Theories of Social Preferences I

1
Theories of Social Preferences I
2
Theories of Social Preferences - Background

• Certain types of behavior (e.g. charitable
  giving) are difficult to explain with standard
  economic theory
• With the emergence of experimental economics this
  became particularly salient
• Why do people depart from equilibrium even in the
  simplest environments?
• 90s and first half of this decade
• Development of theories of social preferences
• Experiments that pit them against each other
• Experiments that test the predictive power of the
  theories

3
Warm-up

• Give some examples of stylized facts from
  experimental economics which are difficult to
  explain with standard theory.
• Some researchers have argued that subjects are
  initially inexperienced and must first learn
  before equilibrium is reached (learning
  hypothesis). What do you think?

4
Theories of social preferences - Models

• Theories of inequality aversion Subjects dont
  like inequality relative to relevant comparison
  groups.
• Fehr Schmidt, QJE 1999
• Bolton Ockenfels, AER 2000.

• Theories of reciprocity Intentions (Kindness
  or Unkindness) are explicitly modeled.
• Rabin, AER 1993.
• Falk Fischbacher, GEB 2006.
• Dufwenberg Kirchsteiger, GEB 2004.

All fairness models modify the utility function
All models retain the rationality assumption.
5
Preferences with linear inequality aversionFehr
Schmidt, QJE 1999.

• Remember the Fehr-Schmidt utility function (two
  players)
• and its graphical representation

6
Explaining behavior in the UG

• One of the puzzles Fehr and Schmidt set out to
  explain was observed in the Ultimatum Game
• rejection of positive amounts in binary game
• behavior in market version in line with standard
  theory
• Just to check how confusing I was in the lecture
  lets go over the FS equilbrium in the ultimatum
  game again...

7
Explaining the UG responder behavior

• When will responder accept an offer?
• Offers above 50 always accepted (why?)
• An offer s below 50 will be accepted if

8
Explaining the UG proposer behavior
9
Explaining the UG proposer behavior (2)
10
Market versions of the UG proposer competition

• Responder accepts any sgt0.5
• Assume that player i makes highest offer
  0.5ltshighlt1
• If player j offers slightly more three
  advantages
• Less disadvantageous inequality relative to
  responder
• Less disadvantageous inequality relative to
  proposer i
• Monetary gain
• In equilibrium at least two proposers offer s1
  and responder accepts (exists for any
  distribution of parameters)

11
Market versions of the UG responder competition

• In UG the responder can control the allocation of
  the pie, by threatening to reject the offer.
• In a market game this is only possible, if also
  competitors decline.
• A single egoistic competitor is enough, to detain
  the majority of fair types of declining unfair
  offers.
• If ß lt (n - 1)/n (for the proposer) there exists
  an equilibrium in which all responders accept any
  s 0 and the proposer offers s 0.
• Solutions of a puzzle Why can there be unfair
  results in competitive games given that people
  care about fairness?
• Answer With competition, direct punishment is
  not possible.
• Conclusion Fairness models are not in
  contradiction to unfair results in experiments.
  One model can explain fair and unfair results.

12
Warm-up

• So what, some critics say, you can explain
  everything by modifying the utility function.
• What do you think?

13
Taking Fehr-Schmidt to the data

• Predictive models of fair behavior
• formalize intuitive ideas and make them testable.
• ? Detect and distinguish between features.
• Provide precise predictions for applications.
• ? Therefore, models
• should be applicable to any game.
• should have a constant parameter set.

14
Taking Fehr-Schmidt to the data

• To make predictions across games we need to find
  out more about the parameters a and ß. How?
• Fehr and Schmidt take data from the UG to find a
  distribution of as and ßs that is consistent
  with observed behavior.
• Given a distribution of as for the responders
  one can also find a distribution of ßs of the
  proposers that is consistent with the data.

15
Taking Fehr-Schmidt to the data

• Fehr/Schmidt go on to show that these
  distributions are also consistent with behavior
  in several other games.

Table taken from Fehr/Schmidt, QJE 1999
16
Warm-up

• Avner Shaked wrote a paper entitled The
  Rhethoric of Inequity Aversion in which he
  fiercely critizes Fehr and Scmidt, among other
  things for their calibration of a and ß. Where do
  you see potential problems with their approach?

17
Taking Fehr-Schmidt to the data
18
Taking Fehr-Schmidt to the data

• Distributions cannot be determined uniquely,
  large set of distributions consistent which are
  compatible with the data
• Shaked accuses Fehr/Schmidt of picking the one
  that best fits their purpose (and not mentioning
  it)

19
Taking Fehr-Schmidt to the data

• Approach only allows for a test at an aggregate
  level.
• Data provide no information about joint
  distribution of a and ß since they are taken from
  different groups.
• How can we test within subject predictions?
• Blanco et al (WP) play different games with same
  subjects
• FS performs well at aggregate level, but less
  well at individual level
• Dohmen et al (forthcoming, EL)

20
Warm-up

• How many friends does it take to moderate Avner
  Shakeds outraged tone?