劳动经济学手册1

TheEconomicsandEconometricsofActiveLaborMarketProgramsJamesJ.Heckman,UniversityofChicagoRobertJ.LaLonde,MichiganStateUniversityandJe¤reyA.Smith,UniversityofWesternOntarioPreparedfortheHandbookofLaborEconomics,VolumeIII,OrleyAshenfelterandDavidCard,editors.WethankSusanneAckumAgellforherhelpfulcommentsonScan-danavianactivelabormarketprogramsandCostasMeghirforveryhelpfulcommentsonSections1-7.1TheEconomicsandEconometricsofActiveLaborProgramsContents1.Introduction2.PublicJobTrainingandActiveLaborMarketPolicies3.TheEvaluationProblemandtheParametersofInterestinEvaluatingSocialPrograms3.1TheEvaluationProblem3.2TheCounterfactualsofInterest3.3TheCounterfactualsMostCommonlyEstimatedintheLiterature3.4IsTreatmentontheTreatedanInterestingEconomicParameter?4.ThePrototypicalSolutionstotheEvaluationProblem4.1TheBefore-AfterEstimator4.2TheDi¤erence-in-Di¤erencesEstimator4.3TheCrossSectionEstimator5.SocialExperiments5.1HowSocialExperimentsSolvetheEvaluationProblem5.2IntentiontoTreatandSubstitutionBias5.3SocialExperimentsinPractice5.3.1TwoImportantSocialExperiments5.3.2ThePracticalImportanceofDroppingOutandSubstitution5.3.3AdditionalProblemsCommontoAllEvaluations6.EconometricModelsofOutcomesandProgramParticipation6.1UsesofEconomicModels6.2PrototypicalModelsofEarningsandProgramParticipation6.3ExpectedPresentValueofEarningsMaximization6.3.1CommonTreatmentE¤ect6.3.2ASeparableRepresentation6.3.3VariableTreatmentE¤ect6.3.4ImperfectCreditMarkets6.3.5TrainingasaFormofJobSearch6.4TheRoleofProgramEligibilityRulesinDeterminingParticipation6.5AdministrativeDiscretionandtheE¢ciencyandEquityofTrainingProvision6.6TheCon‡ictbetweentheEconomicApproachtoProgramEvaluationandtheModernApproachtoSocialExperiments7.Non-experimentalEvaluations7.1TheProblemofCausalInferenceinNon-experimentalEvaluations27.2ConstructingaComparisonGroup7.3EconometricEvaluationEstimators7.4Identi…cationAssumptionsforCross-SectionEstimators7.4.1TheMethodofMatching7.4.2IndexSu¢cientMethodsandtheClassicalEconometricSelectionModel7.4.3TheMethodofInstrumentalVariables7.4.4TheInstrumentalVariableEstimatorasaMatchingEstimator7.4.5IVEstimatorsandtheLocalAverageTreatmentE¤ect7.4.6RegressionDiscontinuityEstimators7.5UsingAggregateTimeSeriesDataonCohortsofParticipantstoEvaluatePrograms7.6PanelDataEstimators7.6.1AnalysisoftheCommonCoe¢cientModel7.6.2TheFixedE¤ectsMethod7.6.3UtFollowsaFirst-OrderAutoregressiveProcess7.6.4UtisCovarianceStationary7.6.5RepeatedCross-SectionAnalogsofLongitudinalProcedures7.6.6TheFixedE¤ectModel7.6.7TheErrorProcessFollowsaFirst-OrderAutoregression7.6.8CovarianceStationaryErrors7.6.9TheAnomalousPropertiesofFirstDi¤erenceorFixedE¤ectModels7.6.10RobustnessofPanelDataMethodsinthePresenceofHeterogeneousResponsestoTreatment7.6.11PanelDataEstimatorsasMatchingEstimators7.7RobustnesstoBiasedSamplingPlans7.7.1TheIVEstimatorandChoice-BasedSampling7.7.2TheIVEstimatorandContaminationBias7.7.3RepeatedCross-SectionMethodswithUnknownTrainingStatusandChoice-BasedSampling7.8BoundingandSensitivityAnalysis8.EconometricPractice8.1DataSources8.1.1UsingExistingGeneralSurveyDataSets8.1.2UsingAdministrativeData8.1.3CollectingNewSurveyData8.1.4CombiningDataSources8.2CharacterizingSelectionBias8.3ASimulationStudyoftheSensitivityofNonexperimentalMethods8.3.1AModelofEarningsandProgramParticipation38.3.2TheDataGeneratingProcess8.3.3TheEstimatorsWeExamine8.3.4ResultsfromtheSimulations8.4Speci…cationTestingandtheFallacyofAlignment9.IndirectE¤ects,Displacement,andGeneralEquilibriumTreatmentE¤ects9.1ReviewofTraditionalApproachestoDisplacementandSubstitution9.2GeneralEquilibriumApproaches9.2.1DavidsonandWoodbury9.2.2Heckman,Lochner,andTaber9.3SummaryonGeneralEquilibriumApproaches10.ASurveyofEmpiricalFindings10.1TheObjectivesofProgramEvaluations10.2TheImpactofGovernmentProgramsonLaborMarketOutcomes10.3TheFindingsfromU.S.SocialExperiments10.4TheFindingsfromNon-experimentalEvaluationsofU.S.Programs10.5TheFindingsfromEuropeanEvaluations11.Conclusions1IntroductionPublicprovisionofjobtraining,ofwagesubsidiesandofjobsearchassistanceisafeatureofthemodernwelfarestate.TheseactivitiesarecornerstonesofEuropean“activelabormarketpolicies,”andhavebeenafeatureofU.S.socialwelfarepolicyformorethanthreedecades.SuchpoliciesalsohavebeenadvocatedasawaytosoftentheshocksadministeredtothelabormarketsofformerEastBlocandLatineconomiescurrentlyintransitiontomarket-basedsystems.Acentralcharacteristicofthemodernwelfarestateisademandfor“objective”knowl-edgeaboutthee¤ectsofvariousgovernmenttaxandtransferprograms.Di¤erentpartiesbene…tandlosefromsuchprograms.Assessmentsofthesebene…tsandlossesoftenplaycriticalrolesinpolicydecision-making.Recently,interestinevaluationhasbeenelevatedasmanyeconomieswithmodernwelfarestateshave‡oundered,andasthecostsofrunningwelfarestateshaveescalated.Thischapterexaminestheevidenceonthee¤ectivenessofsuchwelfarestateactivela-bormarketpoliciessuchastraining,jobsearchandjobsubsidypolicies,andthemethodsusedtoobtaintheevidenceontheire¤ectiveness.Ourmethodologicaldiscussionofalter-nativeapproachestoev