Stata高级应用：Heckman两步法在犯罪学研究中选择偏差修正探讨.pdfVIP

JQuantCriminol(2007)23:151–178

DOI10.1007/s10940-007-9024-4

ORIGINALPAPER

IstheMagicStillThere?TheUseoftheHeckman

Two-StepCorrectionforSelectionBiasinCriminology

ShawnBushwayÆBrianD.JohnsonÆLeeAnnSlocum

Publishedonline:23March2007

Science+BusinessMedia,LLC2007

Issuesofselectionbiaspervadecriminologicalresearch.Despitetheirubiquity,

considerableconfusionsurroundsvariousapproachesforaddressingsampleselection.The

mostcommonapproachfordealingwithselectionbiasincriminologyremainsHeckman’s

[(1976)AnnEconSocialMeasure5:475–492]two-stepcorrection.Thistechniquehas

oftenbeenmipliedincriminologicalresearch.Thispaperhighlightssomecommon

problemswithitsapplication,includingitsusewithdichotomousdependentvariab,

difficultieswithcalculatingthehazardrate,misestimatedstandarderrorestimates,and

collinearityweenthecorrectiontermandotherregressorsinthesubstantivemodelof

interest.Wealsodiscussthefundamentalimportanceofexclusionrestrictions,ortheo-

reticallydeterminedvariabthataffectselectionbutnotthesubstantiveproblemof

interest.Standardstatisticalsoftwarecanreadilyaddresssomeofthesecommonerrors,but

therealproblemwithselectionbiasissubstantive,nottechnical.Anycorrectionfor

selectionbiasrequiresthattheresearcherunderstandthesourceandmagnitudeofthebias.

Toillustratethis,weapplyadiagnostictechniquebyStolzenbergandRel[(1997)Am

SociolRev62:494–507]tohelpdevelopintuitionaboutselectionbiasinthecontextof

criminalsentencingresearch.OurinvestiionsuggeststhatwhileHeckman’stwo-step

correctioncanbeanappropriatetechniqueforaddressingthisbias,itisnotamagic

solutiontotheprob