11 - public goods games (part 2) Flashcards
what are issues with creating experimental games in lab
- error - games might be too complex - subject doesnt understand - subjects might think others wont understand
what is NE
each players strategy maximises their own payoff, given the strategies of the other players
what are the 2 different interpretations of NE
hold unconditionally - subjects play NE perfectly
conditionally on completion of convergence process that players learn - NE is conditional on completing learning process
what are the 3 issues with public good experimental games
- that means wont find real answer
is game theory correct - do people maximise based on what others do?
reasons why NE wont be reached?
- error
- equilibrium - is it conditional or unconditional
- control of preferences
what is the issue with controlling preferences
game theory = maximise utility
experiment = assumes utility is determined by money payoff
but what if players care about other things, money isnt the only thing that captures utility
- subjects will be playing a different game to the one experiments intended
what is the method used to implement a public good in the lab
VCM
voluntary contributions mechanismh
what is a public good
non-rival = one persons consumption of it will not reduce another persons ability to consume it
non- excludable = everybody has access to it
what is a voluntary contributions mechanism
- play game in group
- money payoffs are proportional to points earned
- each person endowed with E token - choose independently and simultaenously what to do with them
- can divide E between private account and public account
- each token in private account gives 1 point
- each token in public earns m points for everyone - m<1
- everyone benefits from public good
what is the dominant strategy in VCM
if player wants to maximise own payoff (points)
- m<1
- so will contribute nothing to public account
- all tokens in private account + whatever anyone else put in public
- if everyone doesnt contribute –> everyone gets E
what would happen if everyone contributes
- mnE points > E points (not contributing)
- everyone is better off if they all contribute everything
- if they cooperated
what are the typical findings of VCM if game is played once
- 20% free riders - dont contribute
- average contribution of endowments - 40-60%
what are typical findings of VCM if game is repeated
and you get feedback on what other members in group contributed
- contribution starts off at 50% of endowment
- decays over next rounds
- contribution rates fall
what 3 reasons explain why players contribute
- when dominant strategy is not to
(has to account for initial contribution and decay)
- error
- learning
- strategic
- preference explanations
why players contribute
error
initially they are confused about the game
- as they learn
contributions decay
why players contribute
strategic
players start by contributing in early rounds because they think it will raise future contribution of others - so they will benefit from this
- want to benefit off of the fact that others are confused (why would they think that others would contribute)
- contributions decay in later rounds to benefit from the cooperation of others
why players contribute
preference explanations
altruism
conditional cooperation
self satisfaction
Keser (1996)
what
wants to test if error is the reason for the contributions
Keser
how does he test for error
- redesigns the game so that NE involves some contribution
- so that error can be seen in either direction - distinguish error from those that actually want to contribute
- possible errors on both side on NE
what did keser find
error
- only 30% of people played the dominant strategy
- majority of people made contributions above the dominant strategy level
- few contributed less than 7
what does keser results suggest
- over contribution is not fully explained by error
- if it was we would see error symettrically on either side of NE
- but saw far more people contribute above
- error is not the sole answer
- but shows that still a few people did make errors and contributed less - so still a factor just not big
Andreoni 1998
what was the aim
- separate the learning hypothesis from the strategic hypothesis
- learning = early contributions are due to error, conditional convergence of NE
- strategic = players contribute in hopes of boosting future contributions - and then decays in later rounds
Andreoni
experiment ran
partners and strangers
play game repeatedly
suprise restart - to test learning hypothesis
what were the predictions in Andreoni 1998
- strategic = predicts higher contribution in partners until final round - can influence others decisions - but not in strangers
- learning = no difference in decay rates between strangers and partners - after suprise restart decay should continue not respark
Andreoni findings
- strangers contribute more than partners - not consistent with strategic hypotehsis
- after restart - partenrs return to contribution levels similar to first round - not consistent with learning hypothesis
Croson 1996
replicates strangers and partners
croson 1996 findings
jump from parterns after restart
strangers contribute less than partenrs - strategic hypothesis
Yamakawa 2016
what experiment
3 different treatments
- human treatment
-VCM, 20 rounds, pairs - computer treatment
- each pair is 1 human and 1 computer
- computers choices are predetermined - by a humans choice in the H treatment
- human in this pair cant influence the computers decisions - human computer treatment
- computer plays on behalf of second human who receives its payoffs
what is the incentive behind having computers
and then having computers but playing for a human
Yamakawa
- no incentive to influence its decisions
- no incentive to play altruistically
- if people contribute to the computer - it must be because of error = no other reason?
- playing for a human - will capture the altruism
- solely pick up the effect of wanting to financially benefit the other
rationale behind Yamakawa
no incentive to contribute because against computer
C = captures errors
HC = captures altruism
Yamakawa
results
H - people contribute and then there is a decay at the end
C - no contribution at all = so people not making errors - people werent confused about the game
HC - closer to C (no contribution) than to H - evidence of strategic motive - altruism isnt a big factor otherwise would of been closer to H
- when humans interact with humans that can respond - they contribute more
what papers say that error/learning doesnt play big factor for why
people crontribute
decays
keser - errors on both sides
Andreoni - restart
Yamakawa - C game no one contributes
