slides

The Consequences of Context for
Co-operation – Contracts,
Comparisons, and Culture
Simon Gächter
University of Nottingham, CESifo, IZA
[email protected]
ESA Copenhagen, 9 July 2010
2. Does the cultural background matter?
1. Social influence effects on
social preferences?
Previous research: Do people
have social preferences?
N
T
A
H
10
0
15
10
30
15
Typical approach: games played in
social & institutional isolation.
Results: social preferences exist, but
are heterogeneous
2
Example: The gift-exchange game
(adapted from Fehr, Gächter & Kirchsteiger, Econometrica 1997)
• Participants are randomly assigned to the roles of
“employer” and “worker”, respectively.
• Structure of the labour relation:
1. Employer:
Wage offer
[0,700]
2. Worker:
3. For both:
Payoffs realised
– Accept/reject offer
– choose costly effort
[1, 2, …, 20]
• Worker payoffs: w – c(e) (costs increasing in effort)
• Employer payoffs: ve – w (revenues increasing in effort)
3
The wage-effort relation
Gächter, Kessler, Königstein (2010)
Random matching, 30 periods
Period 1-10
Period 11-20
Period 21-30
4
But: social behaviour mostly does not take
place in a social vacuum ...
5
This talk
1. How is people’s reciprocity
influenced by other people’s
reciprocity even if there are
no material spillovers
between people?
Îmere social influence effect.
ÎSocial prefs vs. social norms.
2. How does the “cultural”
(societal) background
influence cooperation?
Î Individual heterogeneity vs.
Cultural heterogeneity.
6
Part I: Social influence effects
on reciprocity?
Thöni & Gächter (2010)
Gächter, Nosenzo & Sefton (2010a, 2010b)
7
A Three-Person Gift Exchange Game
Thöni & Gächter (2010)
Based on the two-person gift
exchange game (Fehr, Kirchsteiger, Riedl QJE 1993).
Employer
w
e1
Employee 1
e2
w
Employee 2
Course of action:
1. Employer chooses a wage
w ∈ {50,100,200} for
the two workers
2. The two workers learn w and
choose their effort ei and
ej ∈{1,2,...,20}
3. Payoff functions:
πEmployer = 35(ei+ej) – 2w
πEmployee i = w – 7(ei – 1)
• Two technologically independent
and symmetric employees,
employed by the same firm.
• Employee i’s payoff independent of
Employee j’s effort.
8
What do theories of social preferences
predict about ∂ei/∂ej ? (1)
• Money-maximization: ∂ei/∂ej = 0.
• Reciprocity (Rabin AER 1993; Dufwenberg & Kirchsteiger
GEB 2004): ∂ei/∂ej = 0.
Intuition: Worker j’s action has no influence on
worker i’s payoff and is therefore neither kind
nor unkind.
• Type-based Reciprocity (Levine, RED 1998): ∂ei/∂ej = 0.
9
What do theories of social preferences
predict about ∂ei/∂ej ? (2)
•
•
Inequity aversion: Bolton/Ockenfels (AER 2000): ui(πi , σi ),
own payoff & disutility from unequal income share σi ≠ 1/3.
Assume a very strongly inequity averse worker i who only
cares about σi. He/she chooses ei such that σi = 1/3.
w − 7(ei − 1)
σi =
(v − 7)(ei + e j ) + 14
Assume w = 200; v = 35.
200 − 7(ei − 1)
1
Worker i chooses ei such that σ i =
= .
(35 − 7)(ei + e j ) + 14 3
⇒ ei (e j ) =
607 − 28e j
49
⇒ −1 <
de j
dei
< 0.
→ Efforts are substitutes!
10
What do theories of social preferences
predict about ∂ei/∂ej ? (3)
πP = πi
πi = πj
11
A summary of theoretical predictions about
∂ei/∂ej
• Efforts are unrelated:
– Money maximization
– Levine (1998)
– Dufwenberg & Kirchsteiger (2004)
• Efforts are strategic substitutes:
–
–
–
–
–
–
Fehr & Schmidt (1999)
Bolton & Ockenfels (2000)
Falk & Fischbacher (2006)
Charness & Rabin (2002)
Cox, Friedman & Gjerstad (2007)
Cox, Friedman & Sadiraj (2008)
• Efforts are strategic complements:
– Fehr & Schmidt (1999)
– Charness & Rabin (2002)
12
Comparing predictions with data
Thöni & Gächter (2010)
Most robust prediction by all
theories of social preferences
Results qualitatively consistent
with Fehr-Schmidt ...
... but also with social norms and conformity
13
A related approach using the strategy method
Gächter, Nosenzo & Sefton (CeDEx DP 2010-10)
• Trilateral Gift-Exchange Game: 1 Employer, 2 Employees.
• Sequential play: Employer chooses wages (16 or 32) Î Employee 1
chooses effort (1 to 4) Î Employee 2 chooses effort (1 to 4).
• Employee 2 conditions effort on wages and Employee 1 effort:
• 1) Norm of Reciprocity + Social Influence? 2) Inequity aversion?
Understanding Social Influence Effects:
Social Norms or Social Preferences?
Gächter, Nosenzo & Sefton (soon to be submitted ...)
Design elements:
1. Replicate Gächter, Nosenzo & Sefton (CeDEx DP 2010-10)
2. Measure social norms directly (using a design inspired by
Krupka & Weber 2010)
3. Design a treatment where Fehr & Schmidt predicts that
efforts are unrelated.
15
Measuring social norms
(after Krupka & Weber 2010: “Identifying social norms using coordination games”)
• Describe the situation and ask participants what they think how socially
appropriate a particular effort level of Employee 2 was.
– “very inappropriate”, “inappropriate”, “appropriate”, “very appropriate”
Weighted appropriateness
• Incentive for correctly estimating what majority thinks social norm is
1
0.8
0.6
0.4
0.2
"How socially appropriate is Employee 2's
effort level ..."
Employee 1
Wage1 = 32
chooses:
Wage2 = 32
effort1 = 1
***
0
-0.2
-0.4
***
-0.6
-0.8
-1
1
2
3
4
... if Employee 2 chooses effort level ...
effort1 = 4
Results: Employees 2 in Baseline (n=27)
• When own wage is 32 and co-worker wage is 32, positive relation
between employees’ efforts (Page test: p<0.01).
ÎWe replicate Gächter, Nosenzo & Sefton (CeDEx DP 2010-10)
Î consistent with Fehr & Schmidt and social norms.
Experimental Design: the Random Treatment
One-shot play of modified game:
• After Employee 2 choice, Nature randomly selects one of the two
Employees (50-50 chance). Only wage and effort decisions regarding
the selected employee are implemented.
• Modified payoff functions:
π Employer
⎧20 + 10 ⋅ (e1 ) − w1 if Empl. 1is selected π = ⎧⎨wi − 5 ⋅ (ei − 1)
i
=⎨
⎩0
⎩20 + 10 ⋅ (e2 ) − w2 if Empl. 2 is selected
if empl. i is selected
if empl. i is not selected
• Employee 2 best-reply predictions of F & S model (β>0.5):
Measuring social norms in RANDOM
(after Krupka & Weber 2010: “Identifying social norms using coordination games”)
"How socially appropriate is
Employee 2's effort level ..."
Weighted appropriateness
1
0.8
0.6
Employee 1
chooses:
Wage1 = 32
Wage2 = 32
0.4
0.2
***
0
-0.2
***
-0.4
-0.6
-0.8
-1
1
2
3
4
... if Employee 2 chooses effort level ...
effort1 = 1
effort1 = 4
Results: Employees 2 in Random (n=27)
ÎNo relation between employees’ efforts when both wages are 32
(Page test: p>0.22). Only 3 out of 27 Employees 2 change effort
across different levels of co-worker’s effort.
Î Results consistent with Fehr-Schmidt prediction.
Part II: Cultural influences on
voluntary cooperation?
Gächter, Herrmann & Thöni (2010), based on
Herrmann, Thöni & Gächter, Science 2008
21
Example: Public goods (with punishment)
Gächter, Renner & Sefton (Science 2008)
Experiments in Nottingham, UK; N=207; fixed groups (n=3)
Costs of punishment
2
Mean cost of punishment
1.8
1.6
1.4
1.2
1
0.8
P10
0.6
P50
0.4
0.2
0
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49
Period
22
The free rider problem is ubiquitous, it seems …
Example from Herrmann, Thöni & Gächter, Science 2008
4-Person Public Goods,
fixed matching, t=10,
16 subject pools
Explanation:
Conditional
cooperation.
Fischbacher &
Gächter,
American
Economic
Review, 2010.
23
Cooperation in the presence of punishment
Herrmann, Thöni & Gächter, Science 2008
24
Antisocial Punishment Across Societies
Herrmann, Thöni & Gächter, Science 2008
Punishment
Punishment of
of
Punishment of
free
free riding
riding
co-operators
Antisocial punishment is
significantly correlated
with (at the country level):
The Rule of Law; Norms of
Civic Co-operation,
Democracy, GDP/capita,
Individualism, Social
Equality
25
Cultural influences on cooperation levels
Gächter, Herrmann & Thöni (2010); Data from Herrmann et al. 2008
Cultural areas according to Baker and Inglehart, Am Soc Rev 2000 and Hofstede 2001
26
Cultural influences on (antisocial) punishment
27
Distribution of individual average contributions
28
Distribution of group average contributions
29
Analysis of variance
30
“Out-of-Sample Prediction” (1)
• We observe similar behaviour in culturally similar subject pools.
• Idea: run new experiments in a culturally similar society.
• Experiments in Iasi (Romania, June 2010): culturally similar to Russia,
Belarus, and Ukraine
20
Mean contributions
18
16
14
12
10
8
Minsk
Samara
Dnipropetrovs’k
Iasi
6
4
2
0
1
2
3
4
5
6
7
8
9 10
1
Period
2
3
4
5
6
7
8
9 10
31
“Out-of-Sample Prediction” (2)
2.5
Minsk
Mean punishment
2
Samara
Dnepopetrovsk
Iasi
1.5
1
0.5
0
[-20,-11]
[-10,-1]
[0]
[1,10]
[11,20]
Contribution of punished subject MINUS
contribution of punishing subject
32
Summary
• Previous research has studied social preferences in socially
isolated situations.
• But people do not act in social isolation. Important to study
social context.
• Behaviour of others matters for own behaviour not just for
reasons of reputation or conditional cooperation.
• Behaviour of others tells you what is socially appropriate.
– How big is the moral wiggle room?
• People’s social preferences are also shaped by the wider
society/culture they live in.
– Buchan et al, PNAS 2009
– Henrich et al, Science 2010
– Henrich et al, The Weirdest People in the World, BBS 2010
• THANK YOU!
33