BEEM109BEEM109Experimental Economics and Experimental Economics and
FinanceFinance20092009
Dieter BalkenborgDieter Balkenborg
Times and roomsTimes and rooms
Monday 11-13Monday 11-13 Friday 1-2Friday 1-2 Both in Xfi computer suiteBoth in Xfi computer suite I may want to use FEELE for some I may want to use FEELE for some
Finance and Game experiments Finance and Game experiments (Econport software does not run here.)(Econport software does not run here.)
FIVE lectures then your experiments FIVE lectures then your experiments and presentationsand presentations
Today’s lecture Today’s lecture
IntroductionIntroduction OrganizationOrganization How to run an experimentHow to run an experiment
Next lectureNext lecture
Markets 2Markets 2 Bertrand with complements / MonopolyBertrand with complements / Monopoly CournotCournot CollusionCollusion
Financial MarketsFinancial Markets
Further topicsFurther topics
Financial Market experimentsFinancial Market experiments Experiments in Game TheoryExperiments in Game Theory FairnessFairness NeuroeconomicsNeuroeconomics Anomalies in Decision making under Anomalies in Decision making under
uncertainty and behavioural finance uncertainty and behavioural finance
IntroductionIntroduction
Is economics a science?Is economics a science? The role model: physicsThe role model: physics
Main characteristics:Main characteristics: Rigid mathematical theoriesRigid mathematical theories Rigid experimentally testingRigid experimentally testing
Example: Galilei and the tower of PisaExample: Galilei and the tower of Pisa
s=-1/2ts=-1/2t22g, material not relevantg, material not relevant
And Economics?And Economics?
Rigorous mathematical models: Plenty!Rigorous mathematical models: Plenty! Core model: general equilibriumCore model: general equilibrium Conflict with institutionalists, example of Conflict with institutionalists, example of
economic historyeconomic history Empirical research: PlentyEmpirical research: Plenty
But it assumes rather than tests economic But it assumes rather than tests economic theory theory
Experiments:Experiments: A very recent developmentA very recent development
Need for experimentsNeed for experiments
Ways to evaluate theories: Ways to evaluate theories: plausibility, generality, predictive plausibility, generality, predictive successsuccess
Difficult to sort out cause and effect Difficult to sort out cause and effect when many factors change when many factors change simultaneouslysimultaneously
Arguments against Arguments against experimentsexperiments
Human behaviour not predictable Human behaviour not predictable because of their free willbecause of their free will
Laboratory situations do not capture Laboratory situations do not capture the real world (subjects are paid the real world (subjects are paid peanuts), but our models do.peanuts), but our models do.
Utility maximization is a tautologyUtility maximization is a tautology Utility is not the same as moneyUtility is not the same as moneyExperiments check the validity of Experiments check the validity of
argumentsarguments
Main ProblemMain Problem
Internal validityInternal validity External validityExternal validity
Applies to theory as well!Applies to theory as well!
PioneersPioneers
Chamberlain Chamberlain (Simon*: bounded rationality, satisficing, AI)(Simon*: bounded rationality, satisficing, AI) Nash* / Milnor, RandNash* / Milnor, Rand Sauermann / Selten*Sauermann / Selten* Smith* / PlotSmith* / Plot Allais*Allais* Kahneman* / TverskiKahneman* / Tverski Fouraker / SiegelFouraker / Siegel
How to conduct experimental How to conduct experimental researchresearch
Get an ideaGet an idea read the literature, but not too muchread the literature, but not too much
Get moneyGet money Design the experiment. Design the experiment.
What is the hypothesis to be tested? What is the hypothesis to be tested? Model the situation (like a mathematical Model the situation (like a mathematical
model)model) Simplicity over realismSimplicity over realism Arbitrary design decisionsArbitrary design decisions Programming / PresentationProgramming / Presentation Instructions: words can matterInstructions: words can matter
How to conduct experimental How to conduct experimental researchresearch
Sufficiently many repetitions / budgetSufficiently many repetitions / budget HonestyHonesty
Run pilot: back to the drawing board?Run pilot: back to the drawing board? Recruit subjectsRecruit subjects Run sessions (handrun / computerized)Run sessions (handrun / computerized) Pay subjectsPay subjects Prepare data for evaluationPrepare data for evaluation Evaluate beyond initial hypothesis. Data Evaluate beyond initial hypothesis. Data
study vs econometrics, Statistical methodsstudy vs econometrics, Statistical methods Find theoretical explanation.Find theoretical explanation.
Schools of thought Schools of thought (parody?)(parody?)
Read Selten (1991) on various positionsRead Selten (1991) on various positions Neoclassical economist / BayesianNeoclassical economist / Bayesian
Economics relies on very few psychological Economics relies on very few psychological assumptions which are self-evidently true. Hence assumptions which are self-evidently true. Hence there is not need to conduct experiments or to there is not need to conduct experiments or to change economics whenever there is a new fad change economics whenever there is a new fad in psychology. (Robbins)in psychology. (Robbins)
The neoclassical experimentalistThe neoclassical experimentalist Behavioural anomalies are short-run phenomena Behavioural anomalies are short-run phenomena
which vanish if people have the opportunity to which vanish if people have the opportunity to learn from mistakes and if the experiment is learn from mistakes and if the experiment is properly designed. (Binmore, Smith?)properly designed. (Binmore, Smith?)
Schools of thoughtSchools of thought
Behavioural economist (Fehr / Schmid / Behavioural economist (Fehr / Schmid / Camerer)Camerer) Adjust utility theory such that it fits data better Adjust utility theory such that it fits data better
(fairness) (fairness) Behavioural Finance (Thaler, Schleifer)Behavioural Finance (Thaler, Schleifer)
Find anomalies and behavioural biases in Find anomalies and behavioural biases in financial data which allow you to make moneyfinancial data which allow you to make money
Bounded Rationality / Cognitive Bounded Rationality / Cognitive psychology (Simon / Selten)psychology (Simon / Selten) Procedural models of thinkingProcedural models of thinking ““Bottom up” process of modellingBottom up” process of modelling
Let us run an experiment!Let us run an experiment!
Open web browserOpen web browser Type “FEELE kiosk” in search box top Type “FEELE kiosk” in search box top
rightright Etc.Etc. Experiment: Roise Nagel’s guessing Experiment: Roise Nagel’s guessing
gamegame
How to run an experimentHow to run an experiment
1.1. PurposePurpose2.2. DesignDesign3.3. SubjectsSubjects4.4. Instructions, test roundsInstructions, test rounds5.5. PilotsPilots6.6. RecruitmentRecruitment7.7. PaymentsPayments8.8. Statistical evaluationStatistical evaluation
Purpose (Roth)Purpose (Roth)
Speaking to theoristsSpeaking to theorists Forward vs backward inductionForward vs backward induction
Searching for factsSearching for facts Are economics students different?Are economics students different?
Whispering in the ears of princessWhispering in the ears of princess Roth and kidney exchangeRoth and kidney exchange Test bed, wind channel, Grether and Plott: Test bed, wind channel, Grether and Plott:
testing existing market instututionstesting existing market instututions Which hypothesis do you want to test?Which hypothesis do you want to test?
Design of a sessionDesign of a session
Simplicity, transparency and control vs Simplicity, transparency and control vs realismrealism Internal vs external validityInternal vs external validity
Game character Game character Time limitTime limit ““independent observations”independent observations” Learning vs one-shot games, repetitions, Learning vs one-shot games, repetitions,
which do count?which do count? HonestyHonesty PaymentPayment No need to use computers!No need to use computers!
Design of an experimentDesign of an experiment
HOW MANY SESSIONS?HOW MANY SESSIONS? BUDGET CONSTRAINT!BUDGET CONSTRAINT! Focus vs nuisance variablesFocus vs nuisance variables Treatment variablesTreatment variables Keeping nuisance variables constant Keeping nuisance variables constant Randomization designRandomization design Block designBlock design Fractional factorial designFractional factorial design Double blind experimentsDouble blind experiments Within subject designWithin subject design Ensure that enough subjects show up on time!Ensure that enough subjects show up on time!
Subjects and roomSubjects and room
Typically: studentsTypically: students ““Practitioners” will bring their Practitioners” will bring their
experience to the labexperience to the lab Don’t pay executives!Don’t pay executives! ““field” experimentsfield” experiments Prevent uncontrolled communication Prevent uncontrolled communication
and exchange of information (video and exchange of information (video experiments)experiments)
Instructions, test roundsInstructions, test rounds
The biggest bore, necessary like in a new The biggest bore, necessary like in a new parlour game parlour game
Clarity of instructions is crucial Clarity of instructions is crucial Avoid jargon, misinterpretations, allusions to Avoid jargon, misinterpretations, allusions to
the real worldthe real world Foreign languagesForeign languages Don’t bias subjects by giving away the Don’t bias subjects by giving away the
purpose of the experiment! purpose of the experiment! Test questionsTest questions Test roundsTest rounds
Pilot sessionsPilot sessions
Necessary to check for bugs, both in Necessary to check for bugs, both in software, design and instructionssoftware, design and instructions TyposTypos Incorrect implementationIncorrect implementation How expensive will it get?How expensive will it get? Time length.Time length.
RecruitmentRecruitment
Now all computerized, we get our Now all computerized, we get our experiments filled within hoursexperiments filled within hours
Possibility to restrict access to Possibility to restrict access to certain groupscertain groups
PaymentsPayments
Payment = money Payment = money Payoff = utilityPayoff = utility Controlling for risk aversion (Roth, Controlling for risk aversion (Roth,
Malouf), “irrational” risk avoidanceMalouf), “irrational” risk avoidance Controlling for fairness and spite Controlling for fairness and spite Not every round has to be paidNot every round has to be paid Keep it secret!Keep it secret!
Statistical evaluationStatistical evaluation
Descriptive statisticsDescriptive statistics Often there are no established theories or Often there are no established theories or
these are themselves in questionthese are themselves in question Keep it simple, use non-parametric statistics: Keep it simple, use non-parametric statistics:
Binomial test, sign test, Wilcoxen test, Fisher Binomial test, sign test, Wilcoxen test, Fisher test, Spearman’s rank correlation testtest, Spearman’s rank correlation test
All parametric and econometric tests have All parametric and econometric tests have many more auxiliary assumptions, also many more auxiliary assumptions, also probit and logit approachesprobit and logit approaches
Testing initial hypotheses vs generating new Testing initial hypotheses vs generating new knowledge.knowledge.
Suggestions for Suggestions for ExperimentsExperiments
1.1. Network Externalities. Network Externalities. 2.2. Insurance.Insurance.3.3. Bubble experiment.Bubble experiment.4.4. 2x2 games2x2 games5.5. Hold up problemHold up problem6.6. Sports DraftSports Draft7.7. (Mini) ultimatum games / Dictator games / (Mini) ultimatum games / Dictator games /
reciprocityreciprocity8.8. Bank Runs. Bank Runs. 9.9. Signalling. Signalling. 10.10. Bertrand Complements.Bertrand Complements.11.11. Cournot MarketCournot Market12.12. Behavioural Biases / AnomaliesBehavioural Biases / Anomalies13.13. Cascade experimentCascade experiment14.14. Public Good/PD. Public Good/PD.