Chapter 11-Behavioral Game

Theory
A Course in Behavioral Economics by Erik Angner
Instructor: Tolga Yuret
Motivation
• Analytical game theory is in many ways a huge success story: it is
increasingly becoming the foundation of other subdisciplines of
economics (including microeconomics) and it has migrated to
philosophy, biology, political science, government, public policy, and
elsewhere. But, as we will see in this chapter, its descriptive adequacy
and normative correctness are controversial.
• Behavioral game theory aims to study the degree to which analytical
game theory succeeds in capturing the behavior of real people
engaging in strategic interaction, and proposes extensions of
analytical game theory in the interest of capturing that behavior.
Social Preferences
• Much of the literature on social preferences is driven by data from
two games: the ultimatum game and the dictator game.
• Ultimatom game: In the first stage, Player I proposes a division of the
dollar amount; that is, the proposer offers some share of the $10 to
the other player. In the second stage, the responder accepts or rejects
the proposed division. If she accepts, both players receive their
proposed share; if she rejects, neither player receives anything.
• If you are the proposer how much would you offer?
• If you are the responder what is the minimum dollar amount that
would you accept?
Ultimatom Game
Ultimatom Game
Ultimatom Game
• The ultimatum game has been extensively studied by experimental
economists. According to Camerer’s survey of the results:

The results … are very regular. Modal and median ultimatum offers are
usually 40–50 percent and means are 30–40 percent. There are hardly any
offers in the outlying categories 0, 1–10, and the hyper-fair category 51–100.
Offers of 40–50 are rarely rejected. Offers below 20 percent or so are
rejected about half the time.

• How is this consistent with subgame perfect Nash Equilibrium? Offer 1

dollar and accept!
Ultimatom Game
• Let us assume that each individual is simply trying to maximize his
dollar payoffs, and that u (x)=x. Suppose that proposer offers $4.

• If the responder strategy is to reject all proposed divisions in which

proposer offers less than $4 and accepts all others.

• Is this subgame perfect Nash? Is this Nash?

• Credible vs. Incredible Threat

Ultimatom Game

• In brief, data on the ultimatum game do not represent a puzzle from

the point of view of the theory we learned in Section 10.2, since
observed outcomes are consistent with Nash equilibrium predictions.
(Given the many equilibria of this game, though, this is not saying
much.)

• A prediction based on the idea of subgame-perfect equilibrium, given

our assumption about the two players’ utility function, is in fact
inconsistent with the experimental results.
Ultimatom Game
• Homework: Suppose that you know that responder would accept the
proposer’s offer x (that the proposer gives x dollars to the responder)
with probability x. What should be proposer’s best response?
Dictator Game
• The dictator game resembles the ultimatum game, except for the fact
that the second stage has been eliminated.

• There is only one Nash equilibrium and therefore only one subgame-
perfect equilibrium: the case in which Player I offers nothing to the
responder and keeps all the money for himself.
Dictator Game
Dictator Game
• Experimental evidence suggests that proposers in the (one-shot,
anonymous) dictator game typically offer less than proposers in the
(one-shot, anonymous) ultimatum game. That said, many proposers
are nevertheless willing to share a substantial amount (10–30
percent) of their initial allocation.

• But this is inconsistent with Nash!

Social Preferences
Social Preferences
• P is said to be altruistic and to have altruistic preferences. Some
parents, relatives, friends, and admirers are willing to make real
sacrifices in order to improve another person’s situation.
Social Preferences
• There is no requirement that P derive positive utility from Q’s
attainment. Example:
Social Preferences:
• This specification entails that P’s utility goes up when Q’s attainment
goes down and vice versa. If so, P is said to be envious.
• Some Prius hybrid-car owners derive deep satisfaction from rising
gasoline prices. This cannot be explained by reference to the financial
effects of gasoline prices on the Prius owner: though fuel-efficient, a
Prius remains a gasoline-powered car, so rising gasoline prices will
hurt Prius owners too.
• But it can be explained if we assume that the disutility Prius owners
derive from getting less gasoline for their dollar is outweighed by the
utility they derive from knowing that SUV owners suffer even more.
Rawlsian Justice
• According to a common interpretation of John Rawls’s theory of
justice, societies should be ordered with respect to justice based on
the welfare of the least fortunate in each society.
• Thus, a person with Rawlsian preferences might try to maximize the
minimum utility associated with the allocation. If each individual
derives 𝑥 utiles from his or her private consumption x, the Rawlsian
P might maximize
Inequality Averse Preferences
• Another agent might care about the degree of inequality among the
relevant agents, so as to rank allocations based on the absolute
difference between the best and worst off. Such an agent is said to be
inequality averse and to have inequality-averse preferences. If each
individual derives 𝑥 utiles from his or her private consumption x, the
inequality-averse P might wish to minimize the absolute difference
between each person’s utility.
• This amounts to maximizing
Utilitarian
• Utilitarians like Bentham, whom we came across in Section 1.2,
believe that we should pursue the greatest good for the greatest
number. Thus, a utilitarian agent might try to maximize the total
amount of utility derived from private consumption. Example:

Utilitarian preferences constitute a special case of altruistic

preferences
Dictator Game Payoffs
Exercise

Homework: Solve Exercise 11.4

Utility Specification

• So you can explain all kinds of experiment results when you specify
the utility function as you wish!

• Remember, our Nash description assumed that u(x)=x.

Utility Specification

• In order to accommodate the behavior of the proposer in the dictator

game, we postulate that proposers are largely altruistic. But in order to
accommodate the behavior of responders in the ultimatum game, this
approach is inadequate.

• An altruist would prefer any outcome to ($0, $0). In a subgame-perfect

equilibrium, therefore, an altruistic responder would accept all offers.

• But this is not the case! So it is not easy to agree on a utility function that
explains ALL experimental results.
Intentions
• In a variation of the ultimatum game, responders were found to reject the
uneven division ($8, $2) if the proposer had the choice between ($8, $2)
and the even division ($5, $5), but accept it if the proposer had the choice
between ($8, $2) and the maximally uneven division ($10, $0).

• This makes no sense from the point of view of a responder who evaluates
final outcomes in accordance with either one of the social preference
functions.
• According to each of those models, either ($8, $2) is betterthan ($0, $0) or
it is not; the choices available to the proposer do not matter at all!
Intentions
• One explanation: People are willing to reward people who are perceived as
having good intentions and to punish people who are perceived as having bad
intentions.
• A proposer who offers $2 rather than $0 is interpreted as having good intentions,
even if the resulting allocation is uneven, whereas a proposer who offers $2
rather than $5 is interpreted as having bad intentions.

• Sometimes, these results are discussed in terms of reciprocity or reciprocal

altruism. Respondents are said to exhibit positive reciprocity when they reward
players with good intentions and negative reciprocity when they punish
proposers with bad intentions.

• Thus, a responder in the ultimatum game who rejects a small positive offer from
the proposer is said to exhibit negative reciprocity.
Trust Game
• a sender (Player I) and a receiver (Player II)

• In the first stage, the sender sends some share $x (called invested
amount) of his $10 to the receiver.

• In the second stage, the receiver returns to the sender some share $y
of her total allocation $10 + 3*$x

• The final outcome, then, is ($10 − $x + $y, $10 + 3*$x − $y)

Trust Game
• When both agents maximize u (x) = x, then subgame perfect Nash: $y=0 so
$x=0!

• What is the Pareto efficient outcome?

• What is the outcome if the players have altruistic utility functions?

• In the experiments senders send around $5 and receivers to return around

$4 and $5.
Public Goods Game

• Three players each endowed with $10

• Each contribute some amount. Total is multiplied by two and

distributed back.

• What is Nash?

• What is the pareto optimal solution?

Public Goods Game
• In anonymous, one-shot public-goods games, Robyn M. Dawes and
Richard H. Thaler report:

• While not everyone contributes, there is a substantial number of

contributors, and the public good is typically provided at 40–60
percent of the optimal quantity.
Public Goods Game
• Why do people cooperate in one-shot public-goods games? The
experimental results are consistent with a high level of trust in the
other players and with a desire to reciprocate what they expect to be
generous contributions from the others.

• Contrary to predictions in our discussion about cheap talk, pre-play

communication actually increases cooperation in public goods games.

• It should be noted that when the game is repeated, the level of

contributions tends to decline.
Public Goods Game
• Predictions based on egoistic utility functions in combination with
game theoretic equilibrium concepts suggest that people will be
unable to coordinate their actions even when it is in their interest to
do so.

• Yet, this is a needlessly pessimistic vision of human nature. The

economist Elinor Ostrom won the 2009 Nobel Prize for exploring ways
in which people develop sophisticated mechanisms that allow them
to reach beneficial outcomes in trust and public goods style games
Beauty Contest Game (Guessing Game) -
Keynes
Beauty Contest Game (Guessing Game)
• Here, n players simultaneously pick a number between zero and 100
inclusive. The person whose number is closer to seven-tenths of the
average number wins a fixed prize.

• Nash Equilibrium: Everyone picks zero!

• Real people, however, do not play the Nash equilibrium strategy in

the one-shot game. When played for the first time, answers might fall
in the 20–40 range.
Beauty Contest Game (Guessing Game)
• Interesting things happen when the game is repeated with feedback
about the average number in the previous round. During subsequent
rounds, the average number will decrease and eventually approach
zero. This suggests that, over time, real people converge to the Nash
equilibrium prediction.
Beauty Contest Game (Guessing Game)
Beauty Contest Game (Guessing Game)
• One fascinating feature of this game is that, even if you know the unique
Nash equilibrium strategy, you may not want to play it. As long as you
expect other players to play out-of-equilibrium strategies and choose a
positive number, you will want to do so too.

• And as long as other players expect you to expect them to pick a positive
number, they will want to pick a positive number. And so on.

• Thus, everyone plays a positive number and an out-of-equilibrium strategy.

But, although you want to pick a number greater than zero, the number
must not be too high: the aim is to stay one step ahead of the other
players.
Beauty Contest Game (Guessing Game)