7, 235 – 247.
Ward, A. and Mann, T. (2000). Don’t mind if i do: disinhibited eating under cognitive
load. Journal of Personality and Social Psychology, 78(4), 753–763.
Wason, P. C. (1968). Reasoning about a rule. Quarterly Journal of Experimental
Psychology, 20(3), 273 281.
Weimann, J. (1994). Individual behaviour in a free riding experiment. Journal of
Public Economics, 54, 185 – 200.
Westermann, R., Spies, K., Stahl, G., and Hesse, F. W. (1996). Relative effectiveness
and validity of mood induction procedures: a meta-analysis. European Journal of
Social Psychology, 26, 557–580.
Wilcox, N. T. and Feltovich, N. (2000). Thinking like a game theorist: Comment.
mimeo, University of Houston.
Wright, W. F. and Bower, G. H. (1992). Mood effects on subjective probability
assessment. Organizational Behavior and Human Decision Processes, 52, 276–291.
Xiao, E. and Houser, D. (2005). Emotion expression in human punishment behavior.
PNAS , 102(20), 7398–7401.
Zakay, D. (1993). The impact of time perception processes on decision making under
time stress. In O. Svenson and A. J. Maule, editors, Time Pressure and Stress
in Human Judgment and Decision Making, chapter 4, pages 59–72. Plenum Press,
New York.
BIBLIOGRAPHY
183
Zelmer, J. (2003). Linear public goods experiments: A meta-analysis. Experimental
Economics, 6, 299 310.