斯坦福大学计量经济学应用lecture1

NickBloom,AppliedEconometrics,Winter2010APPLIEDECONOMETRICSLecture1-IdentificationNickBloom,AppliedEconometrics,Winter2010DefiningIdentificationExperimentsNaturalExperimentsInstrumentalvariablesEconometricIdentificationNickBloom,AppliedEconometrics,Winter2010WHATISIDENTIFICATION?GraduateandprofessionaleconomicsmainlyconcernedwithidentificationinempiricalworkConceptofunderstandingwhatisthecausalrelationshipbehindempiricalresults:Thisisessentialforlearningfromempiricalresearch•Time-seriesexample:InterestratesandGDP•Cross-sectionexample:Management&ProductivityNickBloom,AppliedEconometrics,Winter2010WHATISDRIVINGTHISRELATIONSHIP?05101520Federalfundsrate-.15-.1-.050.05.1Detrendedlogindustrialoutput196019701980199020002010ym...DetrendedlogindustrialoutputFederalfundsrateCorrelation=0.233NickBloom,AppliedEconometrics,Winter2010REASONSFORCORRELATIONImaginevariablesYtandXtarecorrelated:Therecanbethreereasonsforthis,whicharenotmutuallyexclusive:•Cause:ChangesinXtdrivechangesinYt•ReverseCause:ChangesinYtdrivechangesinXt•Correlatedvariable:ChangesinZtdrivesXtandYtNickBloom,AppliedEconometrics,Winter2010WHATISDRIVINGTHISRELATIONSHIP?2468Productivity12345ManagementNickBloom,AppliedEconometrics,Winter2010SOHOWDOWEGETIDENTIFICATIONFourbroadapproachesforidentification•Experiments–yougeneratethevariation•NaturalExperiments–youknowwhatgeneratedthevariation•Instrumentalvariables–youhaveavariablethatcanprovideyouvariation•EconometricIdentification–yourelyon(testable)econometricassumptionsforidentificationNickBloom,AppliedEconometrics,Winter2010DefiningIdentificationExperimentsNaturalExperimentsInstrumentalvariablesEconometricIdentificationNickBloom,AppliedEconometrics,Winter2010EXPERIMENTS(1)ExperimentsaretotallystandardinScience&MedicineForexample:•Setupatreatmentandcontrolgroupforanewdrug,makingsurethesearecomparable(orrandomlyselected)•Ensurethesamplesizesarelargeenoughtoobtainstatisticalsignificance•Ensuretheexperimentisunbiased–i.e.thedrugandtheplaceboareassimilaraspossible•RuntheexperimentNickBloom,AppliedEconometrics,Winter2010EXPERIMENTS(2)EconomistsliketousethelanguageofScienceForexampletheUKconsideredintroducinganEducationMaintenanceAllowance,topaykidstostayonatschool.Butwanttotestfirsttoseeifthiswouldthiswork.•Setupatreatmentandcontrolregionstomatchtheseincharacteristics•Selectenoughregionstogetlargesamplesizes•Observeagentsactionstoevaluateimpact(ratherthanselfreportedoutcomes)•RuntheexperimentNickBloom,AppliedEconometrics,Winter2010EXPERIMENTS(3)Experimentsarerareineconomicsbecausetheyareexpensive,althoughtheybecomingmorepopular:Typicalareasforrunningexperimentsinclude:•Developmenteconomics–cheapertorunexperimentsinthethirdWorld(watersupplyormanagementpractices)•Consumereconomics–smallstakesexperimentsthatareeasytoadminister(creditcards)•Individualbusinessapplications–firmscanfinancethese(retailstorelayout)Butsomefieldswillneverhaveexperiment–forexamplemacroeconomicsNickBloom,AppliedEconometrics,Winter2010DefiningIdentificationExperimentsNaturalExperimentsInstrumentalvariablesEconometricIdentificationNickBloom,AppliedEconometrics,Winter2010NATURALEXPERIMENTS(1)Naturalexperimentsarewherefortunatesituationscreategoodunderlyingidentification:Typicallyseveralapproaches:•Taxe.g.ResponseofR&Dtothecostofcapital(Bloom,Griffith&VanReenen,2002),(ChettyandSaez,2003)•Discontinuity(seeover)•Shock-financialcrisisandKibutzim(Abramitzky,2007)•Disasters-EthiopianJewsairlift(Gould,Levy&Passerman,2004)NickBloom,AppliedEconometrics,Winter2010NATURALEXPERIMENTS(2)NaturalexperimentsarealmostthehollygrailofmodernappliedeconomicsIntheabsenceoftrueexperimentstheyprovidethebestwaytoprovidesimpleidentificationCoupleofstandardwaytousenaturalexperimentsinpractice–Discountinunityanalysisand/or–DifferenceindifferencesNickBloom,AppliedEconometrics,Winter2010DISCONTINUITYANALYSIS–example1RegionA(notax)RegionB(50%tax)Imaginea50%taxisleviedoninvestmentintherichcoastalregionAbutnotinthepoorinlandregionB.Ifyousawthegraphbelowcouldyousaywhattheimpactofthetaxisoninvestment?InvestmentEstimatedimpactofthetaxNickBloom,AppliedEconometrics,Winter2010DISCONTINUITYANALYSIS–example2ImpactoftelephonesonpriceoffishinKerala(India)NickBloom,AppliedEconometrics,Winter2010DIFFERNCESINDIFFERENCESt0denotespre-treatmentperiodsforwhichdataareavailablet1denotespost-treatmentperiodsforwhichdataareavailableAveragechangeinoutcome(preandpost-treatment)fortreatmentgroupminusaveragechangeinoutcomeforcontrolgroupControltControltTreattTreattRRRR0101ImpactIdentificationcomesfromthedifferentialchangebetweenthetwogroupspreandpost-treatment–differenceoutunobservedfixedeffects–differenceoutcommontimeeffectsKeyassumptionofcommontimeeffectsforthetwogroupsNickBloom,AppliedEconometrics,Winter2010POLICYEXAMPLEOF“DIFF-IN-DIFF”•SmallfirmsR&Dtaxcreditintroducedin2000forfirmswith250orlessemployees•Socouldlookatfirmsbeforeandaftercredit–Butotherthingsalsochanging(2000peakofdotcomboometc…)–Soneedtosetupacontrolgroupofcompanieslooksimilartofirmsgettingthecreditexceptdon’tgetthecredit•Comp