Altruistic Punishment

in Humans
Ernst Fehr & Simon Gchter
Clemente Jones & Nguyen Lam
Psychology 459
05/08/2014

Introduction
Evolution of Human Cooperation:
Kin Selection Cooperation among genetically
close individuals.
Direct Reciprocity Selfish incentives for
long-term bilateral cooperation.
Indirect Reciprocity Cooperators build a
reputation.
???? Non-repeated cooperation among
genetically unrelated people with no reputation
gains.

Solution
Punishment:
Group better off if free riding is deterred
No one has incentive to punish because it
costs themselves as well as free rider
Punishment of free riders = Second-order
public good
Can work if enough people punish free
riders altruistically, with cost and without
material benefits for the punishers.

The Question
Do humans engage in altruistic
punishment, and if so, how does
this inclination affect the ability of
achieving and sustaining
cooperation?

Participants
240 undergraduate students from the
University of Zurich (Switzerland) and the
Federal Institute of Technology
(Switzerland) voluntarily participated in
the experiments.
31% Females
69% Males
Different majors
33 students control group for the
emotion questionnaire.

Design
Pre-Study
Participants were randomly assigned to one
of ten experimental sessions (with 24
subjects/session).
24 subjects allocated to six groups of four.
Each of the 24 subjects played two 6-period
games:
Punishment
No Punishment
Each participant was placed in front of a
computer in a booth such that subjects could
not see each other.

Design Cont.
Each member of the group received an
endowment of 20 MUs (real monetary stakes)
and each one could contribute between 0-20
MUs to a group project.
For every MU invested in the project, each of the
four group members earned 0.4 MUs, regardless
of whether he or she made a contribution.
Selfish: 20 MUs
Cooperate: 32 MUs
Subjects made their investment decisions
simultaneously and, once the decisions were
made, they were informed about the
investments of the other group members.

Condition: Punishment
Subject could punish each of the other
group members after they were informed
about the others investments.
A punishment decision was implemented
by assigning between 0-10 points to the
punished member.
Each point assigned cost the punished
member 3 MUs and the punisher 1 MU.
Punishment is costly and yield no
benefits!

Results
In the 10 sessions, subjects punished other
group members a total of 1,270 times:
84.3% of the subjects punished at least
once.
34.3% punished more than 5 times.
9.3% punished more than 10 times.
The more a subjects investment fell short of
the average investment of the other three
group members, the more the subject was
punished.

Figure 1

Results Cont.
The presence of punishers establishes a
credible threat that deters non-cooperation:
Punished subjects contribute more in the
next periods.
The act of punishment, although costly for
the punisher, provides a benefit to other
members of the population by inducing
potential non-cooperators to increase their
investments.
The introduction (or elimination) of the
punishment opportunity led to an
immediate rise (or fall) in investment.

Figure 2a

Figure 2b

Why Punish in a One-Shot

Context?
Negative emotions:
Can trigger a willingness to punish
free riders, despite being costly
and yielding no direct benefit.

Emotions
Scenario 1 [2]
You decide to invest 16 [5] francs to
the project. The second group member
invests 14 [3] and the third 18 [7]
francs. Suppose the fourth member
invests 2 francs to the project. You
now accidently meet this member.
Please indicate your feeling towards
this person.

Emotions
Results 1 [2]
Anger/annoyance measured on sevenpoint scale (1 = not at all, 7 = very
much)
Scenario 1:
47% had anger levels of 6 or 7
37% had anger level of 5
Scenario 2:
17.4% had anger levels of 6 or 7
80.5% had anger levels of 4 or 5

Emotions
Scenario 3 [4]
Imagine that the other three group
members invest 14, 16 and 18 [3, 5
and 7] francs to the project. You invest
2 francs to the project and the others
know this. You now accidentally meet
one of the other members. Please
indicate the feeling you expect from
this member towards you.

Emotions
Results 3 [4]
Anger/annoyance measured on sevenpoint scale (1 = not at all, 7 = very
much)
Scenario 3:
74.5% predicted anger levels of 6 or 7
22.5% predicted anger level of 5
Scenario 4:
17.8% predicted anger levels of 6 or 7
80% predicted anger levels of 4 or 5

Controlling for Bias

Same four scenarios presented to 33
subjects that had not participated in
the experiments.
Same emotional patterns from the
240 experimental subjects were
expressed in the 33 controls.

Conclusions
Free riding causes strong negative emotions, which
most people expect.
Most punishment executed by above-average
contributors on below-average contributors (74.2%).
Punishment increases with deviation from the average
investment.
Punishment rendered immediately credible because
most people know they trigger negative emotions by
free riding:
Punishment opportunity leads to an immediate
impact on contributions (as is evident at switch points
between punishment and no-punishment conditions).

Implications
Altruistic Punishment = key force in
establishment of human cooperation.
There is more at work in sustaining
human cooperation than is suggested by
kin-selection, direct reciprocity, indirect
reciprocity, and costly signaling.
Kin Selection
Direct
Reciprocity
Indirect
Reciprocity

Altruistic
Punishment

Limitations
Selective population:
High cognitive ability
W.E.I.R.D.

Sex Ratio:
31% Females
69% Males
NOT representative of the population