Fairness and Bargaining

1
Fairness and Bargaining

• Experimental evidence for fairness
• Models of fairness
• Testing fairness models
• Field Evidence on social preferences

2
Fair Behavior

• Preferences for fair (equal) outcomes
• Unconditional behavior
• Relative to a reference standard
• Reciprocity
• Positive reciprocity
• Rewarding kind behavior
• Negative reciprocity
• Punish unkind behavior
• Even at a cost

3
Evidence for Fairness

• Field evidence
• Collective action (strikes, consumer protest,
  voting)
• Tax compliance (people pay more than is optimal
  given they are rational and selfish)
• Donations
• Questionnaire studies in labor market
• Bewley (1995, 1997)
• Agell and Lundborg (1995)
• Campbell and Kamlani (1997)

4
Experimental Evidence for Fairness

• Bargaining
• Equal offers in bargaining games
• Disadvantageous counter offers
• Ultimatum game rejections of positive offers
• Dictator game positive transfers
• Trust game, moonlighting game and gift exchange
  game (market)
• Positive and negative reciprocity
• Public Goods Games
• Cooperation higher than predicted by standard
  theory
• Conditional cooperation
• Punishment (in public goods games)
• Punishment of defectors

5
The ultimatum game (Güth, Schmittberger and
Schwarze, JEBO 1982)
6
Güth et al. results (JEBO 1982)

• Ultimatum game c DM 4 or DM 10, inexperienced
  subjects.
• All offers above DM 1
• Modal x 50 percent of pie (7 of 21 cases)
• Mean x 37 percent of pie
• A week later (experienced subjects)
• All except one offer above DM 1
• 2/21 offer an equal split
• Mean offer 32 percent of pie
• 5/21 of the offers are rejected
• Systematic deviation from standard prediction

7
Do higher stakes lead to more equilibrium play?

• Hoffman, McCabe, Smith (1999) UG with 10 and
  100
• Offers are not dependent on the size of the cake.
• Rejections up to 30
• Cameron (1995) UG in Indonesia 2.5, 20, 100
  (GDP/Person 670)
• The higher the stakes the more offers apporach
  50/50.
• Responders more willing to accept a given
  percentage offer at higher stakes.
• Without payments we see differences less
  generous offers and more rejections.

8
Source Cameron (1995)
9
Source Cameron (1995)
10
Ultimatum game results (1)
11
Ultimatum game results (2)
12
Does altruism explain the high first mover
offers?

• Forsythe et al. (GEB 1994) compare simple
  ultimatum games with dictator games. In the
  latter, the proposer proposes a division (1-x, x)
  of the bargaining cake, which is then
  implemented.
• Result 1
• In the dictator game the distribution of x shifts
  significantly towards x 0 relative to the
  ultimatum game if real money is at stake (modal
  offer is x 0).
• If only hypothetical questions are asked no such
  shift can be observed.
• Result 2
• Even with real pay there is a concentration of
  offers around equal split (see Fig. 4.4)
• Conclusion Some of the subjects seem to be
  motivated by altruism but the higher
  concentration of offers around the equal split in
  the ultimatum game suggests that behavior cannot
  be fully attributed to altruism.

13
(No Transcript)
14
Single blind vs. double blind (Hoffman, McCabe
and Smith (GEB 1995)

• Conjecture that experimenters exert a kind of
  social control merely by being able to observe
  subjects actions.
• They report that if it is ensured that subjects
  know that the experimenter cannot observe
  individual decisions approximately 70 percent of
  the subjects in the dictator game give nothing
  and almost no offers above 0.3 can be observed.
• Has probably no relevance for the ultimatum game,
  see below
• And does such an environment itself lead to some
  sort of experimenter effect?

15
Punishment versus AnonymityBolton and Zwick (GEB
1995)

• Comparison of one-period ultimatum games with and
  without subject-experimenter anonymity (but
  always subject-subject anonymity).
• Comparison of one-period ultimatum game with the
  impunity game which has the same move structure
  as the ultimatum game but the same incentive
  structure as the dictator game (for first mover).
  In the impunity game only subject-subject
  anonymity prevailed.
• Impunity Game Player 1 proposes a division (1-x,
  x)
• Player 2 accepts or rejects. In case of rejection
  player 2 gets nothing while player 1 still gets
  1-x.
• Punishment option is removed.

16
Impunity game (2)

• Punishment hypothesis First movers in the
  ultimatum game choose high offers because of
  the fear of rejection.
• Prediction Lower offers in the impunity game
  compared to the ultimatum game.
• Anonymity hypothesis First movers in the
  ultimatum game do not want to be judged by the
  experimenter to be greedy and selfish.
• Prediction With subject-experimenter anonymity
  there are significantly lower offers than without
  subject-experimenter anonymity in the ultimatum
  game.

17
Impunity game (3)

• Results Punishment confirmed - Anonymity
  rejected
• In the ultimatum game offers in the first five
  periods are slightly lower under anonymity, in
  the second five periods they are slightly higher.
  In general offers are similar to other non
  anonymous ultimatum games.
• In the impunity game 100 percent of all offers in
  the last five rounds are equilibrium offers.

18
Do subjects accept unfair offers? The best shot
game

• Best Shot Game (Harrison, Hirshleifer, JPE 1989)
• player 1 chooses contribution q1 for a public
  good
• player 2 chooses contribution q2 for a public
  good
• Total contribution to public good is max(q1,q2)
• Costs are linear in contribution
• Revenue is concave in contribution

19
Payoffs in the Best Shot Game
20
Predictions

• If q10, player 2 chooses q2 4. Payoffs (3.7,
  0.42)
• If q11, player 2 chooses q2 0. Payoffs (.18, 1)
  
• If q12, player 2 chooses q2 0. Payoffs (.31,
  1.95)
• If q13, player 2 chooses q2 0. Payoffs (.39,
  2.85)
• If q14, player 2 chooses q2 0. Payoffs (.42,
  3.7)
• Player 1 chooses q10
• (If player 2 answers with contributing 0 as well,
  both would earn zero).

21
Result

• Harrison, Hirshleifer play the game with private
  information about payoffs.
• Convergence to subgame perfect Nash Equilibrium
• Prasnikar Roth (QJE 1992)
• Best Shot game with public information
• They also find convergence to the SPNE
• This is surprising since it implies unequal
  payoffs
• Intuition Intentions (see discussion about
  fairness models)

22
(No Transcript)
23
(No Transcript)
24
Multi-Proposer-Ultimatum game (Prasnikar and
Roth QJE 1992)

• 9 proposers simultaneously make an offer between
  0 and 10 to one responder.
• Responder decides to accept or reject the best
  offer xb.
• In case of rejection all ten players get zero. In
  case of acceptance responder receives xb.
• The proposer whose proposal has been accepted
  receives 10 - xb. All others receive zero.
• Prediction (based on smallest offer unit 0.05)
  
• Responder accepts any xb gt 0.
• Any proposal strictly below 9.95 cannot be an
  equilibrium because by bidding up 5 cents, a
  proposer can increase his payoff.
• 9.95 and 10 are equilibrium proposals (many
  equilibria, e.g., all offer combinations if at
  least two bid 10)

25
(No Transcript)
26
Result

• High offers from the very beginning (mean offer
  8.9)
• Competition plays an important role right from
  the beginning
• Quick convergence to the equilibrium
• How can this result be reconciled with the fact
  that in bilateral ultimatum bargaining subjects
  refuse unfair payoff allocations? (see fairness
  models)

27
Trust game/Investment gameBerg, Dickhaut, McCabe
(1995)

• 2 player sequential game
• Both players are endowed with 10 points
• Player 1 can give 2 up to 6 points (investment).
• Each invested point is tripled
• Player 2 gets to know the investment and can give
  points back to 1 (but does not have to)
• Standard predication
• Player 2 gives nothing to 1 (independent of
  investment)
• Player 1 invests nothing

28
ResultsBerg, Dickhaut, McCabe (GEB, 1995)

• Players 1 do invest
• Players 2 give back points
• Investments of 5 and 10 benefit player 1
• On average players 1 are just compensated
• Evidence against the standard prediction

29
Moonlighting Game(Abbink et al. 2000, Falk et
al. 2000)

• 1. Stage
• Players receive an endowment of 12 points
• Player A chooses action a ? -6, -5, , 5, 6
• a ? 0 A gives B a points
• a lt 0 A takes ?a? points from B
• In case a ? 0 the experimenter triplicates a such
  that B receives 3a.
• If a lt 0 player A takes ?a? points from B and B
  loses ?a? points

30
Moonlighting Game (ii)

• 2. Stage
• B realizes a und chooses b ? -6, -5, , 17, 18
• b ? 0 is a reward for A
• b lt 0 is a punishment
• A reward transfers b points to A
• A punishment costs B ?b? points and reduces As
  income at 3?b?
• Standard Prediction
• b 0 for all a
• a -6

31
Results Falk et al. Testing theories of
Fairness, Intentions matter

• Reward of kind actions
• Punishment of unkind actions

32
Fairness in markets? The gift exchange game

• So far fairness and reciprocity only in bilateral
  or multilateral bargaining environments Also
  relevant in markets?
• The impact of reciprocity on the market outcome
  crucially depends on whether the market is
  complete or incomplete.
• Gift-exchange game (Fehr and Falk JPE 1999)
• Stage 1 Firms and workers conclude contracts.
  Wages are settled in a double auction market,
  with wage ? ?20, 120?
• There is an excess supply of workers (711)
• Unemployment benefit 20

33

• Stage 2 Workers who concluded a contract choose
  an increasingly costly effort, with effort ?
  ?0.1, 1?
• Payoffs
• Firms (120 wage)effort
• Workers wage cost of effort

34

• Control treatment Complete contract 
• Payoffs
• Firms 120 wage
• Workers wage 20
• Standard prediction in both treatments the same
• wage 20
• effort 0.1 in the incomplete contract market

35
Competitive Prediction
36
Reciprocity in Markets Wages
incomplete
complete
37
Underbidding Incomplete Market
38
Underbidding Complete Market
39
Reciprocity in Markets Wage-effort Relation
40
Markets Summary

• In the incomplete contract market, wages are on
  average substantially higher than predicted.
• Underbidding of workers is not accepted by firms.
• Firms pay voluntarily high wages, because there
  is a positive correlation between wages and
  efforts on average.
• When effort is exogenously fixed, wages converge
  towards the predicted equilibrium and firms take
  advantage of underbidding.
• There are many variants of this game (e.g., Fehr,
  Kirchsteiger and Riedl QJE 1993 (one sided
  auction), Gächter and Falk SJE 2002 (bilateral)
• Always the same main results

41
Gift-exchange in the field

• Armin Falk
• Econometrica 2006

42
Contributions of the paper

• Extending the research on reciprocity and social
  preferences to the field
• Social preferences important for many public
  economics applications
• Social preferences research almost exclusively
  confined to laboratory studies
• Better understanding of the motives behind
  charitable giving
• Warm glow vs. gift-exchange
• Amount of donated money is substantial in many
  nations
• US 70 percent of all households donate,
  exceeding 1 percent of GDP (Andreoni et al. 1996)

43
Why conducting a field experiment ?

• In contrast with traditional field studies, it is
  possible to create an exogenous variation in the
  variables of interest
• In contrast to lab experiments people take their
  actions in their natural environment
• Related literature
• List and Lucking-Reiley (JPE, 2002) importance
  of seed money and refund policies
• Frey and Meier (2005) donations to a social fund
  administered by the University of Zurich

44
The field experiment

• International charitable organization that helps
  children in need
• Active in 38 countries and engaged in long-term
  development projects as well as in short-term
  emergency projects
• Shortly before Christmas organization sends out
  roughly 10,000 solicitation letters in the area
  of Zurich
• 2001 mailing for homeless children in Dhaka
  (Bangladesh)

45
Three treatments

• No gift
• Solicitation letter
• Small gift
• Solicitation letter one postcard and envelope
• Large gift
• Solicitation letter four postcards and
  envelopes
• Letter gift from the children from Dhaka,
  which can be kept or given to others.
• Except for gifts and remark on gifts, everything
  was exactly the same across treatments.

46

• A sample postcard

47
Procedure

• Random allocation of donators to treatments
• All letters were sent out on December 5, 2001
• Donations routinely recorded by organization

48
Hypotheses

• Warm glow feelings of internal satisfaction that
  arise from helping people who are in need
  (Andreoni 1989, 1995).
• Willingness to donate unaffected by treatments
• Reciprocity we are obligated to the future
  repayment of favors, gifts, invitations, and the
  like (Cialdini 1992) support from numerous lab
  experiments (Falk/Fischbacher 1999, Fehr/Gächter
  2000)
• Donations no gift lt small gift lt large gift

49
Table 1 Donation frequencies in all treatment
conditions
50
Table 1 Donation frequencies in all treatment
conditions
51
Table 1 Donation frequencies in all treatment
conditions
52
Table 2 Treatment differences of donation
probability
53
Histogram of donations
54
The organizations perspective

• Total donations 92,655 CHF
• Hypothetical total donations if all receive
• No gift 74,472 CHF
• Large gift 120,248 CHF
• Cost of gifts ?2,000 CHF
• Actual net gain 92,655 74,472 2,000 16,183
  CHF (22 percent)
• Hyp. net gain 120,248 - 74,472 4,800 40,976
  CHF (55 percent)

55
Some final thoughts

• Intertemporal substitution?
• No.
• Does gift exchange work over and over again?
• We do not know.
• Does any gift do the job?
• Probably not.

56
Conclusions

• Field experiment
• Relevance of gift-exchange on top of warm glow
• Confirmation of relevance of reciprocity with
  field data
• Important for many public economics applications,
  e.g.,
• Tax evasion
• Design of welfare state
• Incentive schemes

57
What Are the Puzzles ?
58
Evidence Against Fairness (?)

• In many games the experimental outcome is not in
  contradiction to standard theory.
• Convergence to standard prediction in public
  goods games
• Unequal outcomes in complete, competitive markets
  (e.g. double auction or proposer competition).
• Very unequal outcome in best-shot game.

59
Best-shot Game (Mini Version) Harrison,
Hirshleifer (1989), Prasnikar/Roth (1992)
1

• Players 2 accept unequal outcome of (3.7, 0.42)
• Such distributions are rarely accepted in the
  ultimatum game.

dont
contribute
2
2
contribute
contribute
dont
dont
0.42
3.7
0
0.42
3.7
0.42
0
0.42
60
Understanding Fairness

• Predictive models of fair behavior
• Preference based
• How predictive models can be used
• They formalize intuitive ideas and make them
  testable.
• Detect and distinguish between features.
• Provide precise predictions for applications.
• Give framework for evolutionary models.
• Therefore, models
• should be applicable to any game.
• should have a constant parameter set.

61
The Models

• Outcome oriented models
• Fehr and Schmidt (1999) (FS)
• Bolton and Ockenfels (2000) (BO)
• Reciprocity models Rabin (1993)
• Falk and Fischbacher (2005) (FF)
• Dufwenberg and Kirchsteiger (2004) (DK)
• Charness and Rabin (2002) (CR)
• Levine (1999)

62
Outcome Oriented Models

• Ui Ui( pi, p-i)
• Utility depends on own and others payoffs
• How does Ui depend on p-i
• Share pi/Spj (BO)
• all differences pi-pj (FS)

63
Psychological foundations of inequality
aversion Loewenstein, Thompson Bazerman
(JPersSocPsych 1989)

• Measuring social utilities when people compare
  outcomes of allocations to self and others
  depending on
• The type of relationship (positive, negative,
  neutral)
• Dispute type (invention, lot, business)
• 21 positive outcomes 21 negative outcomes
• 42 outcomes three types of relationships 126
  judgments.
• Participants rated their satisfaction with the
  outcomes.
• Measure utility functions (aggregate and
  individual)U c B1SELF B2NEGDIFF
  B3NEGDIFF2 B4POSDIFF B5POSDIFF2

64
Utility of advantageous and disadvantageous
inequality
65
Preferences with linear inequality aversionFehr
Schmidt, QJE 1999.
Interpretation? Assumptions ai?bi ? 0,
bilt1 Negative inequality aversion is more
important than positive. Nobody destroys money to
reduce positive inequality.
66
Explaining the Ultimatum Game
67
Reciprocity Models (Rabin,DK,FF)

• The structure of reciprocity models
• Ui pi ri S kindness j to i pj
• What determines kindness
• Payoff that player i is supposed to receive
  compared to reference payoff
• Absolute reference (FF)
• Relative reference (DK)
• Intentions are incorporated into the theory by
  considering the alternatives.
• Levine model (and CR extended version)
  kindness depend on the type of the other player.

68
Kindness in the DK-model
Others payoff
Reference point
kindness
Range of payoffs
My payoff
69
Kindness in the FF-model

• Kindness term
• Outcome term weighted with intention factor
• Outcome term pi - pj
• Intention factor depends on alternatives
• An action is intentionally kind if the other
  player had any alternative to be less generous
  (give me less).
• An action is intentionally unkind if the other
  player had a reasonable alternative to be more
  generous.
• The model combines the equity standard of the
  outcome oriented models with an intention concept
  similar to the other reciprocity models.

70
Intention Factor for Negative Reciprocity
Others payoff
More generous and reasonable alternative
even less generous alternative
unreasonable alternative
My payoff
71
Intention Factor for Negative Reciprocity
Others payoff
Not intentionally unkind
My payoff
72
Intention Factor for Negative Reciprocity
Others payoff
even less generous alternative
Not intentionally unkind
My payoff
73
Intention Factor for Negative Reciprocity
Others payoff
More generous and reasonable alternative
even less generous alternative
Intentionally unkind
My payoff
74
Intention Factor for Negative Reciprocity
Others payoff
even less generous alternative
Not fully intentionally unkind
unreasonable alternative
My payoff
75
Intention Factor for Positive Reciprocity
Others payoff
Not intentionally kind
My payoff
76
Intention Factor for Positive Reciprocity
Others payoff
Not intentionally kind
More generous alternative
My payoff
77
Intention Factor for Positive Reciprocity
Others payoff
Less generous alternative
Intentionally kind
More generous alternative
My payoff
78
A Questionnaire on kindness (Falk and
Fischbacher, GEB 2006)

• Subjects get list of possible offers.
• Have to evaluate the kindness of these offers.
• Between 100 and 100.

79
What All Models Achieve

• In the ultimatum game, low offers are rejected.
• Therefore, in the ultimatum game the proposers
  make higher offers than the dictators in the
  dictator game.
• In the gift exchange game higher wages are
  rewarded with higher effort. In the investment
  game they predict positive reciprocity.
• In public goods games they predict conditional
  cooperation.
• What about competition? Models predict that here
  subjects accept more inequity
• Intuition Multi proposer competition assume two
  proposers (i, j) offer 0.5 and i refuses to offer
  more. By infinitesimally overbidding, j increases
  material payoff by positive amount (from 1/n1 to
  1) but reciprocity disutility changes only
  infinitesimally is refusal to offer more is not
  an effective tool to achieve a fair outcome, thus
  it is worthless to reject an since the proposer
  cannot be punished

80
Distinctive Feature of the Models

• Differences in preference for distributions.
• Relative importance of inequity aversion,
  efficiency and maximin preferences.
• Differences in reciprocal behavior
• Who is the relevant reference agent?
• individual or group
• Is all punishment driven by inequity aversion?
• difference reduction or retaliation
• What is the role of intentions?
• outcomes or intentions

81
Differences in Reciprocal Behavior
Falk/Fehr/Fischbacher (Ectra. 2005)

• Q1 Who is the relevant reference agent?
• individual or group
• Q2 Is all punishment driven by inequity
  aversion?
• difference reduction (inequity aversion) or
  retaliation (reciprocity)
• Q3 What is the role of intentions?
• outcomes or intentions

82
Q1 Who Is the Relevant Reference Agent?

• Three person one-shot public goods game with
  punishment opportunity
• 1st Stage public goods game
• Contribute 20 points (cooperate) or nothing
  (defect)
• Payoff
• 20 - own contribution
• 0.6 sum of all contributions
• 2nd stage Reduce the other player's payoff at a
  cost
• Punishing cooperators 1 point reduction costs .3
  points.
• Punishing defectors 1 point reduction costs .4
  points.
• It is cheaper to punish cooperators.

83
Q1 Prediction

• BO predict that if cooperators punish, they
  punish other cooperators.
• It is the cheapest way to reduce inequity because
  it reduces the average payoff of the other
  players most (inequity measured towards the whole
  group and not individually)
• The other theories predict that if cooperators
  punish, they punish defectors.
• Because fairness is evaluated for each other
  player separately, those are punished who
  deserve punishment. Either because they have a
  higher payoff (FS) or because they are unkind (DK
  and FF).

84
Q1 Experimental Result (N120)

• 61 percent of the subjects cooperate.
• From the cooperators
• 69 percent punish
• 69 percent punish defectors
• 7 percent punish cooperators
• From the defectors (39 percent)
• 49 percent punish, cooperators and defectors

85
Q2 Is All Punishment Driven by Inequity Aversion?

• One-to-one punishment
• Same three person one-shot public goods game with
  punishment opportunity, but
• 1 point reduction costs 1 points, i.e., there are
  higher costs of punishment
• Inequity aversion models predict no punishment
  because inequity cannot be reduced.

86
Q2 Is All Punishment Driven by Inequity Aversion?

• 51 percent cooperate
• of these cooperators 47 percent punish two
  defectors
• punishment behavior is incompatible with any
  equity model
• Defectors do not punish.

87
Q2 Is All Punishment Driven by Inequity Aversion?

• UG with constant relative share
• Rejection reduces payoffs to 10 percent
• Rejection cannot change the relative share
• Hence, BO predict no punishment
• The other theories predict rejections

P
x
y
R
R
a
r
a
r

8
.8
5
.5

2
.2
5
.5
88
Exp 2 Is All Punishment Driven by Inequity
Aversion?

• UG with constant difference
• Rejection reduces payoffs by 2 points
• Rejection cannot change payoff differences
• Hence, FS and BO predict no punishment
• DK and FF predict rejections
• 82 is unkind and triggers punishment. Punishing
  means a reduction of the other player's payoff.

89
Q2 Experimental Results (N48)

• Punishment does not only occur to reduce
  inequity. Even if inequity cannot be reduced,
  people punish to reciprocate unkindness (20
  percent).

90
Q3 Are Intentions Important? Four Mini Ultimatum
Games
This is like the best shot game
91
Intentions (ii)
92
Predictions of the rejection rates of the 82
offer

• BO and FS predict the same rejection rate for
  both alternatives.
• These theories model fairness in a
  consequentialist way and the consequence of the
  82 offer is always the same.
• DK predict zero rejection rate for the
  alternative 100.
• Subjects do not consider the 82 offer as unkind
  because 100 is even more unkind.
• FF predict positive rejection rates in both
  cases. The rejection rate is higher in the 55
  case.
• Fairness is determined by the outcome and the
  intention of the other subject.

93
Experimental results (N45)Falk, Fehr and
Fischbacher, Economic Inquiry forthcoming)
94
Q3 Proposer Behavior

• Proposer behavior is compatible with selfishness,
  but also with preferences for fairness.

95
Intentions, once more

• Moonlighting game as before (Falk, Fehr and
  Fischbacher, GEB forthcoming, see above).
• But Player As decision is randomly determined
  and players B know that.
• Random mechanisms is based on a human choice
  distribution. Controls for the equality of
  choice probabilities across computer generated
  and and human generated first-mover action.

96
Rewards and punishments with and without
intentions

• The same consequences trigger very different
  behavior.
• Questions consequentialistic notions of fairness.
• Casts doubt on the consequentialistic practice in
  economics to define the utility of an action
  solely in terms of the consequences.

97
Intentions and Random Move Games

• If the move of the first player in the ultimatum
  game is made by a random device, then
• An unfair outcome is not intended by player 1.
• Therefore, unfair offers are less likely to be
  rejected.
• (Blount 1995)
• Same idea in gift exchange game (Charness)
• Here, high wages are rewarded with similar effort
  in the treatment in which a person chooses the
  wage compared to the treatment in which the wage
  is randomly drawn. (though steeper slope)
• In both experiments reward and punishment also
  occur in the random move treatments.

98
Choice plus randomness

• In Charness/Levine (2005) firms either choose
  high or low wage
• Workers respond with low (punish), medium or high
  (reward) effort
• Idea chance move changes outcomes after first
  stage
• Most interesting combination in treatment 2

99
Conclusion

• Fairness can be captured by incorporating
  preferences for fairness into the utility
  function.
• Models reconcile results with equal outcomes as
  well as with unequal outcomes (e.g. UG vs.
  competitive markets).
• Fairness is evaluated individually.
• Inequity reduction is not the only reason for
  punishment.
• Intentions and outcome matter.
• Reciprocity models give a better description of
  human behavior but at a cost in tractability.

100
Trust and Discrimination A Citywide Experiment

• Armin Falk
• University of Bonn
• Christian Zehnder
• University of Zurich

101
Trust is important

• Economic importance of trust derives from the
  fact that it enhances efficiency in the presence
  of limited contract enforcement
• Virtually every commercial transaction has
  within itself an element of trust . (Arrow,
  1972, p.357)
• Trust is part of a groups social capital
• Related to level of foreign investments, growth
  and efficiency of institutions (Guiso/Sapienza/Zin
  gales (2006), Knack/Keefer (1997), La
  Porta/Silanes/Schleifer/Vishny (1997))
• Experimental literature on economic relevance of
  trust
• But little is known with respect to trust and
  discrimination

102
Research questions

• Discrimination Do people trust strangers from
  different groups differently?
• Is trust discrimination taste driven or
  statistical discrimination?
• In-group favoritism, i.e., do people trust
  strangers from their own group more than
  strangers of other groups?
• Individual determinants of discrimination,
  in-group favoritism, trust and trustworthiness?

103
Our study

• The city as a laboratory
• We study discrimination using districts of a city
  as groups
• Districts are natural groups, have a social
  meaning and are sufficiently heterogeneous
• District affiliation is relevant for economic
  transactions
• Social dynamics of cities and city development

104
Related literature

• Discrimination
• Fershtman/Gneezy (2001)
• Bouckaert/Dhaene (2004), Haile et al. (2006), and
  others
• In-group favoritism
• Bernhard et al. (2006)
• Goette et al. (2005)
• Focus on ethnic discrimination
• No individual determinants
• No representative subject pool

105
Design of the field experiment

• One-shot, sequential, two-player trust game
  experiment
• (Berg, Dickhaut and McCabe 1997)
• First and second mover receive endowment of CHF
  20
• First mover
• Investment of CHF 0, 2, 4, , up to 20
• ? Money is tripled and transferred to the second
  mover
• Second mover
• Back transfer of CHF 0, 1, , up to CHF 80

106
Design (2) Payoffs

• First mover
• 20 investment back transfer
• Second mover
• 20 3 x investment back transfer

107
Design (3) District specific social capital

• Each first mover made 12 decisions (plus
  beliefs), one for each of the 12 districts of
  Zurich

Provides information of in-group favoritism and
discrimination on an individual level
108
Zurich and its districts map shown to subjects
109
Design (4) Strategy method

• Second movers responses elicited with help of
  strategy method
• Pay back decision for each possible investment
• Allows
• Specification of types
• Individual determinants of trustworthiness

110
Design (5) Procedures

• 4000 letters sent out
• 984 subjects took part in the study
• Addresses delivered by the statistical office of
  Zurich
• Representative with respect to age, income and
  foreigner status per district
• Ss received letter, instructions, questionnaire
  credible procedure
• Matched and paid out in cash (letters) according
  to their decisions
• Mean earnings per subject about CHF 33.2
• Total spending about CHF 32,600
• We deleted addresses after payment
• Additional newspaper study (see below)

111
Questionnaire plus information from statistical
office

• Socioeconomics
• Gender, age, yearly taxable income, marital
  status, profession, foreigner, number of
  siblings, religion, skill level (isco), district
• Social capital measures and district/city
  specific information
• Trust question (GSS, WVS)
• Club membership (how many)
• Most people are fair or selfish
• Most people helpful or self-serving
• Most people are reliable (yes/no)
• Help provided by unknown in district
• How many friends in district
• Connected to district and city
• How many years in district and city
• How often private phone calls
• Save when walking alone in district at night
• Other
• Political orientation
• Happiness

112
Results

• Discrimination
• Level and distribution of investments
• Do investments differ significantly across
  districts?
• Determinants
• Reciprocation
• Level and distribution
• Are investments and back transfers correlated?
  Statistical discrimination
• In-group favoritism
• Prevalence
• Individual determinants
• Individual determinants of trust and reciprocity

113
Level and distribution of investments

• Mean 13.16 Std. dev 7.07 Median 16

114
District specific discrimination

• Districts differ significantly regression of
  investments on district dummies F-test prob. gt
  F 0.000
• Variation investment 11 percent, coincides with
  expected trustworthiness (Spearmans Rho 0.88,
  plt0.001)

115
Pair-wise comparison of investments in districts
p-values of pair-wise Wilcoxon-signed rank tests
116
Investment ranks are correlated across districts
Consensus among districts about which are the
good and bad districts
Spearmans Rhos, excluding own district Out of 66
correlations, 61 are positive
117
Additional data from a newspaper experiment how
systematic is discrimination?

• Widely read newspaper of Zurich (Tagesanzeiger),
  two articles
• First reporting the idea of the study, not
  showing results
• Invitation to a quiz Which are the two districts
  that received
• the lowest investments?
• the highest investments?
• Among those readers whose answers are correct,
  three are randomly chosen to receive CHF 300
• Second a week later, results of the study were
  reported, together with interviews (president/me)
• 281 readers of the Tagesanzeiger took part

118
Results from the newspaper readers
Corr inv/high Spearman's rho 0.8932, p
0.0000 Corr inv/low Spearman's rho -0.8807, p
0.0002
119
Determinants of trust (discrimination)

• Economic and social status
• Income (?), education and wealth (?) (Alesina and
  Ferrara (2000) Knack and Keefer (1997))
• Heterogeneity
• Ethnic heterogeneity (?) (Costa and Kahn (2002)
  Alesina and Ferrara (2000))
• Religious heterogeneity (?) (Costa and Kahn
  (2002) Alesina and Ferrara (2000))
• Mobility (?) (Glaeser et al. (2001) DiPasquale
  and Glaeser (1998) Alesina and Ferrara (2000))
• Home ownership (?) DiPasquale/Glaeser (1998)
• Political participation (?) Feld/Frey (2000)
• Hierarchical religion (?) (Putnam 2000)
• Own district, in-group bias

120
Descriptive statistics of districts

• Notes Statistical Office of Zurich and
  Statistical Yearbook of the City of Zurich 2003
• Income median per capita income in 1000 Swiss
  Francs
• High education population fraction with at least
  a college degree
• Foreigners population fraction of foreigners
• Religious heterogeneity fragmentation index 1
  - sum of squared population fractions of all
  religions
• Years of residency average number of years with
  residency in the same district per person
• Home ownership fraction of apartments owned by
  inhabitants
• Catholics fraction of Catholics

121
Income
Spearman's rho 0.9161 p 0.0000
122
(No Transcript)
123
2. Back transfers

• How trustworthy are people from Zurich?
• Are there differences across districts?
• Is discrimination statistical or taste driven?
• Correlation between trust and trustworthiness on
  the district level

124
Investments and (expected) back transfers
125
Distribution of individual mean reciprocal
inclination
126
Back transfers

• Back transfers differ significantly across
  districts regression on district dummies prob.
  gt F 0.0206

127
Comparison of ranks for investments and back
transfers statistical discrimination
Spearman's rho 0.6853 p 0.0139
128
Trustworthiness investments and mean reciprocal
inclination by district
Spearman's rho 0.6713 p 0.0168
129
3. In-group Favoritism

• Do people favor strangers from their own
  district?
• Is that driven by taste or the expectation of
  higher trustworthiness?
• Individual characteristics of in-group favoritism

130
In-group favoritism

• Out of the 12 districts, 11 invest higher amounts
  to their own district than they invest on average
  into the other districts

0.003
131
4. Individual determinants of trust (investments)
and trustworthiness (back transfers)

• Largely unknown, since existing evidence almost
  exclusively confined to lab experiments with
  homogenous subject pool
• Bellemare and Kröger (2004) use a Dutch sample
• Fehr et al. (2003) combine experiment with survey
  (441 Ss in Germany)
• Carpenter et al. (2003) cooperation experiments
  in Asian urban slums
• Gächter et al. (2004) cooperation experiments
  with non-student subjects in urban and rural
  Russia
• Dohmen/Falk/Huffman/Sunde (2006) questionnaire
  data, using the GSOEP

132
(No Transcript)
133
Behavioral vs. questionnaire measures of social
capital

• Glaeser et al. (2000), Lazzarini et al. (2003)
• Harvard students, U. of Sao Paolo students
• No relation between general trust question and
  first mover behavior in a trust experiment
• Second movers behavior is correlated
• Similar Burks et al. (2000), Ben-Ner/Putterman
  (1999)

134
VariablesGSS/The World Bank Integrated
Questionnaire for the measurement of Social
Capital/Glaeser et al. (2000)

• trust
• Generally speaking would you say that most people
  can be trusted or that you cant be too careful
  in dealing with people?
• reliable
• Today one cannot rely on strangers anymore. (1 to
  4, binary code)
• fair
• Do you think that most people would exploit you
  or that they would try to be fair to you?
• egoist
• Do you think that most people most of the time
  try to be helpful or only follow their own
  interest?
• help
• If you need help, do you think that a stranger
  from your district would help you? (yes, no)
• phone calls
• How often have you made a private phone call last
  week? (approximate number)
• unsafe
• How safe from crime and violence do you feel when
  walking in your district alone after dark (four
  levels, binary coded)
• club memberships
• In how many clubs are you a member? (number)

135
(No Transcript)
136
(No Transcript)
137
(No Transcript)
138
(No Transcript)
139
Why differences to previous findings?

• More observations
• Non student subject pool
• Responses immediately after decisions

140
Concluding remarks

• Behavioral measure to study trust discrimination
  Significant discrimination
• Determinants economic and social status, ethnic
  heterogeneity
• Differences endogenously reinforced
  segmentation?
• Higher income ? higher social capital ? higher
  income
• If policies want to counteract this segmentation
  they must be able to diagnose the reputation
  differences field experiment
• Trust and trustworthiness correlated statistical
  discrimination
• Strong evidence for in-group favoritism
• Strong correlation of behavioral and survey
  measures of trust
• Important individual heterogeneity and
  determinants regarding trust and trustworthiness,
  discrimination and in-group favoritism
• Age, education, family status, political
  orientation, religion, foreigner

141
Success and Prevalence of Homo ReciprocansIZA
DP 2205

• Thomas Dohmen, Armin Falk, David Huffman and Uwe
  Sunde
• IZA and University of Bonn

142
Questions

• How prevalent is Homo Reciprocans?
• Individual determinants of reciprocity?
• Correlation of positive and negative reciprocity?
• Consequences of reciprocity for economic and
  social outcomes (wages, subjective well-being,
  friendships)?
• Answering these questions requires
• Leaving the lab, and using a large,
  representative samples
• Combine measures of reciprocity with demographic
  variables

143
Data

• Large sample SOEP (2005 wave)
• About 22,000 individuals (age 17), 12,000
  households
• Representative of the population
• Each adult household member is interviewed
• Extensive socio-demographic information
• Individual characteristics
• Family and household background
• First time using questionnaire for large sample,
  complements experiments with non-student
  populations
• Bellemare and Kröger (2004) use a Dutch sample
• Fehr et al. (2003) combine experiment with survey
• Carpenter et al. (2003) cooperation experiments
  in urban slums
• Gächter et al. (2004) cooperation experiments
  with non-student subjects in urban and rural
  Russia
• Falk and Zehnder (2006) Citizens of Zurich

144
Reciprocity measures

• Positive reciprocity
• If someone does me a favor, I am prepared to
  return it
• I go out of my way to help somebody who has been
  kind to me before
• I am ready to undergo personal costs to help
  somebody who helped me before
• Negative reciprocity
• If I suffer a serious wrong, I will take revenge
  as soon as possible, no matter what the cost
• If somebody puts me in a difficult position, I
  will do the same to him/her
• If somebody insults me, I will insult him/her
  back
• 7-point scales
• 1 means does not apply to me at all 7 means
  applies to me perfectly
• Two questions ask explicitly whether the
  respondent would incur costs in order o be
  reciprocal
• 20,774 individuals responded to all six
  reciprocity measures

145
Positive Reciprocity Measures
Negative Reciprocity Measures
146
Distributions of Positive and Negative Reciprocity
147
Is positive and negative reciprocity correlated
within subject?

• Only weakly
• Within-person correlation between positive and
  negative reciprocity 0.024
• Suggests that positive and negative reciprocity
  are distinctive behavioral concepts
• World more complicated than just selfish vs.
  reciprocal
• Different emotional responses?
• Anger appears to be important for explaining
  punishment behavior in experiments (Fehr and
  Gächter, 2002)
• Candidates for positive reciprocity include
  gratitude, or possibly anticipated guilt
  associated with not rewarding
• Different determinants?

148
Determinants of Reciprocity
0.01
Other controls marital status, number of
children, religion, region, student, occupation,
health status, month of interview.
149
Determinants of Reciprocity
0.01
Other controls marital status, number of
children, religion, region, student, occupation,
health status, month of interview.
150
Determinants of Reciprocity
0.01
Other controls marital status, number of
children, religion, region, student, occupation,
health status, month of interview.
151
What about students?

• Lab experiments are often criticized
• Selected student sample
• Do we overestimate the importance of social
  preferences when using students?
• In the GSOEP sample students are significantly
  less reciprocal than non-students (both positive
  and negative reciprocity)
• Field experiment (Falk and Zehnder 2007)
  Investment game, strategy method
• Students are significantly less reciprocal, no
  difference in trust

152
Consequences of reciprocity

• Effort in employment relationships
• Experiments and questionnaires positive relation
  between wages and effort on the job, even when
  not enforceable
• Hypothesized to reflect positive reciprocity
  among workers
• Fehr, Kirchsteiger and Riedl (1993), Fehr and
  Falk (1998), Bewley (1999), Agell and Lundborg
  (1997) etc.
• Do we find evidence from the field that Homo
  Reciprocans works harder?
• Three measures of effort
• Total weekly hours, overtime hours
• Absenteeism

153
Reciprocity and Work Effort
154
Reciprocity and Work Effort
155
Consequences of reciprocity

• Sustaining employment relationships
• Experimental evidence suggests positive
  reciprocity helps sustain long-term relationships
• Brown, Falk, and Fehr (2004) on endogenous
  relations
• Experimental and field evidence suggests workers
  retaliate against employers in response to unfair
  treatment
• Bewley (1999) Mas and Kreuger (2004)
• Does positive reciprocity increase ability to
  stay employed?
• Does negative reciprocity make unemployment more
  likely?
• Firms fire workers to prevent retaliation
• Workers quit as a form of retaliation
• More unemployment among negatively reciprocal
  workers

156
Reciprocity and Unemployment
Additional controls are marital status, number of
children in the household, and religious
background.
157
Success of Homo Reciprocans

• Strategic disadvantage?
• Wasting resources to engage in costly rewards and
  sanctioning
• Strategic advantage?
• Credible signal of reward or retaliation leads to
  better treatment
• Social competence friendships and networks
• Three measures of success
• Number of close friends (SOEP question)
• Income (Mincerian wage regression)
• Happiness (SOEP question)
• Important goal of human life and summarizing
  success and achievement in a general way (see
  Frey and Stutzer 2002)

158
Success of Homo Reciprocans
159
Success of Homo Reciprocans
160
Success of Homo Reciprocans
161
Summary

• Reciprocity rule rather than exception
• Heterogeneity and surprisingly weak correlation
  between positive and negative reciprocity
• Suggests that positive and negative reciprocity
  are distinctive behavioral concepts
• Corroborated by asymmetry in determinants
  (gender, age and height)
• Positive relation between positive reciprocity
  and higher work effort
• Positively reciprocal people report having more
  close friends and a higher overall level of life
  satisfaction.
• In this sense, Homo Reciprocans - in the positive
  domain - is in fact more successful than his or
  her non-reciprocal fellows