23 - experience under uncertainty Flashcards
what are anomalies
choice behaviour that violates standard of rationality embedded in orthodox economics
what is CRE - form of allais paradox behaviour
tendency to choose safe option in problem 1 - and risky option in problem 2 when probabilities of non-zero outcomes are scaled down by multiplying by common ratio
what does CRE violate
expected utility theory
why does CRE occur
subjects evaluating options using decision weights that depart from probability
- overweight rare events
- underweight common events
what is the difference between task experience and outcome experience
task experience = repeating the choice task - familiarity
outcome experience = knowing the outcome of the chosen option - seeing it be resolved and the consequence
Kuilen & Wakker 2006
aim
does outcome experience make a difference to prevalence of CRE
- experience of die rolling, outcomes, foregone outcomes
Kuilen & Wakker
design
- subjects repeatedly make choices between lotteries each round - choices have a common ration structure - face choice pairs
- 2 groups
- no feedback = make choices each round but dont see outcome of chosen lotteries until the end - gain task experience (from repeating)
- feedback = after each round, subjects roll a die to determine outcome of lottery
Kuilen & Wakker
results
with feedback
- outcome experience reduces incidence of anomaly - less EUT violations
- downward trend
- percentage of people choosing consistently with expected value maximisation increases
no feedback
- no trend in level of EUT violations
- task experience does not affect incidence
why does feedback reduce CRE?
Kuilen & wakker
- die rolling - learn how high probability 0.8 is likely to 0.2
- learn from outcomes - that you are more likely to do well in risky option in problem 1 - so switch
- comparing outcome achieved with outcome foregone
Bone et al 1999
design
stage 1. individuals make 12 choices in 4 common ration triples
stage 2. payoffs doubled and subjects in pairs - must discuss and agree their choices
stage 3. same as stage 1
Bone 1999
aim
- are groups / pairs more consistent than individuals
- compare stage 1 and 2
- does experience of group discussion improve the subsequent decisions of indivuals
- comparing stage 1 and stage 3
Bone et al
findings
group is less consistent with EUT than individuals
- proportion of EUT decisions decreases from stage 1 to stage 2
- more CRE
- little evidence of experience of group discussion raising conformity of individuals to EUT
- no difference between 1 and 3
- no evidence of teaching effect - someone who is EUT consistent doesnt teach partner
from papers
does experience reduce the incidence of CRE - and violations of EUT
doesnt
* task repetition
* task discussion
- outcome experience = does
how to consider if descriptions of risks make a difference?
- are patterns of risk preference seen in experiments the same when awareness of uncertainty is learned by experience
- instead arent told the probabilities - gradually learn them yourself - more like in field
why would choices change
depending on whether they discover about risks from complete descriptions or by sampling resolutions
- sampling bias - wont match true probabilities
- people have uncertainty about uncertainty
what is an example of discovering about risks from sampling resolutions
rolling a die
1 = bad
2-10 = good
- most people in sample would never see bad outcome
- from description you learn of the small chance of bad outcome
Cubitt et al (2022)
- choose between 2 options, each option has 2 outcomes
- options are resolved by drawing cards from virtual deck 1/40 - 2 colours
Cubitt
what are the 3 treatment groups
and the control
what do they test for
control = see descriptions of options and probabilities expressed as %s of the deck
unambiguous = subjects see all the 40 cards - know its full deck
ambiguous = subjects dont know that the 40 cards are the whole deck - tests for lack of certainty
restricted = sample is only 18 cards - tests for small sample bias
Cubitt
results
shape of probability weighting function
small effects
- small effect of changing presentation of equivalent information (description vs shown deck)
- small effect of whether subjects know they have seen whole deck (unambiguous vs ambiguous)
Cubitt
results
shape of probability weighting function
large effects
- large effect between ambiguous and restricted
- when restricted sample under represents low probability event they under weight it - dont overweight low probabilities
Cubitt
takeaway
choices and probability weights are affected by whether uncertainties described to subjects or inferred by them from sampling experience
- small sample bias - underweights low probability events
- inverse S probability weighting not artefact of description