3. Incentives Flashcards
What did Vernon Smith say about incentives in 1976?
If incentives are structured in appropriate ways- salience, monotonicity, dominance- can then interpret behaviour of experimental subjects as real economic behaviour
Positives and negatives of using incentives
+ potentially increased effort, attention and concentration
+ less intentional misreporting and affective reactions (inability to predict properly motivated decision)
-expensive
What do T&K 1979 say about incentives?
Hypothetical incentives relies on the assumption that people often know how they would behave in actual situations of choice and on the further assumption that subjects have no special reason to disguise their true preferences
What do Camerer & Hogarth 1992 do?
Review of 74 studies with varied incentive levels and tasks with objective criterion of success
Results of Camerer & Hogarth 1992
-incentives matter but in complex ways
-improve performance in mundane tasks requiring attention
-incentives can cause over arousal and choking
-some tasks have no affect- trading and bargaining
-increasing incentives reduces noise and reduced favourable self presentation
-learning by doing is an important factor
What do Holt & Laury 2002 do?
Elicit risk aversion- ten paired lottery choice decisions with payoffs L, 20L, 50L, 90L. They observe decisions with real and hypothetical choice (within and between subjects)
Key results of Holt & Laury 2002
-Scaling up payoffs made little difference when payoffs were hypothetical
-when payoffs are real, risk aversion increases “sharply” with payoff magnitude
-suggests you need large enough incentives to see true behaviour. Some people may not be able to anticipate how they would act when incentives are real.
Describe the random incentive scheme
When subjects do multiple tasks it is common to implement RIS. Real task is selected at random independently of each subject at the end of the experiment
Positive of RIS
Encourages subjects to think about tasks separately which reduces income effects. It economises an experimenters budget and time
What was Holt’s critique of RIS in 1986?
If subjects using RIS
-view whole experiment as a single choice among compound lotteries (independence axiom doesn’t hold)
-and reduce compound lotteries to simple ones in standard way
Then unless subjects have EU preferences, RIS may bias subjects responses to individual tasks
What would justify RIS without needing EU independence?
If subjects isolate (think about tasks in isolation)
What did Starmer & Sugden 1991 set out to do?
Tests Holt’s theory of RIS not being good
What theory did Starmer and Sugden 1991 hypothesise?
If subjects behave exactly according to Holt’s conjecture we shouldn’t observe Allais’ famous CCE in an experiment using RIS. So they look for CCE in an RIS experiment
Conclusions of Starmer & Sugden 1991
Standard CCE occurrence of picking A and D provides evidence against claim that RIS biases choices in precisely the way assumed by Holt. Leaves open the possibility of some Holt- like bias that there is contamination across tasks
What did Cubitt Et Al 1998 set out to do?
Examine if there is Holt like bias in RIS and if there is contamination across tasks
Design of Cubitt Et Al 1998
Two choice tasks (P’, P”) four conditions between subjects. Each of P’ and P” observed 4 conditions
-single real choice
-hypothetical
-RIS (with high value tasks)
-RIS (with low value tasks)
1) test for CRE across incentive conditions
2) test for contamination effect by comparing responses to (P’,P”) across two RIS conditions
What would we expect from the Cubitt Et Al 1998 study?
Given auxiliary of “fanning out” we would expect less risk averse behaviour in (P’, P”) in the RIS with low value tasks
Results of Cubitt Et Al 1998
-no evidence of contamination
Two possible interpretations
-RIS is okay
-RIS is not okay but our subjects happen to have roughly EUT preferences
CRE is found in hyp, real & RIS
Conclusion of Cubitt Et Al 1998
Tentatively supports the use of RIS
What does Selten Et Al 1999 find?
Money doesn’t induce risk neutral behaviour but lotteries do an even worse job
Whet can be said about theoretical vs behavioural incentive compatibility?
Economists like incentive compatible mechanisms but not all theoretically incentive compatibility mechanisms turn out to be behaviourally incentive compatible e.g case of binary lottery incentives. Achieving behavioural IC requires consideration of real human psychology
In Holt & Laury 2002, what was the makeup of subjects?
50% undergraduate, 33% MBA students, 17% business school faculty. Less subjects used for high payoff tasks due to expenses
In Holt & Laury 2002 describe the design of real and hypothetical tasks
Low payoff decisions always done but high payoff was sometimes real and sometimes hypothetical. Half of subjects did both hypothetical and real for within subject comparison
In Holt & Laury 2002, what was the order of tasks and why?
High hypothetical done before high real to allow us to hold wealth constant and to evaluate the effects of using real incentives
In the Holt & Laury 2002 study, what were the main conclusions?
-people are more risk averse in real tasks and significantly more risk averse in high payoff real tasks
-subjects underestimate the extent to which they will avoid risk. The evidence of risk aversion suggests the potential dangers of analysing behaviour under the simplifying assumption of risk neutral
In Starmer & Sugden 1991, what do they highlight as key benefits of RIS?
-economises the cost of experiments
-avoids problems of reference point and wealth effects that would be created if subjects were paid according to their performance on each of a number of tasks
In Starmer & Sugden 1991 what does prospect theory say about RIS?
If subjects use editing phase of decision making to simplify complex choice problems they will reveal their true preferences in the RIS
Set up of subjects and specific details in Starmer & Sugden 1991
160 subjects in 4 groups of 40.
Half of subjects know which question is the “real” one, the other half don’t.
Half do problem P’ first other half do P” first having already done 20 irrelevant decisions tasks as part of pilots for other experiments
Conclusions of Starmer & Sugden 1991
-can’t claim to prove unbiasedness of RIS, further investigation and larger sample size is needed
-results provide evidence of choice of subjects in RIS not being consistent with reduction principle going against Holt’s hypothesis
-if there are any contamination effects, they seem to be weak
In Cubitt Et Al 1998 what are the main findings?
3 experiments designed to detect cross task contamination effects in RIS. No significant differences between treatments, and no significant contamination effects were found despite evidence of fanning out in experiment 3
What is the reduction hypothesis?
It is the hypothesis put forward by Holt 1986 that the subject treats the whole experiment as a single decision problem reducing all compound prospects to simple ones using the standard probability calculus
How do Cubitt Et Al 1998 test for contamination?
First they test for CRE in hypothesised choice, random lottery and single choice treatments. Second, by comparing responses to the CRE tasks between the two random lottery groups, we can test for contamination effects
What does dominance mean?
The incentives you propose must be big enough that they dominate any other incentives that a subject might have
What ideas about incentives do T&K 1979 miss out on?
That incentives might be required for;
effort
attention
concentration
What are incentive compatible mechanisms?
Tools intended to motivate subjects to reveal things the experimenter wants to observe( e.g beliefs, preferences)
How can the CRE change with incentives?
It is possible that the CRE isn’t in fact a property of true preferences even though it is found in hypothetical choice and random lottery experiments. That interpretation would be consistent with the hypothesis that the stronger incentives are, the less deviation there is from EUT. (Cubitt Et Al 1998)
What are the 4 treatment groups in Cubitt Et Al
G1: p’ for real, p” hypothetical, 18 high value hypotheticals
G2: p’ hypothetical, p” real, 18 high value hypotheticals
G3: RIS with all 20 tasks, high value
G4: RIS with all 20 tasks, low value
How could a strong stance on incentives by viewed by other academics?
Might be seen as a marketing device in order to distance themselves from psychologists. Incentives may thus be seen as a form of product differentiation in the research market
What do psychologists offer as a negative for incentives?
Incentives present an obstacle to less-experienced researchers who find it harder to secure research funding
Why might incentives increasing performance not be desirable?
Incentives may conflict with the objective of an experiment. For example in Castellan 1969 increased incentives cause subjects to choose red light (profit maximising option) more frequently but if the goal of the experiment is to test instinctive judgements incentives may conflict with this.
What is motivational crowding out and who coined it?
The expectation to perform well from added incentives amounts to a distraction from intrinsic motivation (Frey & Stutzer 2006)