6. Self Interest Hypothesis Flashcards
What did Stigler say about self interest and ethical values
“When self interest and ethical values with wide verbal allegiance are in conflict, most of the time self interest theory will win
Self interest hypothesis
People only maximise their own well being
General results of ultimatum game and results from which study are we shown?
-most people offer 50%
-acceptance rate increases as offer increases up until pretty much 100% acceptance when offered 50%.
-slight drop off in acceptance rate for offers over 50%
-Guth Et al 1982 newspaper experiment n=1035. Replicated by Gachter at Nottingham with economic students
What does the strategic fairness paper by Forsythe 1994 do and what do they find?
They compare offers in UG and DG. Offers are much lower in DG so there is a strategic incentive in the ultimatum game. Average offer UG=4.4, Avg offer DG=2.6
Give summary of results in Engel Et al 2011
Meta analysis of 129 studies on dictator games
36% give nothing
16.7% give half
5% give everything
All others are decreasing in likelihood as offer increases
2/3 don’t act in line with self interest hypothesis
Results of Berg Et al 1995
Trust game. People send money and the responder sends money back. This isn’t in line with self interested hypothesis
How does the gift exchange game work?
2 players: principal and agent
Stage 1: principal offers w>0
Stage 2: in case of acceptance, agent chooses effort e
Monetary payoffs
Principal =10e -w
Agent = w-c(e)
Results of public good game (Ledyard 1995)
People cooperate in one shot games but some free ride. Cooperation declined to low levels over time. Partners contribute more than strangers- strategic incentives matter
Results of public goods game with punishment (Fehr & Gaechter 2000)
With punishment, cooperation increases over time in both partners and strangers. Punishment levels in both partner and strangers are similar which indicates punishment is likely emotional rather than strategic
What are the two main theoretical approaches?
Outcome based models of social preferences: individuals care about own payoff and may also care about distribution of material gains between themselves and others
Models of reciprocity: individuals care about their own payoff and may also care about the intentions of others
What assumptions are needed for the Fehr- Schmidt model 1999?
Inequality per se or interpersonal inequity
Relevant reference group?
Relevant reference output?
Advantages vs disadvantaged payoff inequality?
What is self centered equity aversion?
Individuals only care about inequity between own and others payoff
What factors are likely to affect the relevant reference group?
Social context, social proximity among individuals etc.
In lab it is natural to assume the reference group is simply the set of subjects playing against each other
What did Loewenstein 1989 find?
He gave a utility function of player i where alpha measured the aversion against disadvantageous inequity and beta measured the aversion against advantageous inequity
We assume alpha> beta> 0
Beta<1
Which stylised facts about behaviour in the UG experiments died the FS model account for?
Proposers don’t make offers over 50%
Offers of 50% are always accepted
Low offers are likely to be rejected
Set up of Blanco, Engelmann, Norman 2011 in experiment to derive the joint distribution of alpha and beta
Subjects play all of UG, MDG, SPD, PGG. Subjects paid according to results in one randomly chosen game
Findings of Blanco, Engelmann, Norman 2011 in experiment to when finding the joint distribution of alpha and beta
-inequality aversion has predictive power at aggregate level but performs less well at individual level.
-while subjects might follow some notion of fairness across various games, this isn’t systematically captured by the inequality aversion parameters at the individual level
Achievements of the Fehr-Schmidt model
-FS explain several key phenomena observed in experiments
-FS is simple and tractable
-popular and hugely cited
Limits of inequity aversion
-Some model assumptions are driven by focus on individual behaviour in simply economic experiments.
-assumptions are less clear in other contexts or in field
-however, these caveats hold for many models that abstract from important real life features
Theoretically important limitations of the FS model
- Dictator game predictions and linearity assumption in the FS model.
- Cross game consistency of the FS model
- Equity vs efficiency/rawlsian- maximin preferences
- The role of intentions
How do Blanco, Engelmann, Normann estimate subjects alpha and beta?
UG used to estimate subjects alpha, MDG used to estimate subjects beta
Do inequality minimising actions arise because subjects care about inequity or social welfare?
Engelmann & Strobel 2004 use simple distributional games to disentangle inequity aversion, efficiency and maximin motives. One player decides allocation of money among herself and two other players. All 3 motives have seemingly similar levels
How is intentionality tested? And how does it matter?
Compare how human players respond to offers made by computers vs humans. People accept everything when offered it by a computer.
Falk, Fehr, Fischbacher 2003 test role of available yet unchosen alternatives focusing on rejection rate after 8/2 offer
How can reciprocity appear in Battle of sexes Rabin 1993
P1 might conclude that P2 is lowering her own payoff to hurt him so P1 may retaliate. P1’s payoff depends not only on actions but also on beliefs about P2’s motives
What are the three components in the utility function? Rabin 1993
-Utility from material payoff
-Utility from perceived kindness/unkindness
-utility from reciprocity incentive to match (un)kindness with (un)kindness
What is a fairness equilibrium equivalent to?
Fairness equilibrium is equivalent to a NE in a psychological game (mutual BR’s) with the additional requirement of beliefs consistent with equilibrium play
In the battle of sexes what are the fairness equilibria
(O,O) and (B,B) because they are strict NE in material payoffs. (O,B) can be a fairness equilibrium as long as x is small
What constitutes psychological game theory?
When beliefs about what other people will do matter
Psychological nash equilibrium
NE plus requirement that all beliefs match actual behaviour
What is reciprocity
The desire to get even- respond to wrongdoings with revenge and kindness with kindness
What does perceived kindness depend on?
Beliefs about other player’s choices
What is bij? Where ij are subscript
It denotes i’s belief about j’s choice. First order belief
What is Ciji?
Denotes i’s belief about bji. Second order belief
Guilt aversion
Player I suffers from guilt to the extent that he puts less into public goods than average of what he believes his two co-players believe he put into the public account
Describe the Battigalli & Dufwenberg 2007 experimental set up
“Welcome to the (community) experiment”. Treatment of public good already existing vs public good needing to be provided. One shot. Elicit first and second order beliefs after give/take choice
Results of Battigalli & Dufwenberg 2007
Results show that first and second order beliefs are impacted by labelling and valence. Results support reciprocity: contributions and first order beliefs are positively correlated. Results support guilt aversion: contributions and second order beliefs are positively correlated
How do most models view honesty?
They assume that people don’t tell the truth if not incentivised to do so
Describe the set up of seminal experiment on honesty Fischbacher & Felhi- Heusi 2013
Die in a cup experiment. People given a die and a cup in a private lab space and told they will be paid according to reported two rolls of the die where 1=£1, 5=£5, 6=£0
Results of Fischbacher & Felhi-Heusi 2013
0=6.4%
1=7.2%
2=11.6%
3=12.6%
4=27.2%
5=35.0%
People aren’t honest but they are more honest than the standard model predicts
Findings of Abeler, Nosenzo, Raymond 2019
They document patterns in 90 studies of die in a cup type tests. Only 23% of potential monetary gain is realised. In each treatment, more than one state is reported with positive probability. Stakes don’t appear to really matter
Lying costs
Lying causes direct disutility e.g because of injunctive norms or moral rules
Social norms/comparisons
Reporting behaviour informed by comparison with others’ behaviour (including descriptive norms, social preferences)
Reputation for honesty
People care about an observers beliefs about self e.g want to be perceived as honest by others
Drawing in
Lowest state even more underreported when higher state becomes more likely
Drawing out
Lowest state becomes less underreported when higher state becomes more likely
Affinity
Reporting of Wh increases if subject believes Wh is more likely to be reported by others
Aversion
Reporting of Wh decreases if subject believes Wh is more likely to be reported by others
How does changing the observability of states affect results in experiment reporting 4s and 10s?
Subjects told to take out the number of coins that corresponds to their draw out of an envelope. Those who aren’t observed are more likely to lie. Evidence for o-shift when the true state is observable, subjects never lie downwards. Only combining a desire to be and appear honest can explain the data
What inferences can we draw from results of Blanco, Engelmann, Normann 2011?
Suggests individuals don’t maximise a FS utility function consistently across domains. Either they don’t really have FS preferences, or if they do, parameters are sensitive to context
Results of Fehr, Falk, Fischbacher 2003 with role of intentions
-people accept everything when it is randomly chosen by the computer
-Focus on rejection rate after 8/2 offer
-FS predicts that rejection rate will always be the same
-in reality rejection rates are 5/5=45%, 2/8=26%, 8/2=18%, 10/0=11%
What is a consequentialist model?
A model that only looks at outcomes. E.g. Fehr smchidt model
What is pi(j)(bj,aj) in the reciprocity model?
It is the actual payoff I gives to j given that I believes that j is choosing bj
What is pi(j)^e in the reciprocity model?
It is the equitable payoff for j given bj; defined as the average between the highest and lowest payoffs I can give to j, excluding Pareto dominated payoffs
What is pi(i)(ci,bj) in the reciprocity model?
It is the actual payoff I believes j gives to I given that I believes j believes that I plays ci