Game Theory, Behavior and The Law - I

A brief introduction to game theory. Rules of

the game and equilibrium concepts.
Behavioral Games: Ultimatum and Dictator
Games. Entitlement and Framing effects.

Emanuela Carbonara

University of Bologna - Department of Economics

31 January 2018
Outline

Perfect Rationality and Game Theory

A Brief Overview of Game Theory and Equilibrium Concepts

The Ultimatum Game

Framing Effects

Entitlements

The Dictator Game

What is Game Theory?

I Considers environments with strategic interaction:

I i.e. environments in which the payoff of a(t least one) player
depends on her/his own decision but also on other players’
decisions.
I This implies that when a players chooses her/his action she/he
must form her/his beliefs of other players’ actions and how
her/his decision is going to affect other players’ decisions.
Game Theory v. Decision Theory

I Game Theory:
I Strategic Interaction
I Decision theory:
I The problem of optimal decision of an individual with no strategic
interaction
I In some environments, GT or DT depends on the specific case:
I Example: behaving at the restaurant, are you alone or with some
friends (splitting the bill? If you do, how do you order?)
Normative and Positive Game Theory

I Positive
I GT tries to explain and predict behavior in real contexts.
I People learn about how to play, i.e. they learn what is a rational
way of playing (what is best for them), period after period.
I Although they may temporarily depart from the given definition of
”rationality”, they may tend to it and, also, mistakes by many
players may cancel out in the aggregate.
I This positive perspective is very ambitious: departures from
rationality are many and unexpected.
I Normative
I GT investigates what a rational player should do.
I What does rationality mean? In short: choosing the ”best” option,
the ”best response”.
I Somehow recently: new tools for positive analysis (but also for
normative):
I Experimental economics.
I Behavioral economics tries to model non-rational behavior.
Behavioral Game Theory - An Alternative Approach

I Behavioral game theory starts from the observation that not

every player behaves rationally in complex situations.
I Not all players form beliefs based on an analysis of what others
might do (violation of strategic thinking);
I not all players choose the best response given their beliefs, or at
least it is not clear how they define their best response.
I The predictions of traditional game theory are sharp.
I It is not difficult to spot deviations in behavior.
I Deviations are mainly documented by experimental literature.
I Then models are presented, often specialized to the game at
hand, and theories are formulated, aiming to explain deviations.
I The difficult part is to formulate a theory that is general enough
to encompass different types of deviations. Such general theory
is still missing.
How To Deal With Deviations From Rational Behavior

I Focus on specific games.

I Analyze the deviations from equilibria predicted by the
traditional theories and present the most important theories put
forward to explain such deviations.
Setting Up the Stage

I First we need to learn how to describe and define a game

I Then we will learn how to ”solve” it (positive and normative
perspectives)
I To define a game we need (informally):
1. The players
2. The rules of the game
3. The outcomes (for each set of actions taken by players)
4. The payoffs (preferences over outcomes)
I Two descriptions of games: normal (or strategic) form and
extensive form representations
Games in Normal Form

I With few players in some special (mostly pedagogical) cases,

normal form games can be represented with matrices
I The prisoners’ dilemma game
Games in Extensive Form

I In a dynamic game, players may act sequentially, they may act

several times, and their information may depend on what
happened in the past.
I When we need to describe the dynamic in a game, we need
another representation that allows us to answer the following
questions:
I Who plays when?
I What can players do?
I What do they know?
I What are their payoffs?
Games in Extensive Form II

I The extensive form of a game describes:

1. the set of players;
2. at which point in time which player is called to move;
3. which actions are feasible for the players when they are called to
move;
4. what players know about the previous history of the game when it
is their turn to move;
5. the payoff of each player as a function of all the possible final
histories of the game.
Example of an Extensive Form

I In many cases the extensive form of a game can be represented

nicely by a game tree.
Equilibrium and Solution Concepts

I We aim to to find solutions of the game that allow us to

understand how rational players would act.
I The first step will be simply to rely on rationality and common
knowledge of it.
I ...with the concepts of dominance and rationalizability.
I Then we introduce equilibrium concepts to predict of haw
players play.
I Equilibrium requirements include also the assumption that
players have (and this is common knowledge) mutually correct
expectations.
Dominant and Dominated Strategies
Solving a Game Using Dominated and Dominant
Strategies

I Definition: A game is solvable by iterated (strict) dominance if,

for each player i, the set of undominated strategies after a
”sufficient” number of iterations contains a unique strategy.
I We use rationality (& common knowledge of it)
I Consider again the prisoners’ dilemma game
I The game is (trivially, since just 1 iteration) solvable with deletion
of dominated strategies: (D,D)
I Notice: The outcome of the game is not the best for both players
from a cooperative view point!
I How can it be possible that they are not able to cooperate?
I Experimental evidence shows that sometimes (often) individuals
play (C,C).
Iterated Deletion of Dominated Strategies

I How can we solve this? In steps!

I R is strictly dominated and can be eliminated:
I First implication: 2 will never play R
I Second implication: 1 cannot expect 2 to play R
I But now, once R is ”deleted”, D is strictly dominated and can be
also deleted:
I 1 will never play D.
I Hence: 2 cannot expect 1 to play D.
I But now L is strictly dominated.
I We reached a unique undominated strategy: It follows that the
reasonable profile of strategies (the equilibrium) is (U,M).
Nash Equilibrium

I In many games there are no dominant nor dominated strategies.

Which strategies should we expect?
I Nash: strategies should form an ”equilibrium”:
I each player should have a belief on other players strategies such
that his choice is a best response to these beliefs AND these beliefs
turn out to be correct (mutually consistent)
I such consistency implies that nobody has an incentive to deviate
from this strategy
A Formal Definition

Definition
Nash Equilibrium. The strategy profile s∗ = (si∗ , s∗−i ) is a Nash
Equilibrium (NE) if each player’s strategy is a best response to the
other players’ strategies (to the expectation thereof). That is, no
player has incentive to deviate, if no other player will deviate. (If
players find themselves in equilibrium, there is no reason to move
away.)

0 0
ui (si∗ , s∗−i ) ≥ ui (si , s∗−i ) ∀si 6= si∗

Definition 0
Best Response. A strategy si is a best response (BR) for player i to the
other players’ strategies s−i if
0 0
ui (si , s−i ) ≥ ui (si , s−i ) ∀si 6= si
How to Find a Nash Equilibrium

I In a two-player game with a finite number of strategies it is easy

to find all Nash equilibriums:
I Simply check all strategy profiles (”combinations”)
I If strategy spaces are continuous, look at the first order conditions
in order to find ”best response functions”. (Also check the second
order conditions)
Dynamic Games

I Many strategic contexts are dynamic

I This per-se does not necessarily require a new kind of analysis,
in fact
I We can transform almost any dynamic game into its normal form
representation
I And then use the equilibrium concepts analyzed so far for static
games
I However there is a problem with this:
I In dynamic games this procedure may lead to equilibria that do not
make very much sense and the issue is the one of CREDIBILITY of
a player’s strategy
I We then need to REFINE the equilibrium set, eliminating those
unreasonable equilibria
I This is what makes the analysis of dynamic games different!
Dynamic Games and Sequential Rationality
I Another way to see the problem with NE (N;f) in the Entry
game:
I actions at decision nodes not reached by play of the equilibrium
strategies (here, firm I’s action at the decision node following firm
E’s not chosen E) do not affect firm I’s payoff
I As a result, firm I can ”announce” or plan to play whatever
strategy after N
I The problem is that what firm I’s strategy ”says” it will do at the
unreached node can actually insure that firm E, taking firm I’s
strategy as given, wants to play ”out”
I We can rule out those non credible announcements or
predictions such as (fight if firm E plays ”E”) by ALWAYS
requiring that wherever a player is called to act, she must play
an optimal strategy at that point
I It is the principle of sequential rationality: rationality at ANY
point in which you are called to take an action, even if it is out of
the equilibrium path!
I In the example playing f after E is not sequentially rational for I
and then we cannot expect it, we can only expect a after E:
I The NE (N;f) does not satisfy the sequential rationality
requirement!
Sequential Rationality in Action

I For the class of finite games with perfect information there is a

simple way to apply the requirement of SR, it is the backward
induction principle or method
I The idea is: first solve for optimal behavior at the ”end” of the
game (in the entry example: at the post-entry decision node) and
then determine what is the optimal behavior at earlier (upward)
nodes in the game given the prediction of this later behavior
I Note: after solving optimal (rational) behavior at lower nodes,
we can redefine a reduced game where we ”compact” what will
happen at later nodes by anticipating what we have seen is the
optimal behavior at those later nodes
I An example...
Ultimatum Game

I Bargaining game with Take-it-or-leave-it offers.

I There is only one round;
I Sequential moves;
I A proposer (A) and a receiver (B) bargain about splitting a surplus
(say, 10 Euros)
I Player A makes an offer about splitting such surplus;
I If player B does not accept, bargaining is over and both players get
nothing;
I Theory predicts that B should rationally accept any offer that A
makes (including zero).
I Experimental evidence contradicts this conclusion.
Ultimatum Game. Experimental Evidence

I Güth,W., R. Schmittberger, And B. Schwarze (1982): ”An

Experimental Analysis of Ultimatum Bargaining,” JEBO.
I According to theory, players should accept any offer they
receive, no matter how small
I When proposer and chooser are together in the same room
playing the experiment, the most common offer is a 50:50 split
I Also when bargaining is anonymous (through a computer) offers
are seldom worse than 75:25
I More precisely (Camerer, 2003):
I Modal and median ultimatum offers are 40-50% and means are
30-40%.
I Offers in the very low range 0, 1-10% and in the “hyper-fair” range
51-100% are seldom proposed.
I Offers of 40-50% are rarely rejected, whereas offers below 20% are
rejected half of the times.
Why are Responders “Irrational”?

I Many individuals have preference for being treated fairly.

I For many people, a fair split of $ 10 is close to 50:50 (the concept
of fair split may change - but it does not, apparently - when
stakes change).
I Rejecting (substantial) sums could be evidence of “negative
reciprocity”:
I responders reciprocate unfair behavior by punishing the person
who treated them unfairly, even at a cost to themselves (burning
money).
Fairness and Negative Reciprocity

I Where do preferences for fairness come from?

I Evolutionary adaptation: getting angry when treated badly
because getting angry had survival value when interacting within
the same group of people in a relatively small group (Frank, 1988,
Passions within reason: the strategic role of emotions).
I Fairness standards are created by culture, and may differ according
to the closeness of kin relations or the degree of anonymity in
market exchanges.
A Cross-Cultural Study of the Ultimatum Game

I Henrich et al. 2005. ”’Economic Man’ in Cross-Cultural

Perspective: Behavioral Experiments in 15 Small-scale Societies,”
Behavioral and Brain Sciences, 28(6), pp. 795-815 provides a
formidable cross cultural study.
I Culture seems to matter a lot on how people ”play” the
ultimatum game.
I Henrich and colleagues did the ultimatum game (and other
games) in some simple societies (with Pittsburgh and LA as
controls).
Map of Experiments
Results
Discussion

I They showed that some of these societies were playing

remarkably close to the self-interested player, with very little
offers and very low rejection rate.
I Some others were characterized by ”hyperfair” offers, with a
curious interpretation (insulting competitive gift-giving).
I Generally, societies with lower offers are socially disconnected,
with a very low rate of anonymous transactions in the village.
I Seemingly, average offers are positively correlated to the degree
of market development and anonymous interaction.
I The more socially connected societies, the more likely that they
develop social sharing norms and a fairness standard.
Framing Effects in Ultimatum Games. Hoffman et al.
(1994)

I It seems that the way in which strategies are described exerts a

strong influence on the way people behave in the ultimatum
game.
I In one study, the ultimatum game is described as an exchange: a
seller setting a price for a good that the buyer can take or leave
(Hoffman, McCabe, Shachat and Smith (1994), Preferences,
Property Rights and Anonomity in Bargaining Games, Games and
Economic Behavior, 7: 346-80).
I Rejection rates are unchanged, but offers decrease by almost 10%.
Instructions. Hoffman et al. (1994)
Framing Effects in Ultimatum Games. Larrick and
Blount (1997)

I They frame the ultimatum game as a ”common pool ” or

”resource dilemma” ( Larrick and Blount (1997), The claiming
effect: Why Players are More Generous in Social Dilemmas than
in Ultimatum Games, Journal of Personality and Social
Psychology, 72: 810-25).
I A resource dilemma is a sequential game in which players make
claims from a fixed common pool of resources and get nothing if
their claims add up to more than the pool.
I The first player’s claim x from the pool is simply an ”offer” to the
second player, who can then claim the remaining Y − x or ”veto”
the 1st player’s claim by claiming more than Y − x, exactly as in
a ultimatum game.
Larrick and Blount (1997). Results

I They find that offers in the resource dilemma are more generous
(not by much though) and rejection rates lower than in a typical
ultimatum game.
I They conclude that framing the game as a common resource
dilemma creates a ”sense of common ownership that makes
offers more generous on both sides”.
Entitlement Effects

I Sometimes individuals could feel ”entitled” to a certain share of

the surplus.
I It might be an ”endowment effect”: since they have been ”given”
the surplus to share, they feel they have the right to keep the
higher share.
I It might be a matter of self- perception: they feel somehow more
deserving.
I They feel they own the surplus.
How to Generate Entitlement

I A very common way to generate entitlement effects in

experiments is to have subjects earn their endowment by
performing some task (like answering questions, cutting paper,
sealing envelopes, etc.)
I Hoffman et al. (1994) have a treatment in which they allocate the
role of proposer to the individual in the pair who answers the
higher number of general knowledge questions.
I This lowers offers by 10%.
I However, it has the perverse effect of increasing rejection rates in
a statistically significant way.
I The sense of entitlement generated by performing better in the
quiz is not shared by responders (who do not feel more ”obliged”
to accept lower shares).
I Apparently, responders form self-serving beliefs about the
legitimacy of the proposer’s entitlement.
The Dictator game

I Participants to the experiment are matched in pairs

I Player A is designated to decide the split
I Player B is forced to accept whatever split A decides
I This game tests the determinants
I Fear of rejection
I Here there cannot be any
I Sense of fairness
I The result is that a majority of A players chooses to keep no more
than 70
Question: Fairness Standards and Social Norm
Compliance are Irrational Behavior?

I Rejections in the ultimatum game simply mean that responders

are not maximizing their monetary earnings.
I They don’t (necessarily) mean that responders are not capable of
strategic thinking.
I Responders might get extra (intrinsic) utility from punishing
proposers who do not respect fairness standards or violate social
norms.
I Proposers might be reluctant to infringe social norms and offer
too little for fear of a social sanction (hence, proposers might
even be maximizing their monetary earnings avoiding the social
sanction).