Week 1: Introduction - Piotr Evdokimovpiotr-evdokimov.com/Lecture1-Intro.pdf · Week 1:...
Transcript of Week 1: Introduction - Piotr Evdokimovpiotr-evdokimov.com/Lecture1-Intro.pdf · Week 1:...
Week1:Introduction
Economics?
• Econ101:economicsisthestudyoftheallocationofscarceresources
• LionelRobbins(1898-1984):“Economicsisthesciencewhichstudieshumanbehaviorasarelationshipbetweengivenendsandscarcemeanswhichhavealternativeuses”
• Soundskindofboring• Buttruethateconomistsloveconstrainedoptimization
Aseconomictheorists,weorganizeourthoughtsusingwhatwecallmodels.Theword“model”soundsmorescientificthan“fable”or“fairytale”althoughIdonotseemuchdifferencebetweenthem.Theauthorofafabledrawsaparalleltoasituationinreallife.Hehassomemoralhewishestoimparttothereader.Thefableisanimaginarysituationthatissomewherebetweenfantasyandreality.
Anyfablecanbedismissedasbeingunrealisticorsimplistic,butthisisalsothefable’sadvantage.Beingsomethingbetweenfantasyandreality,afableisfreeofextraneousdetailsandannoyingdiversions.Inthisunencumberedstate,wecanclearlydiscernwhatcannotalwaysbeseenintherealworld.Onourreturntoreality,weareinpossessionofsomesoundadviceorarelevantargumentthatcanbeusedintherealworld.
Wedoexactlythesamethingineconomictheory.Agoodmodelineconomictheory,likeagoodfable,identifiesanumberofthemesandelucidatesthem.Weperformthoughtexercisesthatareonlylooselyconnectedtorealityandthathavebeenstrippedofmostoftheirreal-lifecharacteristics.However,inagoodmodel,asinagoodfable,somethingsignificantremains.
• Asinthecaseoffables,absurdconclusionsrevealcontextsinwhichthemodelproducesunreasonableresults,butthismaynotnecessarilymakethemodeluninteresting.• Asinthecaseoffables,modelsineconomictheoryarederivedfromobservationsoftherealworld,butarenotmeanttobetestable.• Asinthecaseoffables,modelshavelimitedscope.(idealizedconditions,etc.)• Asinthecaseofagoodfable,agoodmodelcanhaveanenormousinfluenceontherealworld,notbyprovidingadviceorbypredictingthefuture,butratherbyinfluencingculture.
Yes,Idothinkwearesimplythetellersoffables,butisthatnotwonderful?
…everythingweknowisonlysomekindofapproximation,becauseweknowthatwedonotknowallthelaws asyet.[…]
Theprincipleofscience,thedefinition,almost,isthefollowing:Thetestofallknowledgeisexperiment.Experimentisthesolejudgeofscientific“truth.”Butwhatisthesourceofknowledge?Wheredothelawsthataretobetestedcomefrom?Experiment,itself,helpstoproducetheselaws,inthesensethatitgivesushints.Butalsoneededisimaginationtocreatefromthesehintsthegreatgeneralizations-toguessatthewonderful,simple,butverystrangepatternsbeneaththemall,andthentoexperimenttocheckagainwhetherwehavemadetherightguess.
Doeconomistsinfluenceculture?
• Yes,notalwaysingoodways…
• https://www.nytimes.com/2019/08/21/podcasts/the-daily/business-roundtable-corporate-responsibility.html• 6:36
Causeandeffect
• “Althoughpurelydescriptiveresearchhasanimportantroletoplay,webelievethatthemostinterestingresearchinsocialscienceisaboutcauseandeffect” -AngristandPischke
• Examples:Theeffectofclasssizeontestscores,schoolingonwages,colonialinstitutionsoneconomicgrowth(Acemoglu andRobinson),neighborhoodeffectsonlifeoutcomes(Chetty)
• Typically,economistsuseobservationaldata
Observationaldata
• Ex:Doesgoingtothehospitalmakepeoplehealthier?
• Data:97823participantsin NationalHealthInterviewSurvey
• Y:“ Wouldyousayyourhealthingeneralisexcellent,verygood,good,fair,poor?”(1-5)
• X:“Duringthepast12months,wereyouapatientinahospitalovernight?”
Data
Problem:selectionbias
Counterfactuals
• Toestimatetheeffectofgoingtothehospital,weneedtomodelcounterfactuals
• Y1=healthstatusifyougotothehospital(regardlessofwhetherornotyouwent)
• Y0=healthstatusifyoustayhome(regardlessofwhetherornotyouactuallydo)
• Noticewedon’tobserveY1thosethatstayhomeandwedon’tobserveY0forthosewhogotothehospital
Randomassignmentsolvestheselectionproblem:
Regressionanalysis
Data
Parameterstoestimate
UnobservableerrorGoodgenes,gooddiet,exercise,badhabits,etc
Regressionanalysis
Regressionanalysis
• Noselectionbias⇒ E(h|D)doesnotdependonD
• I.e.,Dcarriesnousefulinformationpredictingtheobservables
• Thisiscalledexogeneity ofindependentvariables
• Oneofthemostimportantassumptionsinallofeconometrics
• IfE(h|D=1)= E(h|D=0)=µ,lete=h-µ
• Re-writethemodelas
Y=a+µ+r*D+h-µ==a’+r*D+e
⦁ NoticeE(e|D)=0(thisishowtheexogeneity assumptionistypicallystated)
⦁ Youcan’tidentifycausaleffectswithoutthisassumption!!!!
Backtocauseaneffect(schoolingonwages,etc)…
• Simplestmodel:y=a+b*X+e
• yisthedependentvariable(e.g.,wages)
• xistheexplanatoryvariable(e.g.,schooling)
• e istheerrorterm
• b isthetreatmenteffect
• Econometricsprovidesuswithasetofmethodsforobtaininganestimate b ofb • E.g.,b=cov(X,y)/Var(X)
• Thesemethodsrequiretheexogeneity assumption,whichisoftenviolated
• Estimatingthetreatmenteffectwithobservationaldataishard
Experimentalmethods
• Inanexperiment,theexplanatoryvariableisexogenousbydesign
• Thisallowsustoidentifycausaleffects
• …andconductsomeofthemostinterestingresearchinsocialscience
Classroomexperimenthere?
History• Chamberlin(1948)
• Firstclassroomexperiment
• VernonSmith• StudentofChamberlin• Moremarketexperimentsinthe60s(Purdue)• NobelPrize(2002,w/DanielKahneman)
• Gametheory• VonNeumannandMorgenstern(1944)• JohnNash(1950)• RAND,Prisoner’sDilemma(1950)
• Source:CharlieHolt’stextbook
Methods
Methods
• Mostlabexperimentsconsistofseveraltreatments
• Betweensubjects• Onesubject>onetreatment
• Withinsubjects• Eachsubjectexposedtomultipletreatments
Hertwig andOrgmann (2011)
• Experimentaleconomicsmethodsdifferfrompsychologyin:• Scriptenactment• Repeatedtrials•Monetaryincentives• Proscriptionagainstdeception
•Why?Lessuncertainty
Canyoupredictwhatstudiesreplicate?
• Canyoupredictwhatstudiesreplicate?• https://80000hours.org/psychology-replication-quiz/• Funarticle:https://www.nytimes.com/2013/04/28/magazine/diederik-stapels-audacious-academic-fraud.html?pagewanted=all
Buteconomicsisn’tdoingmuchbetter…
Fieldvs.labexperiments
• I’llillustratethisusingtheexampleofgiftexchange• ThestorystartswithAkerlof (1982):
Fehr,Kirchsteiger,andReidl (1993)
• Firmmovesfirstandmakesawageofferw∈ {0,5,10,…}• Workermovessecondandchoosesefforte∈ [0.1,1]• Effortiscostly
• Gametheoryprediction:e=0.1,w=0
Gneezy andList(2006)
• Workershiredtocodelibrarybooks• 6hoursfor$12/hour
• Treatment:surpriseannouncementof$20/hour
• Result:subjectswithsurpriseannouncementworkharder
• Explanation?
Thefutureoflabexperiments
Thefutureoffieldexperiments
Thisclass
• Thefocusinthisclasswillbeonlabexperiments
• Inparticular,labexperimentswithagametheoryfocus
• Inparticular,recentlabexperimentswithagametheoryones
• Wewillfoscus onnewresearchthatIpersonallyaminterestedin
Next
1. Gametheorybasics(rationality,rationalizability,dominance).You’llneedtoknowsomemath
2. Someexperimentsonthebasics3. Moregametheory(Nashequilibrium).You’llneedtoknowsomemath4. SomeexperimentsonNashEquilibrium5. Moregametheory6. Moreexperiments7. Etc8. Afewweeksfromnow,stoptalkingaboutgametheoryexperimentsanddiscuss
• Behavioraleconomics(You’llneedtoknowsomemath)• Behavioraleconomicsfromthe“neoclassical”perspective(You’llneedtoknowsomemath)• Decisiontheory(experimentsonrisk,time,etc.You’llneedtoknowsomemath)
9. Ifyoudon’tcareaboutresearch,dropthisclass!!!