Chapter 2

Motivations and incentives

In mainstream economics, what is often loosely described as neoclassical economics,

­economic agents are assumed to be driven by purely economic and monetary motivations.
A large section of behavioural economics takes on the assumptions associated with this
mainstream model to explore other drivers of our everyday choices and decisions. In this
chapter, we will explore some of these non-monetary motivations, and the models and em-
pirical techniques that behavioural economists use to capture them. In understanding what
drives the wider range of incentives and motivations that behavioural economists explore,
it is important to understand well some of the assumptions and microeconomic principles
from neoclassical models because these models focus strongly on the idea that rational
“agents” respond perfectly to incentives. We shall begin this chapter exploring the approach
to analysing incentives that is conventionally adopted by non-behavioural economists.

Incentives in neoclassical economics

In explaining how incentives drive agents’ choices, neoclassical economists make assump-
tions to construct an artificial model of human decision-making. It is useful in under-
standing this artificial model to imagine that economies operate as if they are populated
by a specific type of artificially conceived species: Homo economicus. Homo economicus is as-
sumed to have an exceptional capacity for information processing and decision-making.
In this model, people are assumed to be:

a. Well-informed and able to use information efficiently.

b. Independent in two senses:
i. atomistic – they do not look to others when deciding what to do;
ii. selfish – their utility is determined only by their own comforts.
c. Rational maximizers:
i. using information efficiently by applying mathematical tools to guide their be-
haviour, so their behaviour is systematic and objectively determined, and im-
mune from emotional and sentimental forces, especially if these are associated
with systematic biases in decision-making. As rational maximizers, generally
Homo economicus is assumed to maximize utility and profits in monetary terms;
20 Motivations and incentives

ii. forward-looking in a systematic way, which involves discounting the future

in ways which are consistent over time so that it makes no difference whether
Homo economicus is looking one day ahead or one decade ahead: their choices about
future plans are consistent.
d. Homogenous, i.e. all members of the Homo economicus species behave in the same way
(on average at least) so that one model can capture everyone (on average at least).

These standard assumptions imply that, at the extreme, standard economic models de-
scribe people as robotic, mathematical machines, and so understanding and controlling
behaviour can be seen as, in many ways, more similar to an engineering problem than a
socio-psychological problem. Behavioural economics takes on this approach to develop a
more realistic view of how real people make their economic decisions and choices, start-
ing by exploring the non-monetary incentives and motivations that drive our behaviour.

Extrinsic versus intrinsic motivations

In challenging the standard neoclassical approach to incentives, behavioural economists
start by exploring the wider range of incentives and motivations that drive people’s choices
and decisions – and a prominent theme in this behavioural literature comes in capturing
interactions between individual and cooperative goals. Constructing markets around peo-
ple’s willingess to pay for things that have been traditionally been sustained on the basis
of social values can threaten the existence of socially beneficial actions. Titmuss (1970)
described the negative impact of introducing payments for blood donation: payments
undermined social values and dampened people’s willingness to donate. Frey and Jegen
(2001) examine this phenomenon in the context of motivation crowding theory and draw
on insights from psychology about the “hidden costs of rewards”: monetary incentives
drive extrinsic motivation (motivation driven by external rewards) and undermine in-
trinsic motivations, including internally-driven motivations such as curiosity, helpfulness
and self-realization. Crowding-out of intrinsic effects undermines the focus in standard
economic theory on the importance of monetary rewards and incentives.

Motivating environmental awareness

Many of these insights have been applied in the context of environment decision-making.
In an ideal world, people should be motivated by their sense of social responsibility to care
about the environment – for example, Gowdy (2008) applies insights from behavioural
economics and experimental psychology to the issue of climate change and specifically to
the question of reducing CO2 emissions. He argues that resolving the current crisis of sus-
tainability needs more emphasis on broader facets of human behaviour, including trade-
offs between greed and egoism versus cooperation and altruism. Rational choice theory
does not capture these trade-offs effectively and so will not be a good guide, for example
for policy-makers trying to encourage people to consider more carefully their energy and
environment decisions. Financial incentives may in fact crowd out extrinsic motivations
and feelings of collective responsibility unless environmental policy draws on cooperative,
non-materialistic aspects of human nature. Frey (1997) argues that monetary incentives can
crowd out civic motives but money can also “crowd in” civic motivations when it is used
to acknowledge social worth of individual contributions (Frey and Oberholzer-Gee 1997).
 Extrinsic versus intrinsic motivations 21
Concerns about fairness will also affect the global management of environmental
decisions, especially with respect to the developing world. Dealing with climate change
will need cooperation, trust and reciprocity and even when cooperative frameworks are
imperfect, participation can establish credibility and goodwill (Gowdy 2008). Given that
richer countries in what is known as the global North got rich by burning fossil fuels, is
it fair to tell the developing world to stop using them? Stiglitz (2006) argues that a fair
solution could be to implement a common global carbon tax and allow each country to
keep the carbon tax revenues so that if less developed countries are polluting more then
they will also have more taxation revenue to spend on either reducing the taxation burden
in other areas, or increasing expenditure, including expenditures to support development
of new, environmentally friendly technologies.
Bénabou and Tirole (2003) adapt these insights and incorporate them within an eco-
nomic analysis of principal–agent problems. External rewards offered by a principal (e.g.
an employer) affect the intrinsic motivations of an agent (e.g. an employee) so that exter-
nal incentives are weakly reinforcing in the short-run and negatively reinforcing in the
long-run. There are many examples from experimental economics. In one study, univer-
sity students were asked to solve a puzzle. Those who were not paid, and so were presum-
ably doing the puzzle for the intellectual challenge, put more effort into the task than the
students who were paid (Deci 1975).
Gneezy and Rustichini also report results from a field study of parents collecting
children from school. Parents would often arrive late forcing the school to make sure
a teacher was available to look after the children until their parents arrived. The school
decided to introduce a fine as a deterrent but the plan backfired and more parents arrived
late with a late fine than without one (Gneezy and Rustichini 2000a). This phenomenon
may reflect the fact that, once a fine was introduced, arriving late was then interpreted
as costly service and any guilt that parents may formerly have felt about arriving late was
reduced when they felt they were choosing to pay for the privilege of arriving late. The
extrinsic disincentive (the fine) crowded-out the intrinsic motivation – to cooperate by
trying to arrive on time as often as possible.
Gneezy and Rustichini (2000b) found similar results to these earlier studies in an
analysis of the performance of experimental subjects (students from the University of
Haifa, Israel) completing a series of quizzes and offered a range of different incentives –
with payments made as New Israeli Shekels (NISs). One group was given no payment;
the second, third and fourth groups were paid 10 cents, 1 NIS and 3 NIS respectively and
performance was lowest in the second group. Those who were not paid at all performed
better than those with a small 10 cents payment. Gneezy and Rustichini found similar
results for a study of volunteer work by high school children and conclude that excessively
small payments can be demotivating leading to worse performance than no payment at all
because small payments crowd out intrinsic motivation without offering sufficient exter-
nal incentives to leverage extrinsic motivations.
Bénabou and Tirole (2006) use concepts of intrinsic and extrinsic motivation to ex-
plain pro-social behaviour. In many contexts, people exhibit pro-social behaviour in-
volving altruism or reciprocity. This reflects not only intrinsic motivations but also social
pressures and social norms which link into reputations and self-respect. Taking into ac-
count this mix of motivations, as well as heterogeneity in propensities towards altruism
and self-interest, Bénabou and Tirole construct a model in which choices are the outcome
of three motivators: intrinsic motivation, extrinsic motivation and reputation building.
22 Motivations and incentives

Extrinsic rewards not only crowd-out intrinsic motivations but they also “spoil” the rep-
utational and/or self-image value of pro-social choices.
Ariely et al. (2009a) explore intrinsic motivation, specifically personal preferences for
giving, extrinsic motivation and image motivation. They formulate an “effectiveness hypoth-
esis”: extrinsic rewards are less effective for visible pro-social actions: extrinsic rewards deter
pro-social behaviour because they dilute its signalling value. They explore these hypotheses
using experimental evidence from subjects randomly assigned to one of two US charities,
either the American Red Cross or the National Rifle Association. Charities were labelled as
“good” or “bad” according to participants’ perceptions of the majority’s view. ­Participants
were asked to engage in a simple but effortful task (pressing X and Z on a keyboard). Perfor-
mance was rewarded with donations by the experimenters to the subject’s nominated char-
ity. There were significant differences in effort under “public” conditions (when subjects’
efforts were revealed) versus private conditions (when efforts weren’t revealed). When no
monetary incentives were offered, subjects in the public condition made more effort. Mone-
tary incentives did not increase effort in public condition; in the private condition there were
significant increases in effort with monetary incentives. Ariely et al. interpret this as support
for their effectiveness hypothesis: monetary incentives work better for anonymous giving
but they have a negative impact on public giving perhaps because the social signalling value
of philanthropy is diluted when efforts and payments are public knowledge.

Social motivations
As we explored in the introduction, behavioural economists learn a number of lessons from
other social sciences – including insights from psychology and sociology around the idea
that, as social creatures, we are not purely self-interested. Our incentives and motivations
are also determined by our relationships with other people around us. Nobel Prize-­winning
economist Amartya Sen (1977, 1991) argues that the focus on a single assumption of self-­
interest in standard models suggests that people are “rational fools”, considering only a
single preference ordering given by self-interest without recognizing social preferences over
a range of alternatives. Conversely, in his comparative analysis of Adam Smith’s (1776) An In-
quiry into the Nature and Causes of the Wealth of Nations and his (1759) Theory of Moral Sentiments, Vernon
L. Smith (1998) hypothesizes that altruism and self-interest, whilst apparently contradictory,
are actually consistent – an idea we introduced in Chapter 1. The self-interest in the Wealth of
Nations and the sympathy in the Theory of Moral Sentiments are different facets of the same thing –
a “self-interested propensity to exchange”. In the Wealth of Nations it is exchange of money
and goods; in the Theory of Moral Sentiments it is exchange of friendship. In this way, Vernon L.
Smith asserts that self-interest and other-regarding sympathy are connected.

Behavioural game theory

Behavioural economists address these debates about the limits to our self-interest by build-
ing models incorporating social preferences, sometimes called other-regarding prefer-
ences. The main theoretical vehicle for analysing the links between our social preferences
and social motivations is behavioural game theory – in which social preferences are intro-
duced into standard games explored by economists – providing a starting point for many
of the experimental tests of people’s social preferences. Some versions of behavioural game
 Social motivations 23
theory were developed in response to robust experimental studies of a wide range of games
showing that people do not always play games selfishly and so behaviour does not converge
in the ways predicted by standard game theory – see Camerer (2003a,b) for surveys. Some
of the games commonly used in these experimental analyses, and the standard game the-
ory solutions consistent with perfectly rational choice, are summarized in Box 2.1.

Box 2.1 Box of games

Ultimatum game
Player A (the proposer) is given a sum to divide between herself and player B (the responder). If
B rejects A’s offer, both players get zero. IF B accepts A’s offer, players get the share proposed
by A. Standard game theory predicts that A will offer the minimum possible amount and B will
accept it because a rational maximiser will prefer anything, no matter how small, to nothing.

Dictator game
Player A, the dictator, is given a sum of money and offers a share to player B; player B cannot
veto the offer and must take what he’s given by A. Standard game theory predicts that Player
A will maximize by offering B the minimum possible amount.

Envy games
These games are variants of dictator or ultimatum games but are designed to disentangle
preferences about relative advantage. Player A has a choice between dividing a small amount
equally between themselves and Player B versus a larger amount but with proportionately
more offered to Player B. For example, Player A is told to choose between:

1. £10 divided equally so that A keeps £5 and gives £5 to Player B; and

2. £15 divided unequally with Player A keeping a lesser share of £6 and giving £9 to
Player B.

Standard game theory predicts that A will prefer option 2 because they will prefer more to
less and will ignore how it is distributed.

Public goods games

In these games, each player makes a contribution to a public good. Their benefit may exceed
their contribution but because, by definition, access to public goods is free and non-rivalrous:
in other words, public goods are freely available and accessible for everyone to use, and one
person using the public good does not prevent another person from using it. Under these
conditions, there is no economic incentive for a perfectly rational, maximising individual to
contribute money to something they can access for free. So standard game theory predicts
scenarios in which players free-ride on others’ contributions.

Trust games
Trust games are two stage games:

In stage 1, Player A – the trustor – offers a sum of money for Player B – the trustee. A’s
contribution is multiplied by some factor by the experimenter and then given to Player B.

In stage 2, B decides how much to return to A.

24 Motivations and incentives

If A is generous/trusting and B reciprocates then there is a Pareto improvement be-

cause both players will be better off. However, standard game theory predicts ‘back-
ward induction’, that is, each player reasons back from the very last stage of a game to
figure out which strategies to use in the earlier stages of the game. Person A figures
that B will not send anything back because he has no monetary incentive to do so and
so A doesn’t send anything in the first place. (Sometimes known as investment games
or gift exchange games.)

Centipede games
Centipede games are multistage versions of trust games; sums of money are sent back and
forth between A (the trustor) and B (the trustee) until one or other player “takes” by deciding
how much to keep for themselves. At that point, the game ends. In the meantime, if the player
decides to continue playing then they “pass” at each stage. Standard game theory predicts
backward induction as for the trust game but in real-life experiments people backward in-
duct only a couple of steps (Camerer 2003b).

Solidarity games
In a solidarity game, three players roll a die to determine a win. Before rolling the die, each
player announces how they will allocate “gifts” to the losers.

This game can be used to separate some of the explanations for generosity in simpler
games such as the ultimatum game; e.g. some studies find that the same total amount
is given regardless of the size of the group and therefore apparent generosity is not
about inequity aversion; gifts for each loser are positively correlated with expectations
of gifts from them (Bolton and Ockenfels 2000).

Ultimatum games and dictator games

In playing experimental games, Box 2.1 summarises some of the solutions that a per-
fectly rational maximising player would implement. A major contribution from behav-
ioural economics comes in showing that real people do not play these games in this sort
of strictly rational way. The ultimatum game is possibly the most famous behavioural
game and it has been extensively explored, not only by economists but also experi-
mental psychologists, neuroscientists and even behavioural ecologists. In the ultimatum
game, the experimenter gives Alice (the proposer) a fixed sum – say £100; Alice is in-
structed to offer Bob (the responder) some proportion of that sum – a minimum of £1
to a maximum of £100. If Bob agrees, then they both get the shares proposed by Alice,
but if Bob refuses, both get nothing. Standard game theory assumes “monotonicity”,
that is, that more will be preferred to less and so Alice will want to keep as much for
herself as possible. She will offer Bob £1 and keep £99 for herself. Bob will prefer £1 to
nothing and so will accept Alice’s offer. Thus, standard game theory predicts offers and
acceptances close to zero.
Experimental evidence shows, however, that people do not play in this way in the
real world: Güth, Schmittberger and Schwarze (1982) conducted the first comprehensive
testing of the original ultimatum game and found that players were often guided by what
they thought was a fair or just outcome. Assuming that 50% was perceived as a “fair”
 Social motivations 25
offer, proposers usually offered responders a lot more than zero and relatively close to
50%; and proposers rejected offers around 20%. These findings were replicated across
many studies (including animal studies, for example with monkeys playing for juice). In
a meta-analysis of results from the ultimatum game, Camerer found that the mean offer
was around 30–40% and offers below about 20% were rejected. The findings were also
scalable, with similar results for small versus large sums of money, though there were
some cultural differences (Camerer 2003b; Henrich et al. 2004). Andersen et al. (2011) an-
alysed strategies played by people in poor villages of Northeast India for whom increases
in monetary stakes had a far larger impact. Andersen et al. found that, even when the stakes
are relatively large, responders will still reject low offers and end-up with zero – inter-
preted by some researchers as reflecting the responders’ desires to punish the proposer for
making an insultingly low offer.

Experiments with trust games

Berg, Dickhaut and McCabe (1995) explore trust, reciprocity and social history and try
to answer questions about why we trust and whether or not trust is a primitive response.
They construct a repeated game but one which abstracts from reputation and contractual
pre-commitments by using a trust game designed to test trust and trustworthiness. This
game tests the extent to which people trust others and in turn reciprocate trust by being
trustworthy themselves. Trusting and being trustworthy represent two challenges to nar-
row self-interest: trusting is a risky strategy and may not be reciprocated; trustworthiness
yields no direct, immediate return.
Berg et al. note that a problem with trust games (and similar games including the
ultimatum and dictator games) is that experimental subjects may want to impress
experimenters with their generosity, so they incorporate some experimental design
features to reduce the likelihood of experimental bias, introducing a double-blind
procedure in which the subjects make contributions anonymously using envelopes
and boxes. The experimenter cannot know (and the subjects know that the experi-
menter cannot know) who has been generous and so there are neither sanctions nor
rewards for altruistic choices. In this way, Berg et al. hope that their experimental
design will allow them to abstract from relationships, social influences, communica-
tion, and so on.
Berg et al. use two groups of subjects in two rooms: subject As in Room A and subject
Bs in Room B. All subjects are given $10 to either keep or share. The game then proceeds
in two stages. In Stage 1, subject A decides how much of her $10 to send to an anonymous
counterpart in room B (the trustee). Stage 1 tests the trust of players A; if they do not trust
the people in Room B then they will send no money. In Stage 2, A’s contribution is tripled
and given to player B. Player B must then decide by how much to reciprocate. Essentially,
stage 2 is a dictator game; Player B has complete control over the outcome and Player A
cannot veto his or her choice.
Standard game theory predicts that, if B is purely self-interested, then he has no in-
centive to reciprocate by being trustworthy and returning something to A. Via backward
induction, player A reasons therefore that their best strategy is to send nothing because
anything that they do send will be kept by B. Assuming that A is purely self-interested,
then she has no incentive to send anything to B in the first place.
26 Motivations and incentives

The problem for both players is that they get stuck in something like a prisoner’s
dilemma: both players could have done better by being generous. With no trust and no
reciprocity A keeps $10, B does nothing and both suffer as a result because neither of them
gets any more than their initial payment. On the other hand, if A had been maximally
generous and offered $10 then this would have been tripled to $30. If B had sent half
back to A, that is, $15, then this would have been a better outcome. Both players would
have been unequivocally better off if they had been trusting and reciprocating. This links
into evolutionary analyses in which making and returning kindnesses have evolutionary
advantages in social environments.
Contributing nothing and not reciprocating is the standard game theory prediction
but the experimental results from the trust games above resemble those seen in ultima-
tum games. In Berg et al.’s study almost all Room A subjects were trusting; 30 out of 32
subjects (94%) sent something to Room B, though there was a large degree of variability
in amounts sent. There is less evidence of reciprocity from the players in room B though
57% did reciprocate by sending back more than A sent and, on average, players in room A
did at least get their money back. Apart from that, the extent of trust and reciprocity was
not correlated. To assess the impact of social norms, Berg et al. also analysed the impact
of social history, that is, information about average responses in previous rounds. They
found that these social norms did have some impact especially for players in room B.
The selection of experimental evidence outlined above suggests that people do not
generally make choices consistent with the standard assumptions of game theory. There
are a number of potential explanations: it could just be that people are inherently gen-
erous. Social norms may dictate that people play in a generous way. As explained in the
context of learning models in the previous chapter, it may take time to learn the best
strategies; if people have learnt social norms of generosity outside the lab then it takes
time and experience to unlearn them. Responses may reflect strategic reasoning, for ex-
ample if people believe that stinginess will at some point be punished by the other player.
Overall, there is an empirical problem in separating these hypotheses about motivation
and a number of theoretical models have been devised to explain the results, as explored
in the following section.
In attempting to reconcile the experimental results and varying interpretations, be-
havioural economists have come up with a range of theoretical explanations for the range
of social preferences and motivations revealed in the experiments outlined above. These
models focus on the different reasons for cooperation and altruism including traits of
kindness and fairness, warm glow giving, inequity aversion, preferences for relative as
well as absolute payoffs and strategic thinking. A selection of the models is outlined below.

Punishment and cooperation

Negative incentives can be motivating too. For example, punishment is a powerful
negative incentive, and behavioural economists have explored punishment as a way
to explain some of the experimental findings from behavioural game theory, as out-
lined above. These models present theoretical explanations for other-regarding behav-
iour but what sort of incentives and motivations encourage people to cooperate and/or
to punish people who don’t? How are these motivations reinforced in a social context?
Cooperation and punishment play a crucial role in explaining why people sometimes
 Social motivations 27
cooperate and other times do not. It is helpful to think about what deters people from
violating social norms of reciprocity and what people can do to encourage others to
cooperate.

Altruistic punishment
Studies in behavioural economics and neuroeconomics have shown that people are willing
to pay to punish norm violators. They also find that people who are initially cooperative, will
start to defect if they are not punished. Fehr and Gächter (2000) study altruistic punishment
of social norm violations. They begin by defining social norms as behavioural regularities
based on socially shared belief about how to behave. Social norms are enforced via informal
social sanctions and these sanctions determine work effort, consumption choices, common
pool resources, as well as provision of public goods. Social history is important and people do
punish inappropriate behaviour even when costly to themselves. Fehr and Gächter (2000) ex-
plore these insights in an experimental study of altruistic punishment based on an adaptation
of Fehr and Schmidt’s (1999) version of a public goods game, which we will explore in more
depth below. Their experimental game was conducted in two stages: a contribution stage and
a punishment stage. In the punishment stage, the players were given the opportunity to use
monetized punishment points to punish defectors who had been mean in the contribution
stage of the game. They find that, without punishment, there was almost complete defection
from social norms but with punishment a larger proportion cooperated.
There are further differences dependent on the level of anonymity in the interactions.
They divided the treatments into Partner and Stranger treatments. When subjects were
playing with strangers they were less likely to cooperate; when they were playing with
known partners and violations were punished, 82.5% of subjects cooperated fully. At the
other extreme, in the no-punishment Stranger treatments, contributions converged onto
full free-riding over time. A similar experimental design has been adapted for neuroeco-
nomic analysis by de Quervain et al. (2004) who find that altruistic punishment stimulates
neural responses usually associated with reward processing, as we will explore in more
detail in the neuroeconomics chapters.

Ostracism in social networks

Punishment can be particularly effective in social networks. Fowler and Christakis (2010)
and Randa et al. (2011) examined punishment within social networks. They conducted
online experiments using Mechanical Turk (MTURK) to investigate large-scale cooper-
ation in social networks. The experimental subjects played public goods games either
with or without punishment and were assigned partners either randomly or via their
own choices. When partners are randomly allocated, cooperation is equally beneficial
to all partners. There is no incentive to cooperate and so cooperation decays over time,
confirming Fehr and Gächter’s (2000) study. Also, social factors may operate as much as
a carrot as a stick: people are inclined to cooperate because they benefit in terms of their
own social connectedness. Questions of depersonalization and identification (as explored
in more detail in the identity section below) may also be relevant: if people feel that they
have built a positive cooperative relationship with someone they know and/or identify
with then that will encourage them to partner with that person again.
28 Motivations and incentives

Fowler and Christakis (2010) found that voluntary costly punishment can help sustain
cooperation. Subjects were influenced by the contributions of others, including those
with whom they did not interact initially. These social influences persist over time and
spread through social networks. Randa et al. (2011) developed this study to show experi-
mentally that people make and break social networks in response to cooperation versus
defection by others. Cooperation decays over time if social links are outside the control of
individuals (because they are fixed or determined randomly) or if links are updated infre-
quently. However, if subjects have short-term control of their social connections they can
decide to break links with defectors and form links with cooperators. When people can
choose who they do (and don’t) interact with, it creates an incentive to cooperate because
defectors will be excluded from social networks if they behave uncooperatively.

Social motivation theory

In explaining some of the findings about social motivations, some behavioural economists
have developed a range of theoretical models to capture social preferences and ­motivations –
via social motivation theory. In social motivation theories, theories of utility are developed essen-
tially to extend neoclassical conceptions of utility. These general utility models are designed
to reconcile, albeit in a limited way in terms of links with other social sciences, the range
of experimental evidence – from the market games evidence in which people exhibit pure
self-interest through to the games which reveal generous behaviour in other contexts, for
example the ultimatum, dictator and trust games explored above.

Bolton and Ockenfels’ Equity, Reciprocity

and Competition (ERC) model
Bolton and Ockenfels’ (1998, 2000) equity, reciprocity and competition (ERC) model is a
general model of motivation, devised to capture other-regarding preferences. They argue
that models with altruistic preferences cannot fully explain play in ultimatum games,
dictator games and solidarity games because needs of others and reducing inequality are
not necessarily primary goals.
Bolton and Ockenfels suggest that a good model should explain a number of statements
summarising some empirical regularities they identified from previous experimental analyses:

Statement 1: In dictator games, on average dictators keep at least half but give less
than the whole pie.
Statement 2: In ultimatum games, responders accept all offers of the whole pie and
reject all offers of nothing
Statement 3: In ultimatum games, proposers will propose a payoff to themselves of
at least half but less than the whole pie, i.e. offering some proportion less than half
to the recipient.
Statement 4: Average offers in ultimatum games exceed average offers in dictator
games.

Developing a comparative model from Bolton (1991) and Bolton and Ockenfels (1998),
the ERC model assumes narrow self-interest but with people balancing trade-offs between
their absolute pecuniary payoffs and relative payoffs. Relative payoffs are judged against
 Social motivations 29
a social reference point, assumed to be an equal share to each player, that is a share of ½
for a game with two players. The model is designed to capture play in one-shot games
and it allows for incomplete information. Given imperfect information, “observables” are
used to construct two thresholds at which behaviour deviates from monotonicity – the
proposer’s offer threshold in the dictator game and the responder’s rejection threshold in
the ultimatum game.
From this, Bolton and Ockenfels construct a general motivation function which incorporates
monetary payoffs and relative shares. They illustrate their hypotheses using an example
motivation function for a two-player game, incorporating a social reference point of equal
shares. This function is also constructed so that larger deviations from equal share have
proportionately larger impact in lowering utility. Each player’s type is given by the ratio
of weights they assign to pecuniary versus relative payoffs, with the weights determined
by individual differences, for example in age, education, politics and religion. Narrow
self-interest increases as the weight on pecuniary payoffs increases. The mathematics of
their model is explored in the Mathematical Appendix A2.1.
Bolton and Ockenfels argue that their model is consistent with experimental evidence
from a range of experimental contexts – including dictator and ultimatum games when
generosity is observed, and auction games which reveal competitive play as predicted
by standard game theory models. They argue that their model can reconcile apparently
anomalous findings. This develops findings by Roth et al. (1991) who identified differences
in outcomes from market games versus bargaining games including the ultimatum game.
Roth et al. describe evidence from experiments in Israel, Japan, the US and Yugoslavia.
Market games converged onto the equilibria predicted by standard models but there was
a wide variety of standard and non-standard outcomes for ultimatum games, suggesting
that cultural differences may play a role in the formation of other-regarding preferences.
Bolton and Ockenfels’ emphasis on narrow self-interest means that their model fits
more neatly with mainstream assumptions of rationality but they assert that their motiva-
tion function can also connect different social preferences seen in different contexts, for
example sometimes people exhibit social preferences for fairness; other times they want
to reciprocate and other times they want to compete. So the ERC model can reconcile
evidence about equity, reciprocity and generosity with evidence about competitive self-­
interested behaviour, allowing heterogeneous preferences to reflect differing motivations.
The ERC model is also consistent with interplays of equity and strategy. For example, gen-
erous offers in ultimatum games may not reflect pure altruism and inequity aversion but
may also reflect proposers strategically anticipating how responders might react.

Inequity aversion in Fehr and Schmidt’s Fairness, Competition

and Co-operation (FCC) model
Fehr and Schmidt’s model develops Loewenstein, Thompson and Bazerman’s (1989) anal-
ysis of aversion to unequal outcomes within an experimental context. They analyse sub-
jects’ responses to hypothetical dispute scenarios in which the experimental subject is
either acknowledged or snubbed by a hypothetical partner in a range of scenarios includ-
ing non-business interactions with a peer and business interactions between a customer
and sales person. Subjects were asked to grade their satisfaction with the hypothetical
solution. Loewenstein et al. used the fit of reported satisfactions to a range of social util-
ity functions and found that a utility function which incorporated discrepancies in the
30 Motivations and incentives

payoffs between the members of each pair, that is, a function incorporating payoffs in
which the experimental participant the subject got a better outcome than their partner,
and payoffs in which the participant’s outcome was worse than their partner’s, correlated
well with reported satisfaction.
Fehr and Schmidt develop Loewenstein et al.’s (1989) model to construct their own
model of Fairness, Competition and Co-operation (FCC) capturing the impacts on utility
of what they describe as inequity aversion, that is, dislike of unequal outcomes. In this model,
utility is affected by payoff differentials between players. If a person has a smaller payoff
than other players then this will lower their utility; and if they have a larger payoff then
that will raise their utility, as described mathematically in Mathematical Appendix 8.2.
Fehr and Schmidt describe two forms of inequity aversion: advantageous inequity
aversion – inequity aversion experienced by people in a position of relative advantage; and
disadvantageous inequity aversion – inequity aversion experienced by people in a position
of relative disadvantage. For example, a banker who is unhappy about poverty is expe-
riencing advantageous inequity aversion; a poor person who is unhappy about bankers
being rich is experiencing disadvantageous inequity aversion.
Overall, Fehr and Schmidt’s (1999) insights are somewhat similar to Bolton and Ock-
enfels (1998, 2000) though Fehr and Schmidt are less reliant on an assumption of narrow
self-interest and assert that seemingly irrational behaviour in ultimatum games and dicta-
tor games reflects inequity aversion. Like Bolton and Ockenfels, Fehr and Schmidt claim
that their FCC model can reconcile a range of experimental results showing that people
are sometimes cooperative and sometimes selfish – in contrast to standard approaches
predicting self-interested and uncooperative behaviour in all contexts. The FCC model can
capture the range of responses including the fairness and cooperation seen in ultimatum
games, public good games with punishments and gift exchange games (introduced on
p. 24). It can also explain the uncooperative competitive behaviour observed in market
games and public good games without punishment.
Fehr and Schmidt’s assertion about the generality of their model has created some
controversy. Binmore and Shaked (2010a, 2010b) assert that the empirical support for the
inequity aversion model is not as robust as Fehr and Schmidt claim. Fehr and Schmidt
(2010) respond that Binmore and Shaked have not posed a fundamental challenge to their
approach. The dispute remains unresolved. More nuanced empirical testing of the various
theories of inequity aversion will be helpful in resolving the debate.

Extending behavioural motivation models

The models presented above are an introduction to other-regarding preferences but many
behavioural economists argue that they don’t go far enough in properly embedding the
range of non-economic incentives and motivations driving behaviour. The social mo-
tivation models outlined above incorporate social elements as not much more than an
“add-on” to neoclassical utility theory. Other behavioural economic models designed to
capture social motivations build more links with other social and behavioural sciences,
especially with psychology and sociology. Some of these insights link to elements similar
to the Fehr and Schmidt model but with stronger psychological foundations. These in-
clude models of fairness, kindness and reciprocity, for example Rabin (1993), Kahneman,
Knetsch and Thaler (1986), and Falk and Fischbacher (2006). Fehr and Gintis (2007) assert
that formal models of inequity aversion (Fehr and Schmidt 1999) and reciprocal fairness
 CHOICE OVERLOAD 31
(Falk and Fischbacher 2006) are indispensible tools in analyzing social phenomena but
there is a large number of other theoretical explanations for generous behaviour.
Other models have focused on the emotional satisfaction derived from giving, for
example Andreoni’s (1990) model of warm glow giving capturing the fact that utility
increases with the act of giving, for example in donations to public goods. Theories of al-
truism and warm glow giving have been applied to ordinary daily activities including the
trade-offs between volunteering and work. For example, Sauer (2011) analyses volunteer-
ing and finds that it is optimal when warm glow and expected future economic returns
sufficiently outweigh disutility of extra work effort and childcare costs. Highly-educated
women receive more ‘warm glow’ than men, and there are substantial economic and
non-economic returns to volunteering, with returns to effort particularly high for those
in part-time work.
Another set of theories revolves around the insight that generosity and altruism may
represent investments in our interpersonal relationships and connections, as seen in social
capital and social network theories – enormous literatures in themselves and beyond the
scope of this book. For those interested to learn more, see Dasgupta (2007), Goyal (2009)
on social networks and Halpern (2005), Putnam (2001) on social capital.

CHOICE OVERLOAD
Neoclassical utility theory assumes that people’s choices are constrained just by monetary
constraints – specifically prices, income and wealth. Behavioural economics also captures
a range of non-monetary, psychological/cognitive constraints which interact with some
of the behavioural incentives and motivations explored in this chapter. Some of the main
cognitive constraints include choice overload and information overload. The standard
model in economics focuses on the importance of choice and implicitly it is assumed that
the more choices we have the better we will be. Psychological research has shown, how-
ever, that too much choice – choice overload – can be demotivating. Iyengar and Lepper
(2000) describe findings from three experimental studies illustrating the negative impacts
from having too much choice. For all studies, participants were more satisfied with their
selections and performed better when their set of options was limited. The first study ex-
amined the behaviour of shoppers at an upmarket grocery store. A tasting booth displayed
either a limited choice of six jam flavours or an extensive choice of 24 jam flavours. Mo-
tivation was explored in two ways: the number of shoppers drawn to the booth and the
number of jams they bought. For 242 subjects exposed to the extensive choice treatment,
145 stopped at the booth but only four bought jam. On the other hand, for 260 subjects
exposed to the limited choice treatment, 104 stopped but 30 of those bought some jam.
The study revealed that there was no correlation between the two forms of motivation:
40% of shoppers stopped at the booth displaying a limited range and 30% of them subse-
quently bought some jam. For the group encountering a booth with an extensive selection
of jams: 60% of customers stopped but only 3% of them bought jam. For the second study,
students were offered six or 30 potential essay topics for optional credit. Motivation was
assessed in terms of number and quality of essays written. In the limited choice condition,
74% wrote an essay and the mean grade for these essays was 8.09. In the extended choice
condition 60% wrote essays, with a mean grade of 7.69. For the third study, Iyengar and
Lepper studied the choices of university students choosing chocolates and extended their
experimental design by introducing three treatment conditions: no choice, limited choice
32 Motivations and incentives

(from six types of chocolate) and extensive choice (from 30 types of chocolate). The sub-
jects in the extensive choice group reported feeling that they had been offered too much
choice and those who had a limited set of choices experienced more satisfaction. All these
studies suggest that too much choice can be demotivating.
In capturing some the links with attention, Reutskaja et al. (2011) use eye-tracking
techniques to detect how visual attention shifts when people are engaged in these choice
tasks. The pattern of eye movements can serve as a proxy for how motivated people are by
different stimuli. They apply this to studies similar to Iyengar and Lepper’s (2000) choice
overload experiments. Eye tracking monitors the direction of a person’s gaze and uses this
information to make inferences about which objects are attracting a person’s attention.
Reutskaja et al. assessed the choices of hungry consumers choosing snack items but making
their decisions within a strictly limited timeframe. They found that the experimental sub-
jects assessed objects according to their appeal and this assessment correlated strongly with
willingness to pay. However, the eye-tracking evidence revealed that people had a tendency
to look and choose according to the spatial placements on items within shop displays.
The extent to which people perceive themselves to be overloaded with choices will
also depend on the timeframe within which they must make their decisions. Gabaix and
Laibson (2000) construct a model of decision-making that blends together insights about
bounded rationality, satisficing and heuristics. If people have plenty of time to decide then
they can reach the correct answer but real-life situations often involve time constraints.
Gabaix and Laibson test their hypotheses using experiments on 259 Harvard undergradu-
ates and find that their subjects decided by identifying representative scenarios. In contrast
to the backward induction hypothesized in standard game theory models, their evidence
shows that people economize by simplifying decisions using heuristics quickly to identify
feasible paths.

We have seen in this chapter that behavioural economists extend the range of incentives
and motivations that drive behaviour, including a range of intrinsic as well as extrinsic
motivations and incentives. In exploring this literature, it is important to remember that
behavioural economists do not argue against the idea that money is often a powerful incen-
tive. They are showing that we are driven by other motivations too – and sometimes these
other motivations interact with extrinsic motivations – such as money – in complicated
ways, as captured in motivation crowding and social motivation theories. These complex
influences on decision-making may also be distorted by the cognitive tools that people use
to help them decide quickly – as we shall see in the next chapter, on Heuristics and Bias.

Chapter summary
•• Behavioural economists move beyond conventional economics in exploring a wider
range of incentives and motivations, beyond financial and monetary motivations.
•• Financial and monetary motivations are captured in the category of extrinsic motiva-
tions. Intrinsic motivations are internally generated and are associated with internal
rewards – such as pride in a job done well, satisfaction in performing tasks and desire
to do good.
 A2.1 Mathematics of the Bolton and Ockenfels’ ERC model 33

•• Motivation crowding theory is about how extrinsic motivations crowd out intrinsic
motivations so that when people are paid money they are sometimes less likely to
draw on their intrinsic motivations – for example wanting to be fair to others.
•• Social motivation theory explores how social incentives are a key extrinsic motivation
and are driven by people’s perception of social norms and social reference points.
•• Social motivation theory links to models of social preference, which build on key
ideas from game theory, but with a behavioural–psychological dimension included.
•• Motivations and incentives can be dampened by choice and information overload,
and some behavioural experiments have shown that when participants are given too
many choices, participants’ performance will deteriorate.

Revision questions
1. Building on some of the examples introduced in this chapter, explore some exam-
ples of intrinsic motivation that you think are important to people’s daily lives. Do
these motivations improve behaviour, or not? Justify your answer.
2. Relative to extrinsic motivations such as money, how important are intrinsic moti-
vations to everyday decision-making? Illustrate with some examples.
3. In terms of motivation crowding theory, explore some other examples of how ex-
trinsic motivations can crowd out intrinsic motivations.
4. Outline the essential elements of social motivation theory. What does the empirical
evidence tell us about social motivations and can this evidence be explained using
insights from conventional (non-behavioural) economics? Explain your answer.
5. Is money the only extrinsic incentive? Describe and explain some examples of other
extrinsic motivations and incentives.

Mathematical appendix

A2.1 Mathematics of the Bolton and Ockenfels’

ERC model
Bolton and Ockenfels construct two thresholds at which behaviour deviates from mono-
tonicity which, given incomplete information, are based on experimental observables de-
fining the thresholds of the proposer’s most generous offer and the responder’s minimum
acceptable offer, that is:
ri( c): proposer’s offer thresholds in dictator games
s i( c): responder’s rejection thresholds in ultimatum games
Where c is the total payoff to all subjects and y i is each participant’s payoff – so:

c= ∑y i

General motivation function

From these thresholds, Bolton and Ockenfels construct a general motivation function:
v i = v i ( y i ,σ i )
34 Motivations and incentives

where y i is i’s monetary payoff; σ i is i’s relative share of the total payoff – the proportion
of the total kept by player i. Each player’s relative share is determined by y, c and n, giving:
σ i = σ i( y i , c, n )
The total payout is:
n
c= ∑y j
j=1

It follows that y i = cσ i
For example, in a two-player dictator game, the share kept by the dictator is given by
σ D and so the dictator’s payoff is cσ D and the recipient’s payoff is (1 − σ D )c.

An example motivation function

Bolton and Ockenfels’ example motivation function for a two-player game – ­incorporating
a social reference point = ½ is given by:
bi  1 2
v i( cσ i ,σ i ) = a i cσ i − σ i − 
2 2
where a i ≥ 0, bi > 0
Each player’s type is given by the ratio of weights to absolute pecuniary payoffs (the
first term on the right-hand side) versus relative payoffs (the second term on the right-
hand side), that is, a/b. Strict narrow self-interest increases as a → ∞.
b
The quadratic form allows that larger deviations from equal share have proportion-
ately larger impact in lowering utility. Weights can be assigned to different objectives
depending on context and individual differences.

A2.2 Mathematics of Fehr and Schmidt’s FCC model

Fehr and Schmidt (1999) develop Loewenstein et al. (1989)
For n players i ∈ {1,..., n} with monetary payoffs given by x = x1,..., x n the utility function
for player i is given by:

U i( x ) = x i − α i
1
n −1 ∑max {x − x ,0} − β n 1−1∑max {x − x ,0}
j i i i j
j≠ i j≠ i

In the two player case, this simplifies to:

U i( x ) = x i − α i ∑max {x − x ,0} − β ∑max {x − x ,0}

j i i i j
j≠ i j≠ i

They assume that β i ≤ α i i.e. and 0 ≤ β i < 1

Utility is maximised where x i = x j and the utility loss from disadvantageous inequity
aversion (i.e. x i < x j) is larger than the utility loss from advantageous inequity aversion
(i.e. x i > x j).