Minimum Wage Effects of Employment and School Enrollment: Reply to Evans and Turner

MINIMUMWAGEEFFECTSONEMPLOYMENTAND SCHOOLENROLLMENT:REPLYTOEVANSANDTURNER DavidNeumarkandWilliamWascher* July 1996 * NeumarkisProfessorofEconomicsatMichiganStateUniversityandaResearch AssociateoftheNBER. WascherisSeniorEconomistattheBoardofGovernorsof theFederalReserveSystem.WethankWilliamEvansandMarkTurnerforsharing theirdata. TheviewsexpresseddonotnecessarilyreflectthoseoftheFederal ReserveBoardoritsstaff.

MINIMUM WAGE EFFECTS ON EMPLOYMENT AND SCHOOL ENROLLMENT: REPLY TO EVANS AND TURNER ABSTRACT In earlier work, we presented results suggestingthat minimum wage increases have importantconsequences for both the employment opportunitiesof youths and their decision to enroll in school. In this paper, we show that the recent claim made by William Evans and Mark Turner that our results are sensitiveto changes in the definition of the enrollment rate is based upon an analysis that uses a mismeasured minimum wage index. When the data are constructed properly, our original conclusionsare not affected by changes in the enrollment definition.

I. Introduction Inanearlierpaper(NeumarkandWascher,1995a),wepresentedresultssuggestingthat minimumwageincreasesmayhaveimportantconsequencesforboththeemploymentopportunities ofyouthsandtheirdecisiontoenroll(orremainenrolled)inschool. Inparticular,using aconditionallogitmodelofalternativeemploymentandenrollmentoutcomesappliedtostateleveldatafrom1977to 1989,wefoundthatalthoughthenetdisemploymenteffectsfrom minimumwagesaresmallforteenagersasagroup,thereisanoticeabledeclineinthe teenageschoolenrollmentrateandasign~lcantincreaseintheproportionofteenagerswho areneitherinschoolnoremployed.Wefurtherarguedthatourresultswereconsistentwith thehypothesisthatthereisashiftinlabordemandtowardhigher-productivityteenagers afteraminimumwageincrease,anassertiongivenstrongersupportfromsimilarresearchwe conductedusingmicro-leveldatafrommatchedCPSpanels(NeumarkandWascher,1995b). Recently,WilliamEvansandMarkTurner(1995)arguedthatourresultsusingthestatelevelaggregatedataaresensitivetochangesinthedefinitionofschoolenrollment.In particular,theyclaimthatthenegativeeffectsoftheminimumwageonschoolenrollment becomestatisticallyinsignificantwhentheyswitchfromanenrollmentmeasurebasedonmajor activity(suchasweused)toameasurethatismeasuredindependentlyofemploymentandthus mayalsoincludepart-timeschoolactivity. EvansandTurnerfurtherclaimthatestimated minimumwageeffectsfromtheemploymentequationsestimatedinNeumarkandWascher(1992,

-2- 1994)alsobecomeinsignificantwhentheyreestimatedtheequationsusingtheiralternative definitionofenrollment. Ourfirstinclinationwithrespecttotheenrollmentresultsistoquestionthe assertionthattheexpandeddefinitionofenrollmentusedbyEvansandTurnerispreferable tothemeasurebasedonreportedmajoractivitythatwehadused. Thatis,whiletheir definitionofenrollmentobviouslyismoreinclusivethanours,weareskepticalthatan enrolledteenagerwhoreportshismajoractivityasworkshouldbeconsideredinthesame groupasonewhoreportshismajoractivityasschool. Thisturnsout however,tobeonlya minorissue. WhatdrivesEvansandTurner’sresultsisnotthespecificdefinitionofschool enrollmentthattheyuse,butrathertheir(rnis)useofourminimumwagevariable. II.Reassessin~theEvidence Togenerateestimatesoftheirenrollmentratebackto 1978,EvansandTurnerreliedon theenrollmentsupplementsincludedintheOctoberfilesoftheCurrentPopulationSurvey (CPS)ratherthanthebasicMayfilesthatwehadusedinourearlierpapers. Astheynote, onlytheOctobersurveyincludedanindependentquestiononschoolenrollmentthatfarback. Consistentwiththischangeinmonths,theyalsoconstructedfromthebasicOctoberCPSfiles newvaluesfortheprime-agedmaleunemploymentrateandthefractionoftheworking-age populationbetween16and 19yearsofage. However,theycontinuedtouseourminimumwage variable(fromthedatasetthatwesuppliedtothem),whichwasexplicitlyconstructedto measurethecoverage-adjustedleveloftheminimumwageinMayofeachyearrelativetothe

-3averagewage. ThisminimumwagevariableisinconsistentwiththeiruseofOctoberdatafor theothervariablesfortworeasons. First,somestatesraisedtheirminimumwagebetween MayandOctoberofparticularyears. Second,giventhatnominalwageshavegenerallyrisen overtime,thedenominatorgenerallyunderstatestheaveragewagebyanamountthatlikely variesbystateandyear. EvidenceontheMismeasurementoftheMinimumWageVariable ThetoppanelofTable1documentsthestatutoryincreasesinstateminimumwagelaws misclassifiedbyEvansandTurnerbecauseoftheiruseoftheMayminimumwageindex;given thedefinitionoftheminimumwageindexweused(themaximumofthestateorfederaltinirnum wagemultipliedbyfederalcoverageinthestateanddividedbytheaveragewageinthe state),therelevantmisclassificationsareforstateswithaminimumwagehigherthanthe federalminimum.l Inall,thereare17suchminimumwagechanges,allofwhichoccurred inthemid-tolate-1980s.Anumberofthesereflectaseriesofsmallchangesinseveral NewEnglandstates. However,therealsoweresomelargechangesthatoccurredbetweenMay andOctober,includinga90centperhourincreaseinCalifornia’sminimumwageinJuly 1988, twoincreasesinConnecticuttotaling88centsperhourinOctober1987and1988,anda50 1.Thedatasetalsoincludesfederalminimumwageincreasesin1979,1980,and1981, which-astheyoccurredonJanuary1ofeachyear--arecorrectlyclassifiedbyEvansand Turner.Howeverb,ecauseallofthespecificationisncludeyeardummyvariablesi,tislikely thatmuchoftheidentifyinginformationwithrespecttovariationinthenominalminimumwage comesfromthestateminimumwageinc~ases.

-4centincreaseinOregon’sminimumwageinSeptember1989.Moreimportantly,overthissample period,therewere34statutorystateminimumwageincreases,sothatEvansandTurner misclassifythetimingofexactlyhalfofthese. (WeexcludedtheDistrictofColumbiafrom thislist--despitestatutorychangesintheminimumwageforsomeoccupations--becausethe minimumwageforD.C.inourpreviouspaperswasconstructedasaweightedaverageof occupation-specificminimumwagesandthussomeofthevariationintheminimumwagereflects changesinthedistributionofemploymentacrossoccupations.)Ingeneral,wewouldexpect thismisclassificationtobiastheestimatedeffectstowardzero,sincethemisclassified minimumwageincreasesareassignednottotheyearinwhichtheyoccurred,buttothe followingyear. WealsoexaminedEvansandTurner’smismeasurementoftheaveragewage,bycomparing estimatescalculatedfromtheOctoberCPSfilestothosefromtheMaydata(PanelB). As EvansandTurnerindicate,wagedataforOctoberareonlyavailablebeginningin 1979,andso thesecomparisonsarebasedonaslightlyshortersampleperiodthanweusedinouroriginal paper(andEvansandTurnerusedintheircomment).Ascanbeseeninthefirstcolumn, , usingtherawdata,thecorrelationbetweentheMayandOctoberwageisrelativelyhigh (0.92),andonlyabout13percentofthestate-yearobservationsdisplayadifferently-signed 12-monthchangeinthewage. However,oncestateandyearmeansareremovedfromthedata (whichoccursintheregressionanalysis),thesimilarityinthetwowagemeasuresdrops markedly.Asshownintherightcolumn,thecorrelationcoefficientforthewagelevels

-5fallsto0.29,andtheproportionofobservationsshowingdifferently-signedchanges increasesto0.40. Thecombinedeffectsofthemisclassflcationofstateminimumwagechangesandthe mismeasurementoftheaveragewagearepresentedinPanelC. Intherawdata,thelevelsof theMayandOctobercoverage-adjustedrelativeminimumwagedisplayacorrelationcoefficient of0.81,while28percentoftheobsemationsshowadifferently-signedchangeovertheyear. men stateandyeareffectsareremoved,thecorrelationcoefficientforthelevelsfallsto 0.28,andtheproportionofmis-signedchangesrisestonearlyone-half.Thus,itisclear hornthedatathatthemeasurementerrorsassociatedwithEvansandTurner’suseoftheMay minimumwagevariablewithOctoberemploymentdatacouldleadto spuriousdifferencesin results. Implicationsfor Employment-EnrollmenOt utcomes Toassesstheimpactofmismeasuringthechangesintheminimumwagevariableonthe multinominalolgitmodelofemploymentandenrollmentoutcomes,wereestimatedthemodelusing anewlyconstructedminimumwagevariablethatisspecifictoOctoberofeachyear,sothat-incontrasttoEvansandTurner--allofthedatarefertothesamemonth. Asinourearlier paper(andasrepeatedbyEvansandTurner),weuseafour-categorymodelthatclassifies teenagersasinoneoffourmutuallyexclusivestates:inschoolandnotemployed,inschool andemployed,notinschoolandemployed,andnotinschoolandnotemployed.Inaddition, wereportresultsseparatelyusingtheGLSestimatorthatweemployedinNeumarkandWascher

-6- (1995a)andusingtheOLSestimatorreportedbyEvansandTurner. Finally,foreaseof interpretation,wehavetranslatedthemultinominallogitcoefficientsintoelasticities, evaluatedatthesamplemeansofthedata. ThebasicresultsregardingenrollmentandemploymentarepresentedinTable2. In PanelA,wereporttheresultsfromNeumarkandWascher(1995a).Theseresultsindicatea negativeeffectofminimumwagesontheproportionenrolledinschoolandemployed,although thiseffectissignificantonlyusingtheGLSestimatorandeventhen,onlyatthetenpercentlevel. Theresultsalsoindicateastatisticallysignificantpositiveeffectofthe minimumwageontheproportionofteensneitherinschoolnoremployed(i.e.theproportion “idle”). InPanelB,wereportresultsusingthemixedOctoberandMaydatawiththetwo enrollmentmeasures,repeatingtheOLSresultsreportedbyEvansandTurner(rows4and6), aswellasourreestimatesusingthelessrestrictivecovariancematrixthatpermits heteroscedasticerrorsacrossstatesandfwst-orderautocorrelationintheresidualswithin states. Usingtheenrollmentsupplementschoolingmeasure(rows5and6),theestimated effectsonenrollmentandidlenessaremuchsmallerthanwhentheMaydataareused, especiallyascomparedwiththeestimatesinrow 1;thisresultisevidentwitheither estimator. Incontrast,asshowninrow4,thepositiveandsignificantminimumwageeffect onthepercentageofteensinthenotenrolled,notemployedcategoryinthemodelsusing majoractivitytoclassifyteensbyenrollmentstatusisstatisticallysignificantwhenthe

-7- OLSestimatorisused(EvansandTurnerdidnotfocusonminimumwageeffectsonidleness, whichweviewasperhapsthemostimportantfindinginouroriginalpaper). Whentheerrors arepermittedtoexhibitserialcorrelationandheteroscedasticityh, owever,theestimates forthemodelsusingmajoractivityaresimilartothoseusingtheindependentenrollment question,andmoreimportantly,neithersetofresultsissimilartothoseinpanelA. This pattern,byitself,raisesquestionsaboutEvansandTurner’sclaimthatusingtheir alternativeenrollmentmeasure,ratherthanusingOctoberdata,isthesourceofthe differencebetweentheirestimatesandthoseinNeumarkandWascher(1995a). InPanelC,wereporttheminimumwageelasticitiesfromthemodelreestimatedwithdl data--includingtheminimumwageindex--measuredasofOctoberofeachyear. Becausethe averagewageusedintheminimumwageindexisavailablebeginningonlyin 1979,thesample periodfortheseestimatesistwoyearsshorterthaninpanelsAandB. Nonetheless,the differencesbetweenpanelsBandC--andthesimilaritiesbetweenpanelsAandC--are striking. ConsistentwithEvansandTurner’sclaim(basedontheresultsinrow4ofPanel B),theestimatesusingthemajoractivityvariabletoclassifyenrollmentstatus(rows7 and 8)yieldresultsverysimilartothosewereportedinourearlierpaper(althoughthe negativeelasticityfortheproportionenrolledandemployedissmaller).Incontrastto theirresults,however,theestimatesusingtheenrollmentsupplementtomeasureschooling alsoareverysimilartoourpreviousestimates(rows9and 10). Inparticular,theminimum wageisestimatedtohaveapositiveandsignificanteffectontheproportionofteenagers

-8neitherinschoolnoremployed,regardlessofthemonth,enrollmentmeasure,orcovariance matrixusedinthemodel,aslongasthedataaremeasuredinatime-consistentmanner. Asinouroriginalpaperonthistopic,thestatisticalsignificanceofthenegativeeffects ofminimumwagesonenrollmentisnotrobust butthishaslittle,ifanything,todowith thealternativeenrollmentmeasures.Moregenerally,theseresultsstronglysuggestthatthe differentfindingsthatEvansandTurnerreportarenotduetotheiruseofabroadermeasure ofenrollment,astheyclaim,butrathertotheirfailuretoaccuratelymeasurethetiming andmagnitudeofchangesintheminimumwageindex. Wefurtherillustratethisfindingbyhighlightinginbold,inTable2,the(preferred) GLSestimatesfromouroriginalpaper(row1),theestimatesemphasizedbyEvansandTurner usingthemixedMay/Octoberdataandtheenrollmentsupplement(row6),andtheGLSestimates usingtheenrollmentsupplementandconsistentdataforOctober(row9). Wethinkitisfair tosaythataslongasconsistentdataareused,theresultsarerelativelyinsensitiveto theenrollmentmeasure,andareasweoriginallyreported. Inadditiontoreestimatingthemultinominalolgitmodel,EvansandTurnerreport resultsfromasetofsimpleenrollmentregressionsrelatingthefractionofteenagers enrolledtothesamesetofindependentvariablesusedinTable2. Thisspecification,which wereproduceintable3,providesasimplertestoftheeffectsofminimumwageson enrollment.Basedontheevidencepresentedintheright-handcolumnofpanelsAandB,they againarguethattheresultsarenotrobusttothechoiceofenrollmentmeasure. In

-9particular,whentheyusetheenrollmentsupplementmeasureofschoolingwiththeirmixed October/Maydata,thecoefficientontheminimumwagevariableissmallandinsignificant (row3),incontrasttotheirresultsusingthemajoractivitymeasureofenrollment(row2). However,theirresultobtainsonlyundertheassumptionofseriallyindependentand homoscedasticerrors. Asshownintheleft-handcolumnofpanelB,whenwerelaxthese restrictionsonthecovariancematrix,theeffectofminimumwagesonschoolenrollmentare significantlynegativeandofamagnitudesimilartothoseinpanelA,regardlessofthe enrollmentmeasureused. Itdoesappear,however,thattheresultsfortheenrollment equationsaresomewhatsensitivetochangesinthesample. Whenalldataarebasedon Octobervalues,theminimumwagecoefficientdropssubstantiallyintheequationsusingthe majoractivitymeasureofenrollment(row4),butincontrasttotheclaimsmadebyEvansand Turner(basedonthemixedMay/Octoberdata),theestimatedeffectofminimumwageson enrollmentisstronger(andsignificantatthefive-percentlevel)usingtheenrollment supplementmeasureofschooling(row5). (AsinTable2,wehavehighlightedthemost pertinentestimates.)Again,itisthemixingofMayandOctoberdata,ratherthanthe differentenrollmentmeasure,thatunderliesEvansandTurner’sresults. Implicationsfor theEmploymentEffectsofMinimumWages ThefinalsetofresultspresentedbyEvansandTurnerreexaminetheemployment equationsforteenagersandyoungadultsthatwepresentedinouroriginal1992paperandin our1994replytothecommentbyCard,Katz,andKruege](1994). Asinthecaseofthe

-1oenrollmentequations,EvansandTurnercompareourresultsusingtheMaydatatothoseusing acombinationofMayandOctoberdataandarguethatthenegativecoefficientsontheminimum wagevariablethatwereportedintheseearlierpapersarenotrobusttotheinclusionor definitionoftheenrollmentrate. AswehadreportedinNeumarkandWascher(1995a)--andasshowninrow 1ofPanelAof Table4--thereislittleevidenceofasigntilcantnegativeemploymenteffectfromminimum wagesintheteenagespecificationsthatexcludetheenrollmentrate,aresultthatmotivated ourresearchonjointenrollmentandemploymentdecisions.Aswehavesuggestedelsewhere, however,negativeemploymenteffectsofminimumwageswhenwecontrolforenrollmentare impliedbythehypothesisthatminimumwageincreasesinduceemployerstosubstitutetowards enrolledteenagers(NeumarkandWascher,1995c).Thus,takenatfacevalue,Evansand Turner’sresultsappeartounderminethishypothesis.However,asismadeclearbythe resultsinPanelB(whichonlyreportresultsfromEvansandTurner’spaper),thesmalland insignificanteffectsofminimumwagesfoundbyEvansandTurnerusingvariousalternative definitionsoftheenrollmentrateareonceagainduetotheirmixingofMayandOctoberdata ratherthantothechoiceofenrollmentmeasure. Inparticular,asshowninrow3,whenthe narrowestdefinitionoftheenrollmentrate(basedontheemploymentstatusrecode)is includedintheregression,theestimatedminimumwageeffectsarenegativeandstatistically significantusingconsistentdataforMay. However,whentheOctoberdataareusedwiththe Mayminimumwagevariable,theseestimatedeffectsbecomesmallandinsignificant(row4),

-11andaresimilartotheresultsobtainedusingtheenrollmentsupplementmeasureofschooling (row6). Moreover,ascanbeseeninPanelC,usinganOctober-basedminimumwageindexlargely reversesEvansandTurner’sresults. Intheteenagespecifications,theminimumwageeffect isofthesamemagnitudeandisstatisticallysignificantwithap-valueoflessthan0.1 regardlessoftheenrollmentdefinition;thecoefficientsandstandarderrorsarealsoquite similarintheequationsforyoungadults(withp-valuesclosetoO.1),althoughthe estimatesinpanelCaresmallerthaninpanelA. Indeed,inthissense,theresultsusing theOctoberdataprobablyaremoreintuitivelyappealingthanthoseusingtheMaydata,as theminimumwageeffectisuniformlylargerforteenagersthanforyoungadults. Asinthe previoustables,wehavehighlightedthepreferredestimatesfromourearlierwork(row2), thoseemphasizedbyEvansandTurner(row6),andthoseusingtheenrollmentsupplementwith consistentOctoberdata(row9). Onceagain,theresultsusingeitherenrollmentrateare quitesimilaraslongasconsistentdataareusedandruncountertotheconclusionsreached byEvansandTurnerusingthemixedMay/Octoberdata. Thus,theevidencesuggeststhatthesensitivityofminimumwageeffectsreportedby EvansandTurnerstemsfromusingamonthinwhichtomeasuretheminimumwagevariablethat differsfromthemonthforwhichthedependentvariable(andtheotherindependentvariables) aremeasured.Thismayseemsurprisinggiventhenatureofthemodels,asthespecifications allowforbothcontemporaneousminimumwageeffectsandeffectslaggedoneyear,andtheuse

-12ofMay--ratherthanOctober--wagedataonlyshiftsthetimingoftheeffectsbysixmonths. However,thisshiftintimingcouldbequiteimportantifthecontemporaneousminimumwage effectsaremostprominent,oriflaggedeffectsarestrongestatlengthsofaboutoneyear. SourcesoftheInstabilityintheEvam-TurneRr esults Finally,wecanexaminewhethertheinstabilityintheestimatesusingMay/Octoberdata ismorecloselyassociatedwithEvansandTurner’smisclassificationofminimumwage increasesorwiththeirrnismeasurementoftheaveragewage. Table5presentsestimatesof equationsbasedontheOctoberenrollmentsupplementusingdifferentconstructsofthe minimumwagevariable,alternativelyallowingthenumeratoranddenominatoroftheindexto bemismeasured.TurningfrosttothemultinominalolgitresultsshowninPanelA,itis evidentthatboththemisclassificationofminimumwagechangesandthemismeasurementofthe averagewagemattertosomeextent. Inparticular,correctingindividuallyforthe mismeasurementsi,nbothcases,leadstoanincreaseintheestimatedeffectofminimumwages onidleness,withtheeffectbecomingstatisticallysignificant(comparethesecondorthird rowtothefirstrow). However,itisalsoclearthatthemismeasurementoftheaveragewage isthemoreimportantsourceoferror. Forallfourenrollment-employmenctategories,the estimatesusingtheMayminimumwagewiththeOctoberaveragewagearequiteclosetothe estimatesusingOctoberdataforboththenumeratoranddenominator.Thissamepatternis apparentintheemploymentequations(PanelC). Incontras~intheschoolenrollment

-13equations(PanelB),theestimatesaresensitivetoboththemisclassificationoftheminimum wagechangesandthemismeasurementoftheaveragewage. III. conclusion Insum,areexaminationoftheevidencepresentedbyEvansandTurnerindicatesthat thereisnoreasontobelievethatourearlierestimatesoftheeffectsofminimumwageson theenrollmentandemploymentoutcomesofteenagerswereundulyinfluencedbyouruseofthe majoractivitydefiition ofschoolenrollmentratherthanthemoreinclusivedefinition availableintheOctoberCPSenrollmentsupplement.Instead,thesmallandstatistically insignificantminimumwageeffectsproducedbyEvansandTurnerresultfromtheir mismeasurementoftheminimumwageindexusedinthesestudies.Wethusstandbyour originalconclusionsthatincreasesintheminimumwageresultinsubstitutionbyemployers towardenrolledteenagersandinasignificantincreaseintheproportionofnonenrolled teenagerswithoutajob.

References Card,David,AlanKrueger,andLawrenceKatz.1994.“EmploymentEffectsofMinimum andSubminimumWages:Comment.”IndustrialandLabor RelationsReview.Vol. 47,NO.3,pp.487-497. Evans,WilliamandMarkTurner.1995.“MinimumWageEffectsonEmploymentand SchoolEnrollment:Comment.”Mimeo.UniversityofMaryland. Neumark,DavidandWilliamWascher.1992.“EmploymentEffectsofMinimumand SubminimumWages:PanelDataonStateMinimumWageLaws.”Industrialand LaborRela~“ensReview,Vol.46,No.1,pp.55-80. . 1994.“EmploymentEffectsofMinimumand SubminimumWages:ReplytoCard,Katz,andKrueger.”IndustrialandLabor RelationsRevie~. Vol.47,No.3,pp.497-512. . 1995a.“MinimumWageEffectsonEmploymentand SchoolEnrollment.”JoumalofBusinessandEconomicsStatistics.Vol.13, No.2,pp. 199-208. . 1995b.“TheEffectsofMinimumWageson TeenageEmploymentandEnrollment:EvidencefromMatchedCPSSurveys.” ResearchinLaborEconomics.forthcoming. . 1995c.“ReconcilingtheEvidenceonEmployment EffectsofMinimumWages:AReviewofourResearchFindings,”Financeand EconomicsDiscussionSeriespaperno.95-53,FederalReserveBoard, December,1995.

Table1:ConsequenceosfUsingMayMinimumWageVariableforOctober A. Misclassified Minimum Wa~e Increases State MinimumWageIncrease DateofIncrease Massachusetts $.20 July1,1986 RhodeIsland $.20 July 1,1986 Vermont $.10 July1,1986 Massachusetts $.10 July1,1987 RhodeIsland $.10 July 1,1987 Vermont $.10 July 1,1987 Connecticut $.38 October1,1987 California $.90 July 1,1988 Massachusetts $.10 July1,1988 RhodeIsland $.35 July1,1988 Vermont $.10 July2, 1988 Connecticut $.50 October1,1988 Wisconsin $.30 July1, 1989 Vermont $.10 July2, 1989 RhodeIsland $.25 August1,1989 NorthDakota $.05 August14,1989 Oregon $.50 September1,1989 Totalnumberofstateminimumwageincreases,May1978-October1989: 34 Numberincorrectlyassignedtoyearafierincrease: 12 Numberincorrectlyomitted: 5 B. Mismeasured Average Wape RawData NetofYearandStateEffects CorrelationbetweenMay andOctoberaveragewage .92 .29 Proportionofobservations withdifferent-signedMay-May andOctober-Octoberchanges inaveragewage .13 .40 C. Overall Effects onCoverape- Adiusted Relative Minimum Wape RawData NetofYearandStateEffects CorrelationbetweenMay andOctober .81 .28 Proportionofobservations withdifferent-signedMay-May andOctober-Octoberchanges .28 .45 PanelAexcludesWashington,D.C. Changesinstateminimumwagesbelowthefederal minimumwagearealsoignored.Notethattheminimumwageincreasesoccuringin 1989 werenotincludedinthedatasetusedinNeumarkandWascher(1995a,1992,1994), becausethosedataextendedthroughMayof 1989.However,theseincreasesoccurredby October1989,andhenceshouldhavebeenincludedinthedatausedbyEvansandTurner.

Table2:EstimatedMinimumWageEffects(CurrentandLagged)onEmployment andEnrollmenSt tatus,ElasticitiesfromConditionaLlogitEstimates,16-19Year-Olds Enrollment Enrolled, Enrolled, NotEnrolled Not Enrolled, Measure NotEmr)love~ ~mt)love~ Emn1oye~ NotEmDloyed A.Earlier Results 1978-1989,all data hornMay: 1)AR(l), heteroscedastic Major -.02 -.47* .14 .64** errors activity (.18) (.29) (.26) (.15) 2)Independent, Major -.15 -.17 .12 .60** homoscedastiecrrors activity (.24) (.38) (.12) (.20) B. Resultswith Mixed Mav/OctoberData 1978-1989,enrollment andemploymentdata fromOctober,Minimum wageindexfromMay: 3)AR(l),heteroscedastic Major .00 -.40 .28 .16 errors activity (.21) (.33) (.29) (.17) 4)Independent, Major -.05 -.54 .35 .46** homoscedasticerrors activity (.27) (.40) (.39) (.22) 5)AR(l), heteroscedastic Enrollment .03 -.13 .04 .15 errors supplement (.21) (.29) (.34) (.17) 6)Independent, Enrollment -.02 -.04 -.07 .33 homoscedasticerrors supplement (.27) (.37) (.43) (.22) C New Results 1680-1989,all data fromOctober: 7)AR(l), heteroscedastic Major -.04 -.19 -.09 .74** errors activity (.14) (.22) (.23) (.12) 8)Independent, Major .07 -.28 -.06 .41** homoscedasticerrors activity (.22) (.33) (.33) (.18) .64** 9)AR(l), heteroscedastic Enrollment -.12 -.33 .43* errors supplement (.16) (.23) (.26) (.13) 10)Independent, Enrollment .05 -.24 -.08 ,54** homoscedasticerrors supplement (.22) (.31) (.37) (.19) ‘*’and‘**’indicatestatisticallysignificantat the 10and5 percentlevels,respectively. Earlierresul@arefromNeumark andWascher(1995a). Theestimatesinrow 1correcta slightprogrammingerrorin thatpaperthatBillEvansandMark Turnerpointedout;fie resultsarequalitativelysimilar. TheestimatesinPanelBarebasedonourspecificationussingtie datainEvansandTurner(1995).

Table3:fitimated MinimumWageEffects(CurrentandLagged)on EnrollmentRates, 16-19Year-Olds %In School Enrollment AR(l), heteroscedastic Independent,homoscedastic Measure - - ~. Earlier Results 1978-1989,alldata fromMay: -.26** 1) Major -.27* activity (.12) (.15) B. Results with Mixed avlOctober Data 1978-1989, enrollment andemploymentdata from October,Minimum wage indexfromMay: -*33** 2) Major -.25** activity (.13) (.15) Enrollment -.20* -.08 supplement (.11) (.13) C. New Results 1980-1989, all data from October: 4) Major -.06 -.07 activity (.10) (.13) -.18** 5) Enrollment -.09 supplement (.09) (.11) ‘*’and ‘**’indicatestatisticallysignificantatthe10and5 percentlevels,respectively.Earlierresultsare basedon thedatain NeumarkandWascher(1995a). .,.-

Table4: EstimatedMinimumWageEffects(CurrentandLagged)on EmploymentRates 16-19Year-OldS 16-24Year-Olds Enrollment Enrollment Enrollment Enrollment Enrollment Measure Excluded H Excluded Included A.Earlier Results 1973-1989,alldata fromMay: 1) Employment -.04 ..24** -.29** -.28** statusrecode (.13) (.08) (.09) (.07) 2) Major ... -.13 ... -.27** activity (.11) (.08) B. Evans/Turner Results 1978-1989,alldata horn May: 3) Employment -.06 -.23** -.16 ..24** statusrecode (.15) (.10) (.10) (.08) 1978-1989,enrollment andemploymentdata horn October,Minimum wageindexfromMay: 4) Employment -.08 -.13 -.14 -.06 statusrecode (.14) (.10) (.10) (.08) 5) Major .*. -.18* ... -.07 activity (.13) (.09) 6) Enrollment ... -.09 ... -.04 supplement (.14) (.10) C. NewResult$ 1980-1989,alldata fromOctober: .,15* .,. 7) Employment -.19 -.17** -,13* .. ! 4 statusrecode (.12) (.08) (.09) (.07) 8) Major ... ..21* -.13 activity (.11) ““” (.08) 9) Enrollment ... .*22* -*14* supplement (.12) ““” (.08) ‘*’and‘**’indicatestatisticallysignificantat the 10and5 percentlevels,respectively. EarlierresultsarefromNeumark andWascher(1992,1994). EvansandTurnerresul~ arefromEvansandTurner(1995). In PanelC,thedatabeginin 1980becausewagedatain Octoberarefirst availablein 1979.

Table5:Effectof MisclassifiedMinimumWagesandMismeasuredAverageWages, UsingEnrollmentSupplement A. Table 2 Results MinimumWage Enrolled, Enrolled, NotEnrolled NotEnrolled, Variable NotEmt)loYe~ EmDlove~ EmD1oved NotEmDloYed May minimum, .03 -.13 .04 .15 May average(Row 5) (.21) (.29) (.34) (.17) Octoberminimum, -.02 -.03 -.10 .27* May average (.20) (.29) (.31) (.17) May minimum, -.10 -.34 .40 .64** Octoberaverage (.18) (.25) (.29) (.15) Octoberminimum, -.12 -.33 .43* .64** Octoberaverage(Row9) (.16) (.23) (.26) (.13) B. Table 3Resultq MinimumWage - % In School Mayminimum, -.20* May average(Row 3) (.11) -*33** Octoberminimum, Mayaverage (.10) May minimum, -.12 Octoberaverage (.10) Octoberminimum, -.18** Octoberaverage(Row 5) (.09) c. Table 4 Resultq MinimumWage 16-19Year-Olds, 16-24Year-OIds, Variable EmD1ovmen~ EmDloYmen~ Mayminimum, -.09 -.04 May average(Row6) (.14) (.10) Octoberminimum, -.04 -.02 Mayaverage (.15) (.10) -*21* Mayminimum, -.13 Octoberaverage (.12) (.08) -,22* .*14* Octoberminimum, Octoberaverage(Row9) (.12) (.08) Alldataothertian relativeminimumwagevariablearebasedon OctoberCPS’S.Allspecificationsuse enrollmentsupplementto:measureenrollment,andallowAR(l), heteroscedasticerrors. ‘*’and ‘**’ indicatestatisticallysignificantatthe 10and5 percentlevels,respectively.