Does Going Public Affect Innovation

Accepted ArticleDoesGoingPublicA?ectInnovation?

SHAIBERNSTEIN?

ABSTRACT

Thispaperinvestigatesthee?ectsofgoingpubliconinnovationbycomparingtheinnovationactivityof?rmsthatgopublicwith?rmsthatwithdrawtheirIPO?lingandremainprivate.NASDAQ?uctuationsduringthebook-buildingphaseareusedasaninstrumentforIPOcompletion.Usingpatent-basedmetrics,I?ndthatthequalityofinternalinnovationdeclinesfollowingtheIPO,and?rmsexperiencebothanexodusofskilledinventorsandadeclineintheproductivityoftheremaininginventors.However,public?rmsattractnewhumancapitalandacquireexternalinnovation.Theanalysisrevealsthatgoingpublicchanges?rms’strategiesinpursuinginnovation.

StanfordUniversity.IamdeeplygratefultoFritzFoley,JoshLerner,AndreiShleifer,andJeremySteinfortheirinvaluablecomments.IalsothankPhilippeAghion,E?Benmelech,LauraField,PaulGompers,RobinGreenwood,SamHanson,OliverHart,VictoriaIvashina,DirkJenter,BillKerr,JacobLeshno,GustavoManso,RamanaNanda,FranciscoPerez-Gonzalez,DavidScharfstein,AntoinetteSchoar,AmitSeru,AdiSunderam,RickTownsend,andJe?Zwiebelforhelpfulcomments.IamgratefultoseminarparticipantsatChicagoUniversity,ColumbiaUniversity,DartmouthCollege,EntrepreneurialFinanceandInnovationConference,HarvardBusinessSchool,HebrewUniversity,LondonBusinessSchool,LondonSchoolofEconomics,NBERProductivityLunch,NorthwesternUniversity,SearleConferenceonInnovationandEntrepreneurship,StanfordUniversity,Tel-AvivUniversity,UniversityofBritishColumbia,andUniversityofPennsylvaniaforhelpfulcommentsandsuggestions.AndrewSpeenprovidedsuperbresearchassistance.IamgratefulforfundingfromtheEwingMarionKau?manFoundation.

?

This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1111/jofi.12275.

This article is protected by copyright. All rights reserved.

Doesthetransitiontopublicequitymarketsa?ectinnovation?Thisquestionisimportantgiventhecriticalroleofinnovationinpromotingeconomicgrowth(Solow(1957))andtheprevalenceoftechnological?rmsintheinitialpublico?ering(IPO)marketoverrecentyears.1Althoughalargebodyofresearchexaminestheperformanceof?rmsaroundtheirIPO,littleisknownaboutthee?ectsofgoingpubliconinnovation.Thispaper’smaincontributionistoshowthatgoingpublica?ectsthreeimportantdimensionsofinnovationactivity:thecreationofinternallygeneratedinnovation,theproductivityandmobilityofindividualinventors,andtheacquisitionofexternalinnovation.

Theoretically,infrictionless?nancialmarkets,sellingequitiespubliclyshouldhaveno

Accepted Articlebearingonsubsequentinnovationactivity.Under?nancialfrictions,however,thetransitiontopublicequitymarketsimproves?rms’accesstocapital,whichcanleadtoanincreaseininnovationactivity,becausesuchactivityisparticularlylikelytobesensitiveto?nancingconstraints(Arrow(1962),HallandLerner(2010)).Ontheotherhand,agencyproblemsassociatedwiththetransitiontopublicequitymarketsmayundermine?rmincentivestoinnovate(BerleandMeans(1932),JensenandMeckling(1976)).

Ultimately,empiricalinvestigationisnecessarytodisentanglethepositiveandadverse

e?ectsofpubliclistingoninnovation.Estimatingthesee?ects,however,ischallengingduetoaninherentselectionbiasassociatedwiththedecisiontogopublic.Astandardapproachtostudyingthedynamicsof?rmoutcomesaroundtheIPOistousewithin-?rmvariation.ButasJainandKini(1994)note,?rmschoosetogopublicataspeci?cstageintheirlifecycle,andasaresultthisapproachmayproducebiasedestimatesoftheIPOe?ect.Forinstance,if?rmschoosetogopublicfollowinganinnovationbreakthrough,asarguedbyP′astor,Taylor,andVeronesi(2009),2thenpost-IPOperformancemayre?ectreversiontothemeanandtherebymixlifecyclee?ectswiththeIPOe?ect.

ToovercomethisselectionbiasandestimatetheIPOe?ectoninnovation,Iconstruct

asampleofinnovative?rmsthat?leaninitialregistrationstatementwiththeSecuritiesandExchangeCommission(SEC)inanattempttogopublic,andtheneithercompleteor

1

This article is protected by copyright. All rights reserved.

withdrawtheir?ling.Combinedwithstandardpatent-basedmetrics,thissampleallowsmetocomparetheinnovationactivityof?rmsthatgopublicwiththatofprivate?rmsthatintendedtogopublicandhenceareatasimilarstageintheirlifecycle.Ofcourse,comparingcompleteandwithdrawnIPO?lingsintroducesanewbiasassociatedwiththedecisionof?rmstowithdrawtheIPO?lingandremainprivate.

Toaddressthisconcern,IuseNASDAQ?uctuationsoverthetwomonthsfollowingthe

Accepted ArticleIPO?lingdateasaninstrumentforIPOcompletion,relyingon?lers’sensitivitytoaggregatestockmarketmovementsduringthebook-buildingphase.Consistentwithpriorliterature,I?ndthattheseshort-runNASDAQ?uctuationsstronglypredictIPOcompletion,withthee?ectconcentratedatmarketdeclines(e.g.,Benvenisteetal.(2003),Busaba,BenvenisteandGuo(2001)).Intheanalysis,theIPOe?ectisidenti?edfromdi?erencesinlong-runinnovation(i.e.,innovationover?ve-yearperiod)between?rmsthat?letogopublicinthesameyearbutexperiencedi?erentpost-?lingNASDAQreturns.

Fortheinstrumenttobevalid,itneedstosatisfytheexclusionrestrictioncondition,

thatis,two-monthNASDAQreturnsmustberelatedtothelong-runinnovationmeasuresonlythroughtheIPOcompletionchoice.I?ndempiricalsupportforthevalidityoftheinstrument(seeSectionII.Cforadetaileddiscussion).Whilereduced-formanalysisshowsthattheinstrumentsigni?cantlyexplainslong-runinnovation,itmayalsoa?ectinnovationthroughotherchannels.Totheextentthatthisisthecase,two-monthNASDAQreturnsshouldalsoexplainlong-runinnovationoutsidethebook-buildingphase.Aplacebotestrevealsthatoutsidethebook-buildingphase,whenownershipchoiceis?xed,NASDAQreturnshavenoe?ectonlong-runinnovation.This?ndingisconsistentwiththenotionthatshort-termNASDAQreturnsduringthebook-buildingphasea?ectlong-runinnovationonlythroughtheIPOcompletionchoice.

Usingtheinstrumentalvariablesapproach,I?ndasigni?cantlinkbetweenpublicowner-

shipandinnovation:goingpubliccausesasubstantial40Tclineininnovationnoveltyasmeasuredbypatentcitations.Atthesametime,I?ndnochangeinthescaleofinnovation,

2

This article is protected by copyright. All rights reserved.

asmeasuredbythenumberofpatents.Theseresultssuggestthatthetransitiontopublicequitymarketsleads?rmstorepositiontheirR&Dinvestmentstowardmoreconventionalprojects.

Havingshownthatgoingpublica?ectsthecompositionofinnovativerelativetoconven-

Accepted Articletionalprojects,Istudythee?ectsofgoingpubliconindividualinventors’productivityandmobilityovertime.I?ndthatthequalityofinnovationproducedbyinventorswhoremainatthe?rmdeclinesfollowingtheIPO,andkeyinventorsaremorelikelytoleave.The?rmsthatgopublicarealsomorelikelytogeneratespin-o?companies,suggestingthatinventorswholeaveremainentrepreneurial.Thesee?ectsarepartiallymitigatedbytheabilityofpublic?rmstoattractnewinventors.

Ialso?ndastarkincreaseinthelikelihoodthatnewlypublic?rmsacquirecompanies

intheyearsfollowinganIPO.Tobetterunderstandwhethertheseacquisitionsareusedtopurchasenewtechnologies,Icollectinformationontargets’patentportfolios.I?ndthatpublic?rmsacquireasubstantialnumberofpatentsthroughM&A:acquiredpatentsconstitutealmostathirdof?rms’totalpatentportfoliointhe?veyearsfollowingtheIPO.TheacquiredpatentsareofhigherqualitythanthepatentsproducedinternallyfollowingtheIPO.

Theseresultssuggestthatthetransitiontopublicequitymarketsa?ectsthestrategies

that?rmsemployinpursuinginnovation.Whilepubliclytraded?rmsgeneratemoreincre-mentalinnovationinternally,theyalsorelymoreheavilyonacquiringtechnologiesexternally.ThisshifttakesplaceconcurrentwithasubstantialinventorturnoveraftertheIPO.Whatleadstothesechanges?Astockmarketlistingmayincreasethescopeforagency

problems,leading?rmstopursuelessinnovationfollowingtheIPO(BerleandMeans(1932),JensenandMeckling(1976)).Alternatively,agencyproblemsmaybeconcentratedinprivate?rmsthatpursuetoomuchinnovation.Finally,itmaybethecasethattheresultsarenotdrivenbyagencyproblems,butrathertheimprovedaccesstocapitalallows?rmstofocusoncommercializationaftertheIPO.Suchastrategyisnotviablefor?rmsthatremainprivate

3

This article is protected by copyright. All rights reserved.

andthereforefocusoninnovationonly.IdiscussthesehypothesesindetailinSectionIV.I?ndsupportiveevidenceformanagerialcareerconcerns,consistentwithrecentpapersonagencyproblemsinpublicequitymarkets(e.g.,Aghion,VanReenen,andZingales(2013),Asker,Farre-Mensa,andLjungqvist(2014),Fismanetal.(2013),andGao,Hsu,andLi(2014)).

Thepaperisrelatedtoseveralstrandsoftheliterature.First,byproposinganidenti?-

Accepted ArticlecationstrategytoisolatetheIPOe?ect,andfocusingon?rms’innovationactivitiesaroundtheIPO,thispapercontributestotheIPOliteraturethatexplores?rmbehaviorfollowingtheIPOanddocumentsadeclinein?rmperformancemeasuressuchaspro?tabilityandproductivity(Chemmanur,He,andNandy(2010),DegeorgeandZeckhauser,(1993),JainandKini(1994),Mikkelson,Partch,andShah(1997),Pagano,Panetta,andZingales(1996,1998),andP′astor,Taylor,andVeronesi(2009)).3

Second,byfocusingontheex-postconsequencesofbecomingapubliclytraded?rm

oninnovation,andillustratingacomplextrade-o?betweenpublicandprivateownershipstructures,thepapercontributestoagrowingbodyofworkthatcomparesthebehaviorofpublicandprivate?rmsalongvariousdimensions,suchasinvestmentsensitivity(e.g.,Asker,Farre-Mensa,andLjungqvist(2014),Sheen(2009)),debt?nancingandborrowingcosts(SaundersandSte?en(2011),Brav(2009)),dividendpayouts(MichaelyandRoberts(2012)),andCEOturnover(Gao,Harford,andLi(2014)).

Thisworkalsocontributestoagrowingliteraturethatexplorestheroleofgovernance,

capitalstructure,andownershiponcorporateinnovation.Forexample,Seru(2014)exploresthee?ectsofmergersandacquisitionsoninnovation,Lerner,Sorensen,andStromberg(2011)examinetheimpactofprivateequityinvestments,andAghion,VanReenen,andZingales(2013)exploretheimportanceofinstitutionalinvestorsandconcentratedownership.Inaddition,thispaperisconnectedtothetheoreticalliteratureontheorganizationofR&D(e.g.,AghionandTirole(1994)),andmorebroadlytoempiricalworkontheboundariesofthe?rm(e.g.,Robinson(2008),Beshears(2013),andSeru(2014)).

4

This article is protected by copyright. All rights reserved.

Therestofthepaperproceedsasfollows.SectionIdescribesthedataandpresentssummarystatisticsandSectionIIpresentstheempiricalstrategy.SectionIIIprovidesthemainresultsandSectionIVdiscussespotentialchannels.SectionVconcludes.

Accepted ArticleI.Data

ThedatainthisanalysiscompriseinformationonIPO?lings,patents,hand-collected

?nancialinformation,andinformationonother?rmcharacteristics.

A.IPOFilings

ToapplyforanIPO,a?rmisrequiredtosubmitaninitialregistrationstatementtothe

SEC(usuallyFormS-1),whichcontainstheIPO?ler’sbusinessand?nancialinformation.FollowingthesubmissionoftheFormS-1,?lersmarkettheequityissuancetoinvestors(thebook-buildingphase)andhavetheoptiontowithdrawtheIPO?lingbysubmittingFormRW.WithdrawalsarecommoninIPOmarkets,withapproximately20%ofallIPO?lingsultimatelywithdrawn.

IcollectallIPO?lingsusingThomsonFinancial’sSDCNewIssuesdatabase.Thesample

startsin1985,whenSDCbegancoveringwithdrawnIPOssystematically,andendsin2003,astheanalysisexplorestheinnovationoutcomesof?rmsinthe?veyearsaftertheIPO?ling.FollowingtheIPOliterature,IexcludeIPO?lingsof?nancial?rms(SICbetween6000and6999),unito?ers,closed-endfunds(includingREITs),ADRs,limitedpartnerships,specialacquisitionvehicles,andspin-o?s.Iidentify5,583completeIPOsand1,599withdrawnIPO?lingsfrom1985to2003.

5

This article is protected by copyright. All rights reserved.

B.PatentData

B.1.

MeasuringInnovativeActivity

Accepted ArticleAnextensiveliteratureontheeconomicsoftechnologicalchangedemonstratesthat

patentingactivityre?ectsthequalityandextentof?rminnovation,andtheuseofpatentingactivityasameasureofinnovationiswidelyaccepted(Hall,Ja?e,andTrajtenberg(2001),Lanjouw,Pakes,andPutnam(1998)).Importantlyforthisanalysis,patentinformationisavailableforbothpublicandprivate?rms,unlikeR&Dexpenditures,andallowstomeasure?rminnovationoutputalongseveraldimensions.

Themostbasicmeasureofinnovationoutputisasimplecountofthenumberofpatents

granted.However,patentcountscannotdistinguishbetweenbreakthroughinnovationandincrementaldiscoveries(e.g.,Griliches(1990)).Asecondmetricthereforere?ectsthenoveltyofapatentbycountingthenumberofcitationsapatentreceivesfollowingitsapproval.Thenumberofpatentcitationsalsocapturestheeconomicimportanceoftheinnovation,asitcorrelateswitha?rm’smarketvalue(Hall,Ja?e,andTrajtenberg(2005)).Indeed,Koganetal.(2012)showthatthestockmarketstrongly,andpositivelyreactstotheapprovalofpatentsthatareeventuallyhighlycited,andthatsuchpatentspredictboth?rmproductivityandcapitalandlabor?owsfromnon-innovativetoinnovative?rmswithintheindustry.Bothcitationratesandpatentcountsvaryovertimeandacrosstechnologies.Variation

maystemfromchangesintheimportanceoftechnologiesorfromchangesinthepatentsystem.Therefore,acomparisonofrawpatentsandcitationsisonlypartiallyinformative.Toadjustpatentcitations,IfollowHall,Ja?e,andTrajtenberg(2001)andconstructScaled

Citationsbyscalingeachpatentcitationcountbytheaveragenumberofcitationsofmatchedpatents,wherematchedpatentsarepatentsgrantedinthesameyearandinthesametechnologyclass.Similarly,toadjustpatentcounts,Iweighteachpatentbytheaveragenumberofpatentsgrantedby?rmsinthesameyearandtechnologyclass.Hence,patentsgrantedintechnologiesinwhich?rmsissuemorepatentsreceivelessweight.TheScaled

6

This article is protected by copyright. All rights reserved.

NumberofPatentsisasimplesumofthescaledpatentsa?rmgeneratesinayear.AsadditionalmeasuresIemployOriginalityandGenerality,whichusethedistributionofcitationstocapturethefundamentalnatureofinnovation(Trajtenberg,Henderson,andJa?e(1997)).Apatentthatcitesabroaderarrayoftechnologyclassesisviewedashavinggreateroriginality,whileapatentthatiscitedbyamoretechnologicallyvariedarrayofpatentsisviewedashavinggreatergenerality.4Similarlytopatentcountsandcitations,ScaledOriginalityandScaledGeneralityarenormalizedbythecorrespondingaverageoriginalityorgeneralityofallpatentsgrantedinthesameyearandtechnologyclass.

B.2.PatentDataSources

ThepatentdatacomefromtheNationalBureauofEconomicResearch(NBER)patent

Accepted Articledatabase,whichincludesdetailedinformationonmorethanthreemillionpatentssubmittedtotheU.S.PatentandTrademarkO?ce(USPTO)from1976to2006(Hall,Ja?e,andTrajtenberg(2001)).

IusetheNBERbridge?letoCOMPUSTATtomatchpatentsto?rmsthatcompleted

theIPO?ling,andmanuallymatchpatentstowithdrawnIPO?lings.5Irestrictthesampleto?rmswithatleastonesuccessfulpatentapplicationovertheperiodfromthreeyearsbeforeto?veyearsaftertheIPO?ling.Thisyields1,488innovative?rmsthatwentpublicand323thatwithdrewtheIPOapplication.

Thegoalistocollectinformationon?rms’innovationactivityinthe?veyearsafterthe

IPO?lingregardlessofwhetherthe?rmsareacquiredorgopublicinasecondattempt,toavoidbiasesthatmayarisefromtruncating?rmactivity.Afterall,?rmexitsareyetanotherconsequenceoftheIPOe?ectthatin?uences?rms’innovationpaths.Collectingpatentinformationsubsequentto?rmexitsisfeasibleasinmostcases,evenifa?rmisacquired,itspatentsarestillassignedtotheacquiredratherthantheacquiringcompany.6

Icalculatethenumberofcitationsapatentreceivesinthecalendaryearofitsapprovalandinthesubsequentthreeyears.Thistimeframeisselectedto?tthenatureofthesample.

7

This article is protected by copyright. All rights reserved.

SincemanyoftheIPO?lingsinthesampleoccurtowardtheendofthe1990s,increasingthetimehorizonofcitationcountswouldreducesamplesize.Furthermotivatingtheselectedcitationhorizonisthefactthatcitationsareconcentratedinthe?rstfewyearsfollowingapatent’sapproval,withaconsiderableserialcorrelationincitationrates(AkcigitandKerr(2013)).

SincetheNBERpatentdatabaseendsin2006,Isupplementitwithinformationfrom

Accepted ArticletheHarvardBusinessSchool(HBS)patentdatabase,whichcoverspatentsgrantedthroughDecember2009.Thisenablescalculationofcitationsforpatentsgrantedtowardtheendofthesample.Overall,thesampleconsistsof39,306grantedpatentsof?rmsthatwentpublicand4,835grantedpatentsof?rmsthatwithdrewtheir?ling.[INSERTTABLEI]

TableIcomparesthepatentingactivityofwithdrawnandIPO?rmsinthethreeyears

priortotheIPO?ling.I?ndnosigni?cantdi?erencesacrossanyofthepatentingmeasures.SinceavalueofoneforScaledCitationsimpliesthata?rmisproducingpatentsofaveragequality,itisinterestingtonotethatbothIPO?rmsandwithdrawn?rmsproducepatentsthataresubstantiallymorecitedthancomparablepatentsintheyearsprecedingtheIPO?ling(80%higherforwithdrawn?rmsand89%higherforIPO?rms).Thisevidencesuggeststhat?rmsthatchoosetogopublicarelikelytodosofollowinginnovationbreakthroughs,raisingconcernsofpost-IPOmeanreversion,regardlessoftheIPOe?ect.Thisfurthermotivatestheempiricalapproachinthispaper.7

C.FinancialInformationandFirmCharacteristics

Theanalysisofprivate?rmsiscomplicatedduetodatalimitations.Whilepatentsare

usefulincapturingtheinnovationactivityofbothpublicandprivate?rms,no?nancialinformationisreadilyavailableforwithdrawn?rmsinstandard?nancialdatabases.Toovercomethisconstraint,Icollectwithdrawn?rms’?nancialinformationfrominitialregis-8

This article is protected by copyright. All rights reserved.

trationstatementsbydownloadingFormS-1?lingsfromtheSEC’sEDGARdatabase,whichisavailableasof1996.Icollect?nancialinformationforIPO?rmsfromCOMPUSTATandCapitalIQ.

Additionalinformationon?rmcharacteristicscomesfromvarioussources.Iobtaindata

Accepted Articleonventurecapital(VC)fundingfromSDC,VentureXpert,andregistrationstatements.Isupplementthesedatawithinformationona?rm’sageatthetimeoftheIPO?linganditsunderwriters’rankingfromregistrationforms,VentureXpert,JayRitter’swebpage,andtheSDCdatabase.Finally,Icollectinformationon?rmexits,thatis,eventsinwhicha?rmisacquired,goespublicinasecondattempt(forwithdrawn?rms),or?lesforbankruptcy.IuseCOMPUSTATandCapitalIQtosearchforacquisitionsandbankruptcies,andtheSDCdatabasetoidentifysecondIPOsofwithdrawn?rms.Iperformextensivecheckstoverifythenatureofprivate?rms’exitsusingtheDealPipelinedatabase,Lexis-Nexisandwebsearches.

TableIcomparesthecharacteristicsofIPO?rmsandwithdrawn?rmsatthetimeof

?ling.I?ndnosigni?cantdi?erencesin?rmsize,measuredbythelogof?rmassets,orinR&Dexpendituresnormalizedby?rmassets.However,withdrawn?rmshaveahighercash-to-assetsratioandlowernetincometoassets.Ialso?ndnosigni?cantdi?erencesinthereputationoftheleadunderwriter,whichisoftenusedasaproxyfor?rmquality(CarterandManaster(1990),Carter,Dark,andSingh(1998),andLoughranandRitter(2004)),orin?rmageatthetimeof?ling.

Therearestarkdi?erences,however,intheNASDAQ?uctuationsthat?rmsexperience

aftertheIPO?ling.Speci?cally,?rmsthatgopublicexperienceonaveragea3%increaseintwo-monthNASDAQreturnsfollowingtheIPO?ling,while?rmsthatwithdrawexperienceonaverageasharpdropof6%overasimilarperiod.Thedi?erencesinNASDAQreturnsinthethreemonthspriortotheIPO?ling,however,arefairlysmall(5%increasefor?rmsthatultimatelyremaineprivateversus7%forthosethatgopublic).Thesedi?erencesprovidefurthermotivationforusingNASDAQ?uctuationsasasourceofvariationthata?ectsIPO

9

This article is protected by copyright. All rights reserved.

completionchoices.Idiscussthisindetailinthenextsection.

TableIalsosummarizes?rmexitsinthe?veyearsfollowingtheIPO?ling.I?ndthat29%ofthewithdrawn?rmsand24%oftheIPO?rmsareacquiredoverthisperiod,withthedi?erencebetweenthetwogroupsinsigni?cant.Only2%ofbothIPOandwithdrawn?rmsgobankrupt.Additionally,18%ofthewithdrawn?rmsultimatelygopublicinasecondattemptinthe?veyearsfollowingtheIPO?ling.

Thesmallfractionofwithdrawn?rmsthatreturntopublicequitymarketsinasecond

Accepted Articleattemptishighlightedintheliterature(DunbarandFoerster(2008),Busaba,Beneviste,andGuo(2001)).SeveralexplanationsexistforthelowfractionofsecondattemptIPOs.ReturningtoIPOmarketsinasecondattemptmaybedi?cultasthewindowofopportunitymayhaveclosedduetotheboomandbustnatureofIPOmarkets(IbbotsonandRitter(1995)).BrauandFawcett(2006)?ndthatthemostimportantsignalwhengoingpublicisa?rm’spasthistoricalearnings.Ifgoingpublicrequiresseveralyearsoffastgrowthtoattractinvestors’attention,suchgrowthmaybedi?culttoreplicateinasecondattempt.DunbarandFoerster(2008)furthersuggestthattherearereputationalcostsassociatedwiththedecisiontowithdraw,whichmayprevent?rmsfromreturningtoequitymarkets.

II.EmpiricalStrategy

A.EmpiricalDesign

Identifyingthee?ectsofgoingpubliconinnovationand?rmoutcomesischallengingdue

toinherentselectionissuesthatarisefromthedecisiontogopublic.ThisconcernisevidentinFigure1,whichpresentswithin-?rmchangesininnovationquality,asmeasuredbyScaledCitations,intheyearsaroundtheIPO.8FirmsexperienceamonotonicdeclineinpatentnoveltybetweenthetwoyearsbeforetheIPOeventandthe?veyearsafter.Duetothechoicetogopublic,post-IPOperformancecouldbedrivenbyreversiontothemeanandlifecyclee?ects,ratherthanthee?ectofbecomingapubliclytraded?rm.

10

This article is protected by copyright. All rights reserved.

[INSERTFIGURE1]

Whatisareasonablecounterfactualfor?rmsthatgopublic?Simplycomparingpost-IPOinnovationtoanaverageprivate?rmmaygeneratebiasedestimates,asmostprivatecompaniesarefundamentallydi?erentandnevergopublic.ThisisevidentinTableI,whereonecanseethatpre-?linginnovationqualityof?rmsthatgopublicis80%higherthantheaverageinnovationintherespectivetechnologicalclasses.Toovercomethisconcern,Ifocusonthose?rmsthatsubmitaninitialregistrationstatementtotheSECinanattempttogopublic.

Icomparethelong-runinnovationof?rmsthatgopublicwith?rmsthat?ledtogopublic

Accepted Articleinthesameyearbutultimatelywithdrawtheir?lingandremainedprivate.Thissetupisattractiveasitallowsforcomparisonofthepost-IPOperformanceof?rmsthatgopublicwiththatofprivate?rmsatasimilarstageintheirlifecycle.Thebaselinespeci?cationis

Yipost=α1+β1IPOi+γ1Yipre+Xi??δ1+νk+μt+ε1i,

(1)

whereYipostistheaverageinnovationperformanceinthe?veyearsfollowingtheIPO?ling

(averagescaledcitations,averagescaledoriginality/generality,andaveragescalednumberofpatentsperyear),YipreistheequivalentmeasureinthethreeyearspriortotheIPO?ling,andIPOiisthedummyvariableofinterest,indicatingwhethera?lergoespublicorremains

private.Underthenullhypothesisthatgoingpublichasnoe?ectoninnovation,β1should

notbestatisticallydi?erentfromzero.Thismodelincludesindustry(vk)andIPO?ling

year(μt)?xede?ects.

IfthedecisiontowithdrawanIPO?lingisrelatedtounobservedinnovationpolicyor

innovativeopportunities(capturedintheerrorterm),thenβ1maybebiased.Therefore,IinstrumentfortheIPOcompletionchoiceusingNASDAQreturnsinthe?rsttwomonthsofthebook-buildingphase.Thedecisiontouseatwo-monthwindowissomewhatarbitrary.OnecouldusetheNASDAQreturnsovertheentireperiodofthebook-buildingphase.

11

This article is protected by copyright. All rights reserved.

However,sincethelengthofthebook-buildingphaseisoftencorrelatedwiththelikelihoodofwithdrawing,Ichoosea?xedwindowthatissu?cientlyshorterthantheaveragelengthofthebook-buildingperiod.

The?gurebelowillustratesthetimelineoftheIPO?lingandtheNASDAQ?uctuations

Accepted Articleduringthebook-buildingphase.Onaverage,theownershipchoiceismadewithinthefourmonthsfollowingtheIPO?ling.Firm-levelinnovationismeasuredoverthe?ve-yearperiodaftertheIPO?ling.9

Toimplementtheinstrumentalvariablesapproach,Iestimatethefollowing?rst-stage

regression:

IPOi=α2+β2NSDQi+γ2Yipre+Xi??δ2+νk+μt+ε2i,

(2)

whereNSDQiistheinstrumentalvariable.Thesecond-stageequationestimatestheimpactofIPOoninnovationactivity:

??Yipost=α3+β3IPOi+γ3Yipre+Xi??δ3+νk+μt+ε3i,

(3)

??whereIPOiarethepredictedvaluesfrom(2).Iftheconditionsforavalidinstrumentalvariablearemet,β3capturesthecausale?ectofanIPOoninnovationoutcomes.Iimplementtheinstrumentalvariableestimatorusingtwo-stageleastsquares(2SLS).Ialsouseaquasi-maximumlikelihood(QML)Poissonmodeltoestimatethespeci?cation(BlundellandPowell(2004)),whichisthestandardestimationmethodusedinboththeinnovationliteratureandcountdataanalysismoregenerally.

Itisimportanttonotethattheestimatesintheinstrumentalvariablesanalysisare

12

This article is protected by copyright. All rights reserved.

comingonlyfromthesensitive?rms(ImbensandAngrist(1994)).Inotherwords,theestimatesarecomingonlyfromthose?rmsthatwouldaltertheirIPOcompletiondecisioniftheyexperienceaNASDAQdrop,andthereforearesensitivetochangesintheinstrument.IntheInternetAppendixIprovideasimpleexampletoillustratethispoint.InthenextsectionIdiscusstheassumptionsthatneedtoholdfortheinstrumenttobevalid.

Accepted ArticleB.NASDAQFluctuationsandIPOCompletion

Fortheinstrumenttobevalid,itmuststronglya?ectIPOcompletionchoice.Indeed,

issuersarehighlysensitivetostockmarket?uctuationsduringthebook-buildingphase(Benvenisteetal.(2003),Busaba,Benveniste,andGuo(2001),Dunbar(1998),DunbarandFoerster(2008),EdelenandKadlec(2005)).ThissensitivityisalsoillustratedinFigure2,whichplotsthefractionofmonthly?lingsthatultimatelywithdrawagainstthetwomonthsofNASDAQreturnscalculatedfromthemiddleofeachmonth,whichapproximatesthestockmarket?uctuationsduringtheinitialpartofthebook-buildingphase(thecorrelationofthetwoplotsequals-0.44andissigni?cantlydi?erentfromzeroatthe1%level).Consistentwiththisevidence,asurveybyBrauandFawcett(2006)?ndsthatCFOsthatwithdrawanIPOregistrationindicatethatmarketconditionsplayedaroleintheirdecision.[INSERTFIGURE2]

Whywould?rmsnotsimplywaitformorefavorablemarketconditions?Thereare

severalpossiblereasons.First,a?lingregistrationautomaticallyexpires270daysafterthelastamendmentoftheIPO?ling,whichlimitsthetimetocompletetheIPO?ling(Lerner(1994)).Additionally,waitingiscostly:aslongastheapplicationispending,?rmscannotissueprivateplacements,andareforbiddenfromdisclosingnewinformationtospeci?cinvestorsorbanks.Indeed,?rmsarerequiredtoupdatetheregistrationstatementperiodicallytore?ectthecurrenta?airsofthecompany.Theseconsiderationslead?rmstowithdrawpriortotheautomaticexpirationoftheIPO?ling.

13

This article is protected by copyright. All rights reserved.

The?rst-stageresults,presentedinTableII,demonstratethee?ectofNASDAQreturnsduringthebook-buildingphaseonIPOcompletion.Thedependentvariableisequaltooneifa?rmcompletestheIPO?ling,andzerootherwise.Allspeci?cationsinclude?lingyearandindustry?xede?ectsusingOLS.Incolumn(1),I?ndthatthecoe?cientontwo-monthNASDAQreturnsequals0.704andissigni?cantatthe1%level.AonestandarddeviationdeclineinNASDAQreturnstranslatesintoa8.72TcreaseinthelikelihoodofcompletingtheIPO.Moreover,theF-statisticequals47.79andexceedsthethresholdofF=10,whichsuggeststhattheinstrumentisstrongandunlikelytobebiasedtowardtheOLSestimates(Bound,Jaeger,andBaker(1995),StaigerandStock(1997)).[INSERTTABLEII]

Incolumn(2),Icontrolforthethree-monthNASDAQreturnspriortotheIPO?ling,

Accepted Articlethenumberofpre-?lingpatents,aVCbackingdummyvariable,andthelocationofthe?lerwithintheIPOwave.10Thecoe?cientremainshighlystatisticallysigni?cant,suggestingthattheinstrumentisalmostorthogonaltotheaddedcontrolvariables.Incolumns(3)and(4),Ilimitthesampletothepre-2000period.Theresultsremainsimilar.Incolumns(5)and(6),IcalculatetheNASDAQreturnsovertheentirebook-buildingperiod.11Thecoe?cientisstillsigni?cantatthe1%level,butthemagnitudeofthecoe?cientissomewhatlower.Incolumns(7)and(8),Iconstructadummyvariablethatequalsoneifthe?rmdoesnotexperienceaNASDAQdrop,wherea?rmisclassi?edashavingexperiencedaNASDAQdropifitstwo-monthNASDAQreturnfollowingtheIPO?lingisamongthelowest25%ofall?lerswithinthesameyear.Thedummyvariableincolumn(7)ishighlystatisticallysigni?cant,re?ectinga10.6%increaseinthelikelihoodofIPOcompletion.Column(8)addstheadditionalcontrolvariables.Theresultsagainremainunchanged.

Figure3illustratesthenonparametricrelationbetweenthetwo-monthNASDAQ?uctu-

ationsandthelikelihoodofcompletingtheIPO?ling.The?gureshowsthataslongastheNASDAQ?uctuationsarenegative,thereisapositiveandmonotonicassociationbetween

14

This article is protected by copyright. All rights reserved.

asmeasuredbynumberofpatents.Inthatcase,theadditionoflow-qualityinnovationprojectsmaygeneratetheresultsratherthanarepositioningofresearchtolower-impacttopics.TheanalysisinTableVIIIaddressesthisquestionbyexploringchangesinthescaleofinnovation.15

[INSERTTABLEVIII]

Theendogenousmodelincolumn(1)indicatesthatIPO?rmsproducesigni?cantly

Accepted ArticlemorepatentsperyearfollowingtheIPO?ling,witha37.75%increaserelativetothepre-IPOaverage.Column(2),however,indicatesthattheabovee?ectisinsigni?cantwhenthereduced-formspeci?cationisestimated.The2SLSestimateincolumn(3)showsthatthecoe?cientontheIPOvariableisinsigni?cantandthemagnitudedeclinesto28.17%.Infact,whenusingtheIVPoissonspeci?cationincolumn(4),thecoe?cientontheIPOvariableisclosetozeroandinsigni?cant.

A.3.

RobustnessChecksandInterpretation

InthissectionIsummarizeseveralsupplementalanalysesthattesttherobustnessand

interpretationoftheabove?ndings.IstartbyexploringwhetherthedeclineintheaverageinnovationqualityofIPO?rmscanbedrivenbyalowerpatentingthresholdaftertheIPO.Thismayleadtotheadditionoflow-qualitypatentsandhenceloweraveragepatentingquality.Todoso,Iexamine?rms’best(i.e.,most-cited)patent,whichisunlikelytobea?ectedbytheadditionoflow-qualitypatent?lings.I?ndthatthequalityofIPO?rms’bestpatentdeclinesaswell,withacomparablemagnitudeasthedeclineinaverageinnovationqualityreportedinTableVI.Thisevidence,whichisreportedintheInternetAppendix,suggeststhatgoingpublica?ectstheentirepatentdistributionratherthansimplydrivingaverageperformancedownbytheadditionoflow-qualityprojects.16

Next,cash-rich?rmsmaybeassociatedwithfewercitationsbecauseciting?rmsfacehigherlitigationrisk,whichwouldmechanicallygeneratetheresultsinthepaper.Totest

20

This article is protected by copyright. All rights reserved.

thisconcern,IfocusonpatentsapprovedbeforetheIPO?lingandexaminewhetherpatents’annualcitationratechangesafterthe?rmsgopublic(relativeto?rmsthatwithdrawthe?ling).InresultsreportedintheInternetAppendix,I?ndthatchangesinthecitationrateofexistinginventionscannotbeexplainedbythetransitiontopublicequitymarkets.IalsoexplorewhethertheresultsaredrivenbytheInternetbubbleyears.Asillustrated

Accepted ArticleinTableIV,theinstrumentstronglypredictsIPOcompletionevenwhenall?rmsthat?ledduringtheInternetbubbleandthereafterareexcluded.InrobustnesstestsIreestimatetheinnovationnoveltyregressionsafterexcludingall?rmsthat?letogopublicasof1999.Theresults,reportedintheInternetAppendix,remainsigni?cantandqualitativelyunchanged.Inthemainanalysis,Icollectinformationontheinnovationtrajectoryof?rmsinthe?ve

yearssubsequenttotheIPO?ling.Ininterpretingtheresults,itisnaturaltowonderhowtheendogenoustransitionof18%ofthewithdrawn?rmstopublicequitymarketsa?ectstheestimates.Toexaminethisquestion,Irepeattheinstrumentalvariablesanalysisusinganendogenousvariablethatequalsoneifa?rmgoespublicinthetwo(orthree)yearsaftertheIPO?ling,regardlessofwhetheritwithdrawsinthe?rstattempt.Thus,ifa?rmwithdrawsits?lingandreturnstotheIPOmarketinthe?rsttwo(three)yearsafterthe?ling,this?rmwouldbeconsideredpartofthetreatmentgroup(andnotthecontrolgroup).17As

reportedintheInternetAppendix,thesespeci?cationsgenerateestimatesthataresimilartothosereportedabove.

Finally,itisimportanttounderstandwhethertheestimatesaredrivenbychangesin

?rmsthatwithdrawtheirIPO?lingandremainprivateorby?rmsthatgopublic.Forexample,withdrawn?rmsmaychoosetoincreasetheirinnovationinordertosuccessfullygopublicinasecondattempt,ratherthanIPO?rmsexperienceadeclineininnovation.However,I?ndthatallIPO?lersexperienceadeclineinscaledinnovationmeasures,both?rmsthatexperienceaNASDAQdrop(andthusaremorelikelytoremainprivate)and?rmsthatdonotexperienceaNASDAQdrop.18However,thedeclineismoresubstantialamong?rmsthatendupgoingpublic.Similarly,changesintheacquisitionofexternalinnovation,

21

This article is protected by copyright. All rights reserved.

asdiscussedinSectionIII.C,aredrivenbythose?rmsthatgopublicratherthanthosethatremainprivate.19

Accepted ArticleB.InventorMobilityandProductivityChanges

AsubstantialportionofR&Dinvestmentisintheformofwagesforhighlyeducated

scientistsandengineers,whoencompassthe?rm’sknowledge.ItisthereforenaturaltoexplorewhethergoingpublicleadstoinventormobilityandproductivitychangesfollowingtheIPO.

B.1.InventorLevelData

Thepatentdatabaseprovidesaninterestingopportunitytotrackinventors’mobility

across?rms,aseachpatentapplicationincludesboththenameoftheinventoranditsassignee(mostoftentheinventor’semployer).Analysisofinventor-leveldata,however,ischallengingforseveralreasons.First,patentsareassociatedwithinventorsbasedontheirnameandgeographiclocation,butinventors’namesareunreliableas?rstnamescanbeabbreviatedanddi?erentinventorsmayhavesimilarorevenidenticalnames.Second,whileitispossibletoinferthataninventorchanged?rms(forexample,aninventorisassociatedwithapatentforcompanyAin1987andapatentforcompanyBin1989),theprecisedateoftherelocationisunavailable.Additionally,transitionsforwhichinventorsdonotproducepatentsinthenewlocationarenotobservable.Hence,thismethodidenti?esrelocationsofthemorecreativeinventorswhopatentfrequentlyandpresumablymatterthemost.Toovercomethehurdleofnamematching,IusetheHBSpatentingdatabase,whichin-

cludesuniqueinventoridenti?ers.Theuniqueidenti?ersarebasedonre?neddisambiguationalgorithmsthatseparatesimilarinventorsbasedonvariouscharacteristics(Lai,D’Amour,andFleming(2009)).Whenpatentapplicationsincludemultipleinventors,Iattributeapatentequallytoeachinventor.Overall,Iobtaininformationonapproximately36,000in-ventorsinmysample.Irestricttheanalysistoinventorsthatproduceatleastasinglepatent

22

This article is protected by copyright. All rights reserved.

bothbeforeandaftertheIPO?lingandexaminethepatentingbehaviorofinventorsinthethreeyearsbeforeand?veyearsaftertheIPO?ling.Iidentifythreeinventortypes:1.Stayer–AninventorwithatleastasinglepatentbeforeandaftertheIPO?lingatthesamesample?rm.

2.Leaver–Aninventorwithatleastasinglepatentatasample?rmbeforetheIPO?ling,andatleastasinglepatentinadi?erentcompanyaftertheIPO?ling.20

3.Newcomer–AninventorwithatleastasinglepatentaftertheIPO?lingatasample?rm,butnopatentsbefore,andatleastasinglepatentatadi?erent?rmbeforetheIPO?ling.

Ofthe36,000inventorsinmysample,Icanclassify16,108inventorsintotheabove

Accepted Articlecategories.21Theseinventorsaccountforapproximately65%ofthesamplepatents.

B.2.

InventorLevelAnalysis

Iexplorechangesininventor-levelactivityusingtheinstrumentalvariablesapproach

introducedinSectionII.A.Istartbyinvestigatingchangesintheinnovationqualityofstayers.Next,Iexamineinventormobilitybystudyinginventors’likelihoodofleaving,spinningo?acompany,orjoiningthe?rmfollowingtheIPO?ling.

TheresultsarereportedinTableIX,wheretheunitofobservationisatthelevelofthe

inventor.Incolumn(1),Iexplorechangesinstayers’productivity.Ifocusonthesetofinventorsthatremainatthe?rm,andthedependentvariableistheaveragescaledcitationsproducedbyinventorsinthe?veyearsaftertheIPO?ling.Icontrolfortheinventor’spre-IPO?lingcitationsperpatent,aswellas?lingyearandindustry?xede?ectsandtheotherstandardcontrols.Standarderrorsareclusteredatthe?rmlevel,toallowforcorrelationsbetweeninventorsinthesame?rm.Iestimatethe2SLSincolumn(1),and?ndthattheIPOcoe?cientequals-1.094,whichisstatisticallysigni?cantatthe1%level.Themagnitudeislarge,correspondingtoa48Tclineininventors’innovationnoveltyinIPO?rmsrelativeto

23

This article is protected by copyright. All rights reserved.

thepre-IPO?lingperiod.22These?ndingssuggestthatthedeclineinIPO?rms’innovationactivitycouldbeattributedatleastinparttoachangeinthequalityofinnovationproducedbyinventorswhoremainatthe?rm.[INSERTTABLEIX]

Toestimatewhethergoingpublicmaya?ectinventors’departure,Ifocusoninventors

Accepted Articlethatproducepatentsatthe?rmbeforetheIPO?ling,andexaminethelikelihoodthattheyleavethe?rm.Incolumn(2),Irunasimilarspeci?cationtothatreportedincolumn(1),butnowthedependentvariableequalsoneiftheinventorisclassi?edasaleaverandzeroifaninventorisastayer.The2SLSestimatesshowthatinventorsinIPO?rmsare18%morelikelytoleavethe?rmaftertheIPO,withthecoe?cientstatisticallysigni?cantatthe1%level.23ThisresultdemonstratesthatthedeclineinthequalityofinnovationofIPO?rmsispotentiallyalsodrivenbythedepartureofinventors.

Doestheincreasedinventordepartureratealsotranslateintoahigherrateofspin-o?s?

FollowingAgrawaletal.(2012),Iidentifyallpatentsproducedbyleaverspost-departureandthe?rmstowhichthesepatentsareassigned.Anassigned?rmisconsideredaspin-o?ifthenumberofappliedpatentsbeforetheleaver’spatentiszero.24Thisapproachprovidesanapproximationofwhetheraleaverstartedanewcompany.Whilenotperfect,inaccuraciesaresimilarlylikelytoa?ectleaversofIPOandwithdrawn?rms.

Irepeatthespeci?cationofcolumn(2),focusingoninventorswhoproducedpatentsina

sample?rmbeforetheIPO?ling,andexplorethelikelihoodthatcreateaspin-o?.The2SLSestimatesreportedincolumn(3)ofTableIXillustratethatIPO?rmsare9.5%morelikelytogeneratespin-o?scomparedtowithdrawn?rms,suggestingthatIPO?rms’leaversremainentrepreneurialfollowingtheirdeparture.Thisresultisalsoconsistentwiththe?ndingthattheleaversofIPO?rmsgeneratehigherqualitypre-IPOpatentsrelativetothestayers,incontrasttothecaseofwithdrawn?rms(SeetheInternetAppendix).

Finally,IexplorewhetherIPO?rmsaremorelikelytoattractnewinventors.Incolumn

24

This article is protected by copyright. All rights reserved.

(4)Irestricttheanalysistoinventorsthatgeneratepost-?lingpatentsinasample?rm,andasadependentvariableIuseanindicatorforwhetheraninventorisanewcomer.Usingthe2SLSspeci?cationI?ndthatIPO?rmsaresubstantiallymorelikelytohirenewinventors.Themagnitudeofthecoe?cientislarge,correspondingtoa38.8%increase.25

Theresultsrevealthatthetransitiontopublicequitymarketshasimportantimplications

Accepted Articleforthehumancapitalaccumulationprocess,asitshapes?rms’abilitytoretainandattractinventors.FollowingtheIPO,thereisanexodusofinventorsleavingthe?rm,andmorespin-o?saregeneratedfollowingtheirdeparture.Additionally,goingpublica?ectsthepro-ductivityoftheinventorswhoremainatthe?rm.TheaveragequalityofpatentsproducedbystayersdeclinessubstantiallyatIPO?rms.However,thise?ectismitigatedinpartbytheabilityofIPO?rmstoattractnewinventorswhoproducepatentsofhigherqualitythantheinventorswhoremainatthe?rm.

C.AcquisitionofExternalTechnologies

Inthissection,Iexplorehowthetransitiontopublicequitymarketsa?ects?rms’reliance

onexternaltechnologiespurchasedthroughmergersandacquisitions.26Icollectinformation

onallacquisitionsconductedbyIPOandwithdrawn?rmsusingtheSDCdatabase.27In

resultsreportedintheInternetAppendix,I?ndthatIPO?rmsexhibitasharpincreaseinM&Aactivityinthe?veyearsfollowingtheIPO,whilethereisnomeaningfule?ectforwithdrawn?rms.

Acquisitions,however,areusedforavarietyofreasons.Thequestionremainswhether

acquisitionsareusedtobuyexternaltechnologies.Icollectinformationonpatentsgeneratedbytarget?rmsintheyearspriortotheacquisitionbymatchingacquisitiontargetstotheNBERpatentsdatabase.28Approximately7,500patentswereacquiredthroughmergersandacquisitionsinthe?veyearsfollowingtheIPO?ling,relativetoapproximately30,000patentsproduced.IntheInternetAppendix,IshowthatbeforetheIPO?lingbothwithdrawnandIPO?rmsrarelyacquireexternalpatentsthroughM&A.FollowingtheIPO,however,the

25

This article is protected by copyright. All rights reserved.

fractionofexternalpatentsinIPO?rms’portfolioincreasesto31%whileitremainssmallforwithdrawn?rms(8%).

Similarpatternsarisewhenusingtheinstrumentalvariablesapproach.InTableX,the

Accepted Articledependentvariableisthenumberofexternalpatentsboughtthroughmergersandacquisi-tionsinthe?veyearsfollowingtheIPO?ling.Allspeci?cationsfollowthemodeldescribedinSectionII.A,controllingfor?lingyearandindustry?xede?ects.Ialsoaddalladditionalcontrolvariablespreviouslymentioned.Iestimatethe2SLSincolumn(3)and?ndthatIPO?rmsacquire2.6morepatentsperyearcomparedtowithdrawn?rms.[INSERTTABLEX]

Theidentifyingassumptionhereisthat,fortwo?rmsthatattempttogopublicinthe

sameyear,changesinshort-termNASDAQ?uctuationshaveasimilare?ectonthelikelihoodofacquiringexternalpatents.Tofurtheralleviateconcernsabouttheexclusionrestriction,incolumn(4)Ialsocontrolfortheindustryacquisitionintensityinthe?veyearsfollowingtheIPO?ling.29I?ndthattheresultscontinuetohold.IntheInternetAppendix,theseresultscanbealsoillustratedbycomparingsimpleaveragesbetween?rmsthatexperienceaNASDAQdropandother?lersinthesameyear,similartotheapproachtakeninTableV.Giventheincreasedrelianceonexternalpatents,itisinterestingtocomparetheexternal

andinternalpatentsofIPO?rms.IntheInternetAppendix,I?ndthatonaverageexternalpatentsexhibithigherqualitythanpatentsgeneratedinternally,asmeasuredbyaveragescaledcitations.30Overall,theresultssuggestthatgoingpublicleadstoenhancedreliance

onexternaltechnologies.

IV.PotentialChannels

Theempirical?ndingsthusfarillustratethatthetransitiontopublicequitymarkets

changes?rm’sinnovationstrategy.FollowingtheIPO,?rmsexperienceadeclineininno-vationquality,substantialinventorturnover,andanincreaseintheacquisitionofexternal

26

This article is protected by copyright. All rights reserved.

innovation.

Inthissection,Idiscussseveralpotentialexplanationsforthe?ndings.FollowingtheIPO,various?rmcharacteristicschange,andhenceanumberofpotentialstoriesmayarise.Ifocusonfoursuchexplanations.Thecorporate?nanceliteraturehaslongarguedthatpublicownershipentailsagencyproblemsbetweenmanagersandshareholdersduetotheseparationofownershipandcontrol(BerleandMeans(1932),JensenandMeckling(1976)).Therefore,I?rstfocusontwohypothesesbasedonagencyproblemsthatcanleadpubliclytraded?rmstoconducttoolittleinnovation.Speci?cally,motivatedbyAghion,VanReenen,andZingales(2013),Ifocusonthemanagerialcareerconcernsandthequietlifehypotheses.However,itmaybethecasethatagencyproblemsaremorepervasiveinprivate?rms,leadingtotoomuchinnovation.Thisalternativetheoryiscapturedbytheexcessiveinnovationhypothesis.Finally,Iconsiderthecaseinwhichtheresultsmaynotbedrivenbyagencyproblems.Inparticular,thisagency-freehypothesisismotivatedbytheassumptionthatcommercializationiscapital-intensiveandthuspossibleonlyfollowingtheIPO.

Idiscussthesehypothesesindetailbelow,andthenprovidesuggestiveempiricalevidence

Accepted Articleinanattempttodi?erentiatebetweenthesealternativestories.

A.Hypotheses

A.1.

CareerConcernsHypothesis

The?rsthypothesisisbasedonavariationofthecareerconcernstheory(Holmstrom

(1982)).Underthishypothesis,theprincipal-agentproblemoccursbetweenshareholdersandthemanager.Themanagergainsprivatebene?tsfromretainingherjob.Innovationmaycostthemanagerherjob,however,ifshareholdersattributeinnovationfailures,eventhoseduetopurelystochasticreasons,topoormanagerialskill.Asaresult,concernsabouttheimpactofinnovationoninvestors’perceptionofthemanager’sability,thatis,careerconcerns,leadtoanaversiontoinnovation.

Increasedmonitoringmayallowinvestorstodistinguishbetweenunluckynegativedraws

27

This article is protected by copyright. All rights reserved.

andmanagerialskill(Aghion,VanReenen,andZingales(2013)),whichwouldhelpinsulatethemanagerfromadversereputationalconsequencesofinnovation.Inprivate?rms,whereownershipisconcentrated,investorshavegreaterincentivetomonitorthemanager,whichshouldleadtomoreinnovation.Inpublicownership,incontrast,dispersedownershipleadstolowerincentivestomonitorwhichmaydissuadethemanagerfrominnovating.31

Anadditionalpredictionofthishypothesisisthatanentrenchedmanager,whosejobis

Accepted Articlemoresecure,maybewillingtoinnovaterelativelymoreaftertheIPO,asinnovationdoesnotexposethemanagertoasmuchriskoflosingherjob.

A.2.

QuietLifeHypothesis

Anotherhypothesisrelatedtotheagencyproblembetweenshareholdersandtheman-

agerisbasedontheideathatinnovationisadi?culttaskthatimposesaprivatecostonthemanager(forexample,itmayrequiredeviatingfromstandardroutines),deterringthemanagerfrominnovationactivity(basedonasimplevariationofHolmstromandTirole(1997)).Thishypothesis,whichissimilartothequietlifehypothesisofBertrandandMul-lainathan(2003),predictsthatinprivatelyownedcompanies,investormonitoringmayleadthemanagertoloseherjobifshedoesnotinnovate,whileunderpublicownership,weakerinvestorincentivetomonitormayallowthemanagertoinnovateless.ThishypothesisalsopredictsthatanentrenchedmanagerwillchoosetoinnovatelessfollowingtheIPO,asherjobismoresecure,incontrasttothecareerconcernshypothesis.

A.3.

ExcessiveInnovationHypothesis

Itmayalsobethecasethatprivate?rmsinnovatetoomuch.Iftheentrepreneurgains

privatebene?tsfrominnovation,asopposedtocommercialization,andsheisalsothelargestownerinaprivate?rmandthereforehascontrol,thentheentrepreneurmaychoosetofocustoomuchoninnovation.Here,theprincipal-agentproblemarisesbetweentheentrepreneur,whoisthelargestshareholder,andminorityshareholders.32Inpubliclytraded?rms,in

28

This article is protected by copyright. All rights reserved.

contrast,theentrepreneurisnotthelargestshareholderandhaslesscontrol,andthereforehaslessabilitytopursueexcessiveinnovation.

A.4.Agency-FreeHypothesis

Underthelasthypothesis,?rmschoosetogopublictocommercializeexistinginnovation.

Accepted ArticleCommercializationalonecannotexplaintheresults,asallsample?rmsareattemptingtogopublic.However,ifcommercializationiscapitalintensive(morethaninnovation),thenonlythose?rmsthatsuccessfullycompletetheIPOcanpursuecommercializationstrategy.Ifthefocusoncommercializationleads?rmstoinnovateless,thismayexplainthedeclineininnovationquality.Incontrast,?rmsthatremaineprivatedonothavetheresourcestocommercializeandthuscontinuetoinnovate.Importantly,forthishypothesistohold,itisnecessarytoassumethatsomelimitedresource,suchasmanagerialattention,prevents?rmsfrompursuingcommercializationandinnovationsimultaneously,despitetheimprovedaccesstocapitalfollowingtheIPO.

B.EmpiricalTests

Thissectionprovidessuggestiveempiricalevidenceinanattempttodistinguishbetween

thefourhypothesesabove.

B.1.

ExcessiveInnovationHypothesis

Theexcessiveinnovationhypothesispositsthatwhentheentrepreneurhascontrol,pri-

vate?rmsmayconducttoomuchinnovationthatmaynotnecessarilycontributeto?rmvalue.Toexplorethishypothesis,Icompare?rmsthatgopublicwithasetofprivate?rmsinwhichtheentrepreneurhasnocontrol.Speci?cally,Iconsiderprivate?rmsownedbyVC?rmsandmanagedbyaprofessionalmanager(thatis,anonfoundermanager).Theseprivate?rmsarelikelytosu?ertheleastfromagencyproblemsthatleadtoexcessiveinno-vation.IrepeatthemaininstrumentalvariablesanalysistoexaminetheIPOe?ectforsuch

29

This article is protected by copyright. All rights reserved.

private?rms(seetheInternetAppendix).I?ndthatevenVCownedprivatecompaniescon-ductsigni?cantlyhigherqualityinnovationintheyearsfollowingtheIPOrelativeto?rmsthatgopublic.Moreover,inventorsinsuchprivate?rmsaresigni?cantlylesslikelytoleave.Thisevidencesuggeststhatprivate?rmsdonotoverinvestininnovation,asconjecturedbytheexcessiveinnovationhypothesis,ascompaniesthatarelikelytosu?ertheleastfromagencyproblemsproducehigherqualityinnovation.

B.2.

Agency-FreeHypothesis

Accepted ArticleCantheagency-freehypothesisexplaintheresults?Ifimprovedaccesstocapitalallows

?rmstofocusoncommercialization,leadinginturntoadeclineininnovationquality,thensuchadeclineshouldbelessapparentin?rmsthatalreadyhavesigni?cantresourcesatthetimeoftheIPOorthatraiselittlecapitalthroughtheIPO.Such?rmsmaybethelargest?rmsthatgopublic,?rmswithhighcashrelativetoassetsatthetimeoftheIPO,and?rmsthatraisetheleastcapitalthroughprimaryshares.Suchadeclineshouldalsobelessapparentin?rmsthatfacelowercommercializationcosts,suchas?rmsinthesoftwareandservicesindustries.Toexaminethee?ectoftheIPOoneachofthesegroups,Irepeatthemaininstrumentalvariablesapproach,andreporttheresultsintheInternetAppendix.InallcasesIstill?ndasigni?cantdeclineininnovationquality.Whilethisevidenceisnotconclu-sive,itsuggeststhattheaccesstocapitalandpotentialabilitytopursuecommercializationafteranIPOdoesnotexplainthedeclineininnovationquality.

Inlightofevidenceintheliteraturethatinnovationiseconomicallyimportant(e.g.,Hall,

Ja?e,andTrajtenberg(2005),Koganetal.(2012)),itmaybesurprisingif?rmschoosenottopursueinnovationaftertheIPO.33Forexample,Koganetal.(2012)?ndthathigh-qualityinnovationpredicts?rmproductivityaswellascapitalandlabor?owsfromnon-innovativetoinnovative?rmswithinanindustry.Moreover,inthe“useofproceeds”sectionintheIPO?lingprospectus,I?ndthat75%of?rmsstatethattheyalsointendtouseproceedsfromtheIPOfortechnologicaldevelopmentandinnovation.34WhenIrepeatthemaininstrumental

30

This article is protected by copyright. All rights reserved.

variablesstrategyrestrictingthesampleto?rmsthatstateanintentiontoinnovate,asreportedintheInternetAppendix,I?ndasimilardeclineininnovationqualitysimilartothemain?ndings.35

B.3.

CareerConcernsandQuietLifeHypotheses

Accepted ArticleInlightoftheaboveevidence,Inextconsiderwhetherpubliclytraded?rmsunderinvest

ininnovationduetoagencyproblemsthatarisebetweenmanagementandshareholders.Consistentwiththemainresults,boththequietlifeandcareerconcernshypothesessuggestthatinnovationwilldeclinefollowingtheIPO.However,thesehypotheses,yielddi?erentpredictionswheninteractedwithmanagerialentrenchment.Managerialentrenchmentmayimproveinnovationunderthecareerconcernshypothesisbutdecreaseinnovationunderthequietlifehypothesis.Thesecontrastinge?ectsofmanagerialentrenchmentarediscussedbyFismanetal.(2013).

Iexploretheinteractionofgoingpublicwithmanagerialentrenchmentinane?ortto

distinguishbetweenthesetwohypotheses.Asaproxyformanagerialentrenchment,IuseadummyvariablethatcapturescasesinwhichtheCEOisalsothechairmanoftheboard(ShleiferandVishny(1989)).IhandcollectthisinformationfromIPOprospectuses.IntheInternetAppendix,Irepeatthemaininstrumentalvariablesanalysistoexaminethee?ectoftheIPOon?rminnovationseparatelyforentrenchedandnon-entrenchedCEOs.I?ndthat?rmswithnon-entrenchedmanagersexperienceasigni?cantdeclineininnovationqualityfollowingtheIPO,andinventorsofthese?rmsaremorelikelytoleave.Incontrast,suche?ectsdonotholdfor?rmswithentrenchedmanagement.Theseresultssupportthecareerconcernshypothesis.

C.Discussion

Overall,thissectiono?erssuggestiveevidencethatagencyproblemsbetweenmanage-

mentandshareholders,intheformofcareerconcerns,leadpubliclytraded?rmstopursue

31

This article is protected by copyright. All rights reserved.

lowerqualityinnovation.Theseresultsareconsistentwithrecentevidencethatexploresagencyproblemsinpublicequitymarketsinvarioussettings.36Forexample,Gao,Harford,andLi(2014)?ndthatpublic?rmsaremorelikelytoreplaceaCEO,relativetoprivate?rms,andthatsucheventsaremoresensitiveto?rmcontemporaneousperformance.Moreover,performancefollowingCEOdepartureismoremodestinpublic?rms.Related,Aghion,VanReenen,andZingales(2013)?ndthatinstitutionalinvestorsinpublic?rmsleadtomoreinnovation,andthelikelihoodofCEOturnoverfollowingarevenuedropislowerfor?rmswithgreaterinstitutionalownership.Thesepapersareconsistentwithcareerconcerns,whichmayleadmanagerstoselectmoreincrementalandlessriskyprojects.

Ifthepaper’smainresultsaredrivenbyagencyproblems,thenwhydo?rmsgopublic?

Accepted ArticleThisquestiongoesbacktoJensenandMeckling(1976).Therearemanycostsassociatedwiththedecisiontogopublic.Underwritingdirectexpenses,underpricing,increasedtrans-parency,andshort-termmarketpressuresarejustafewexamples.Thesecostsarebalancedbyvariousbene?tsfromgoingpublic,suchasimprovedaccesstocapital,increasedliquidity,greaterinvestordiversi?cation,andenhanced?rmreputation,amongothers(see,forexam-ple,Pagano,Panetta,andZingales(1998)andRitterandWelch(2002)).Anotherexampleofsuchbene?tsistheincreasedabilityof?rms,followingtheIPO,toacquireexternalinno-vationandattractnewinventors.Thesearetwoimportantmitigatingforcesthatalleviatethedeclineininternalinnovationandthedepartureofexistingemployees.

V.Conclusion

Inthispaper,IinvestigateanimportantyetunderstudiedaspectofIPOs,namely,theef-

fecton?rminnovation.The?ndingsinthispaperrevealacomplextrade-o?betweenpublicandprivateownership.Followingthetransitiontopublicequitymarkets,internalinnovationbecomeslessnovel,incomparisonwith?rmsthatwithdrawtheirIPO?lingandremainepri-vate.Inaddition,?rmsexperienceanexodusofskilledinventors.However,duetoincreased

32

This article is protected by copyright. All rights reserved.

accesstocapital,IPO?rmscanrelyonacquisitionoftechnologiesasanadditionalsourceofinnovationandcanattractnewhumancapital.Theseresultssuggestthatgoingpublicchangesthestrategiesthat?rmsemploywhenpursuinginnovation.Iproposefourpotentialexplanationsofthese?ndings,and?ndsuggestiveevidencethatsupportsmanagerialcareerconcerns.

Theseresultsrelatetoconcernsraisedbycorporatemanagers,bankers,andpolicymak-

Accepted ArticleersaboutwhethertherecentdeclineinIPOsmarksabreakdownintheengineofinnovationandgrowth(WeildandKim(2009)).ThepassageoftheJumpstartOurBusinessStar-tups(JOBS)Actaimstoeasetheabilityofyounger,fast-growingcompaniestoraisefundsthroughtheIPOmarket.Thispapershowsthat,beyondraisingcapital,goingpublica?ects?rms’internalprojectselection,humancapital,andoutsourcingstrategies,andthereforehasimportantimplicationswithrespecttotheoptimaltimingofgoingpublic.

33

This article is protected by copyright. All rights reserved.

REFERENCES

Aggarwal,Vikas,andDavidHsu,2013,Entrepreneurialexitsandinnovation,ManagementScience60,867-887.

Aghion,Philippe,andJeanTirole,1994,Themanagementofinnovation,QuarterlyJournalofEconomics109,1185-1209.

Aghion,Philippe,JohnVanReenen,andLuigiZingales,2013,Innovationandinstitutionalownership,Americaneconomicreview103,277-304.

Agrawal,AjayK.,IainM.Cockburn,AlbertoGalasso,andAlexanderOettl,2012,Whyaresomeregionsmoreinnovativethanothers?Theroleof?rmsizediversity,NBERWorkingpaper.

Akcigit,Ufuk,andWilliamR.Kerr,2013,Growththroughheterogeneousinnovations,NBERWorkingpaper.

Anton,JamesJ.,andDennisA.Yao,2004,Littlepatentsandbigsecrets:Managingintellectualproperty,RANDJournalofEconomics35,1-22.

Arrow,KennethJ.,1962,Economicwelfareandtheallocationofresourcesforinvention,inTheRateandDirectionofInventiveActivity(PrincetonUniversityPressandNBER).Asker,John,JoanFarre-Mensa,andAlexanderLjungqvist,2014,Corporateinvestmentandstockmarketlisting:Apuzzle?,Workingpaper,NewYorkUniversity.

Bennedsen,Morten,FranciscoP′erez-Gonz′alez,andDanielWolfenzon,2010,Thegover-nanceoffamily?rms,inRonaldAndersonandKurtBaker,Eds:CorporateGovernance:ASynthesisofTheory,Research,andPractice(Wiley&Sons).

Benveniste,LawrenceM.,AlexanderLjungqvist,WilliamJ.WilhelmJr,andXiaoyunYu,2003,Evidenceofinformationspilloversintheproductionofinvestmentbankingservices,JournalofFinance58,577–608.

Accepted ArticleBerle,AdolfA.,andGardinerC.Means,1932,TheModernCorporationandPrivateProp-erty(Macmillan,NewYork).

Bertrand,Marianne,andSendhilMullainathan,2003,Enjoyingthequietlife?Corporategovernanceandmanagerialpreferences,JournalofPoliticalEconomy111,1043-1075.

Beshears,John,2013,Theperformanceofcorporatealliances:EvidencefromoilandgasdrillingintheGulfofMexico,JournalofFinancialEconomics110,324-346.

Blundell,RichardW.,andJamesL.Powell,2004,Endogeneityinsemiparametricbinaryresponsemodels,ReviewofEconomicStudies71,655-679.

Bound,John,DavidA.Jaeger,andReginaM.Baker,1995,Problemswithinstrumentalvariablesestimationwhenthecorrelationbetweentheinstrumentsandtheendogeneousexplanatoryvariableisweak,JournaloftheAmericanStatisticalAssociation90,443-450.

34

This article is protected by copyright. All rights reserved.

Brau,JamesC.,andStanleyE.Fawcett,2006,Initialpublico?erings:Ananalysisoftheoryandpractice,JournalofFinance61,399-436.

Brav,Alon,andPaulA.Gompers,1997,Mythorreality?Thelong-rununderperformanceofinitialpublico?erings:Evidencefromventureandnonventurecapital-backedcompanies,JournalofFinance52,1791-1821.

Brav,Omer,2009,Accesstocapital,capitalstructure,andthefundingofthe?rm,JournalofFinance64,263-308.

Busaba,WalidY.,LawrenceM.Benveniste,andRe-JinGuo,2001,TheoptiontowithdrawIPOsduringthepremarket:Empiricalanalysis,JournalofFinancialEconomics60,73–102.

Carter,RichardB.,FrederickH.Dark,andAjaiK.Singh,1998,Underwriterreputation,initialreturns,andthelong-runperformanceofIPOstocks,JournalofFinance53,285-311.Carter,RichardB.,andStevenManaster,1990,Initialpublico?eringsandunderwriterreputation,JournalofFinance4,1045-1067.

Accepted ArticleChemmanur,ThomasJ.,ShanHe,andDebarshiK.Nandy,2010,Thegoing-publicdecisionandtheproductmarket,ReviewofFinancialStudies23,1855-1908.

Degeorge,Francois,andRichardZeckhauser,1993,ThereverseLBOdecisionand?rmperformance:Theoryandevidence,JournalofFinance4,1323-1348.

Dunbar,CraigG,1998,Thechoicebetween?rm-commitmentandbest-e?ortso?eringmethodsinIPOs:Thee?ectofunsuccessfulo?ers,JournalofFinancialIntermediation7,60-90.

Dunbar,CraigG.,andStephenR.Foerster,2008,Secondtimelucky?WithdrawnIPOsthatreturntothemarket,JournalofFinancialEconomics87,610–635.

Edelen,RogerM.,andGregoryB.Kadlec,2005,IssuersurplusandthepartialadjustmentofIPOpricestopublicinformation,JournalofFinancialEconomics77,347–373.Fisman,RaymondJ.,RakeshKhurana,MatthewRhodes-Kropf,andSoojinYim,2013,GovernanceandCEOturnover:Dosomethingordotherightthing?ManagementScience60,319-337.

Gao,Huasheng,Po-HsuanHsu,andKaiLi,2014,Publicequitymarketsandcorporateinnovationstrategies:evidencefromprivate?rms,Workingpaper,UniversityofBritishColumbia.

Gao,Huasheng,JarradHarford,andKaiLi,2014,InvestorhorizonandCEOturnover-performancesensitivity,Workingpaper,UniversityofBritishColumbia.

Griliches,Zvi,1990,Patentstatisticsaseconomicindicators:Asurvey,JournalofEconomicLiterature28,1661-1707.

35

This article is protected by copyright. All rights reserved.

Hall,BronwynH.,ZviGriliches,andJerryA.Hausman,1986,PatentsandR&D:Istherealag?InternationalEconomicReview27,265-283.

Hall,BronwynH.,AdamJa?e,andManuelTrajtenberg,2001,TheNBERpatentcitationdata?le:Lessons,insightsandmethodologicaltools,NBERWorkingpaper8498.Hall,BronwynH.,AdamJa?e,andManuelTrajtenberg,2005,Marketvalueandpatentcitations,RANDJournalofEconomics36,16–38.

Hall,BronwynH.,andJoshLerner,2010,The?nancingofR&Dandinnovation,inBronwynH.HallandNathanRosenberg,eds.:HandbookoftheEconomicsofInnovation,(Elsevier-North,Holland).

Holmstrom,Bengt,1982,Managerialincentiveproblems-Adynamicperspective,inEssaysinEconomicsandManagementinHonorofLarsWahlbeck,(SwedishSchoolofEconomics,Helsinki).

Holmstrom,Bengt,andJeanTirole,1997,Financialintermediation,loanablefunds,andtherealsector,QuarterlyJournalofEconomics112,663-691.

Ibbotson,Roger,andJayR.Ritter,1995,Initialpublico?erings,inRobertA.Jar-row,VojislavMaksimovic,andWilliamT.Ziemba,eds.:HandbooksofOperationsResearchandManagementScience(North-Holland,Amsterdam)

Imbens,GuidoW.,andJoshuaD.Angrist,1994,Identi?cationandestimationoflocalaveragetreatmente?ects,Econometrica62,467-475.

Accepted ArticleJa?e,AdamB.,andManuelTrajtenberg,2002,Patents,Citations,andInnovations:AWindowontheKnowledgeEconomy(MITPressBooks).

Jain,BharatA.,andOmeshKini,1994,Thepost-issueoperatingperformanceofIPO?rms,JournalofFinance49,1699-1726.

Jensen,Michael,andWilliamH.Meckling,1976,Theoryofthe?rm:Managerialbehavior,agencycostsandownershipstructure,JournalofFinancialEconomics3,305-360.Kogan,Leonid,DimitrisPapanikolaou,AmitSeru,andNoahSto?man,2012,Technologicalinnovation,resourceallocation,andgrowth,NBERWorkingpaper.

Lach,Saul,andMarkSchankerman,1989,DynamicsofR&Dandinvestmentinthescien-ti?csector,JournalofPoliticalEconomy97,880-904.

Lai,Ronald,AlexanderD’Amour,andLeeFleming,2009,Thecareersandco-authorshipnetworksofUSpatent-holderssince1975,WorkingPaper,HarvardUniversity.

Lanjouw,JeanO.,ArielPakes,andJonathanPutnam,1998,Howtocountpatentsandvalueintellectualproperty:Theusesofpatentrenewalandapplicationdata,JournalofIndustrialEconomics46,405–432.

36

This article is protected by copyright. All rights reserved.

Lerner,Josh,1994,Venturecapitalistsandthedecisiontogopublic,JournalofFinancialEconomics35,293–316.

Lerner,Josh,MortenSorensen,andPerStr¨omberg,2011,Privateequityandlong-runinvestment:Thecaseofinnovation,JournalofFinance66,445-477.

Accepted ArticleLoughran,Tim,andJayR.Ritter,2004,WhyhasIPOunderpricingincreasedovertime?,FinancialManagement33,5-37.

Michaely,Roni,andMichaelR.Roberts,2012,Corporatedividendpolicies:Lessonsfromprivate?rms,ReviewofFinancialStudies25,711-746.

Mikkelson,WayneH.,MeganM.Partch,andKshitijShaw,1997,Ownershipandoperatingdecisionsofcompaniesthatgopublic,JournalofFinancialEconomics44,281-307.Moser,Petra,2007,Whydon’tinventorspatent?NBERWorkingpaper.

Netter,Je?ry,MikeStegemoller,andM.BabajideWintoki,2011,Implicationsofdatascreensonmergerandacquisitionanalysis:Alargesamplestudyofmergersandacquisitionsfrom1992to2009,ReviewofFinancialStudies24,2316-2357.

Pagano,Marco,FabioPanetta,andLuigiZingales,1996,Thestockmarketasasourceofcapital:Somelessonsfrominitialpublico?eringsinItaly,EuropeanEconomicReview40,1057-1069.

Pagano,Marc,FabioPanetta,andLuigiZingales,1998,Whydo?rmsgopublic?Anempiricalanalysis,JournalofFinance53,27-64.

P′astor,Lubos,LucianTaylor,andPietroVeronesi,2009,Entrepreneuriallearning,theIPOdecision,andthepost-IPOdropin?rmpro?tability,ReviewofFinancialStudies22,3005.Ritter,JayR.,andIvoWelch,2002,AreviewofIPOactivity,pricing,andallocations,JournalofFinance57,1795-1828.

Robinson,David,2008,Strategicalliancesandtheboundariesofthe?rm,ReviewofFi-nancialStudies21,649-681.

Saunders,Anthony,andSaschaSte?en,2011,Thecostsofbeingprivate:Evidencefromtheloanmarket,ReviewofFinancialStudies24,4091-4122.

Seru,Amit,2014,Firmboundariesmatter:EvidencefromconglomeratesandR&Dactivity,JournalofFinancialEconomics111,381-405.

Sheen,Albert,2009,Dopublicandprivate?rmsbehavedi?erently?Anexaminationofinvestmentinthechemicalindustry,Workingpaper,UCLA.

Shleifer,Andrei,andRobertW.Vishny,1989,Managemententrenchment::Thecaseofmanager-speci?cinvestments,JournalofFinancialEconomics25,123–139.

37

This article is protected by copyright. All rights reserved.

Solow,RobertM.,1957,Technicalchangeandtheaggregateproductionfunction,ReviewofEconomicsandStatistics39,312–320.

Staiger,Douglas,andJamesH.Stock,1997,Instrumentalvariablesregressionwithweakinstruments,Econometrica65,557-586.

Stein,JeremyC.,1989,E?cientcapitalmarkets,ine?cient?rms:Amodelofmyopiccorporatebehavior,QuarterlyJournalofEconomics104,655–669.

Accepted ArticleTrajtenberg,Manuel,RebeccaHenderson,andAdamJa?e,1997,Universityversuscorpo-ratepatents:Awindowonthebasicnessofinvention,EconomicsofInnovationandNewTechnology5,19–50.

Weild,David,andEdwardKim,2009,Awake-upcallforAmerica.GrantThornton(CapitalMarketsSeries).

38

This article is protected by copyright. All rights reserved.

TableI

SummaryStatistics

Thistableprovideskeysummarystatistics,comparing?rmsthatgopublicwith?rmsthatwithdrawIPO?lingandremainprivate.Allvariablesarede?nedintheAppendix.Averageinnovationmeasuresarecalculatedoverthethreeyearsupto(andthrough)theIPO?lingyear.FinancialinformationandIPOcharacteristicsareatthetimeoftheIPO?ling.Firmexitssuchasacquisition,bankruptcy,andsecondIPOarecalculatedoverthe?veyearsaftertheIPO?ling.*,**,and***indicatethatdi?erencesinmeansarestatisticallysigni?cantatthe10%,5%,and1%level,respectively.

Accepted ArticleCompletedWithdrawnMeanMedianS.D.MeanMedianS.D.Di?erence

InnovationMeasuresintheThreeYearsBeforeIPOFilingCitations12.697.2521.6010.916.0016.831.78ScaledCitations1.891.411.731.801.311.940.09NumberofPatents8.202.0050.067.002.0015.001.21ScaledNumberofPatents2.960.8511.162.720.935.070.24Generality0.450.470.210.460.500.22?0.01Originality0.470.500.210.480.490.23?0.01ScaledBestPatent4.302.895.714.002.494.920.31FinancialInformationatIPOFiling(from1996)LogTotalAssets3.072.910.052.972.93R&D/Assets0.290.210.310.290.19NetIncome/Assets?0.31?0.110.48?0.44?0.21Cash/Assets0.280.200.260.360.32IPOCharacteristics

LeadUnderwriterRankingFirmAgeVCBacked

Post-FilingNASDAQReturnsPre-FilingNASDAQReturnsPioneer

EarlyFollower

0.110.310.470.29?0.090.010.13***?0.08***

8.1611.940.460.030.070.020.059.001.278.179.008.0010.9811.147.000.000.500.511.000.030.11?0.06?0.050.060.120.050.050.000.140.030.000.000.220.070.00IPOFiling0.000.170.000.42--

1.3310.380.500.140.160.170.26?0.010.80?0.05*0.09***0.02***?0.01?0.02

FirmExitsintheFiveYearsAftertheBankruptcy0.02Acquisition0.24SecondIPO-

0.02

0.290.180.000.000.000.150.460.360.00?0.05-

39

This article is protected by copyright. All rights reserved.

TableIIFirstStage

Thistablereportsthe?rst-stageestimationoftheinstrumentalvariablesanalysis.Thedependentvariableisadummyvariablethatequalsoneifa?rmcompletestheIPO?ling,andzerootherwise.NASDAQReturnsisconstructeddi?erentlyacrossspeci?cations.Incolumns(1)to(4),NASDAQreturnsarethetwo-monthreturnsaftertheIPO?lingdate.Incolumns(5)and(6),NASDAQreturnsarecalculatedovertheentirebook-buildingperiod,thatis,fromthedateoftheinitialregistrationstatementtothecompletionorwithdrawaldate.Finally,columns(7)and(8)useadummyvariablethatisequaltooneifa?rmdoesnotexperienceaNASDAQdrop,wherea?rmissaidtohaveexperiencedaNASDAQdropifthetwo-monthNASDAQreturnsfromthedateoftheIPO?lingarewithinthebottom25%ofall?lersinthesameyear.Incolumns(3)and(4)thesampleisrestrictedtoIPO?lingsbefore2000.Whencontrolvariablesareincluded,thefollowingvariablesareaddedtothespeci?cation:three-monthNASDAQreturnspriortotheIPO?ling,numberofpatentsinthethreeyearsbeforetheIPO?ling,VCBacked,Pioneer,andEarlyFollower.Allvariablesarede?nedintheAppendix.ThemodelisestimatedusingOLS,androbuststandarderrorsarepresentedinparentheses.*,**,and***indicatethatthecoe?cientisstatisticallysigni?cantatthe10%,5%,and1%level,respectively.

(1)FullTwoMonths

(2)FullTwoMonths

(3)Pre-2000TwoMonths

(4)Pre-2000TwoMonths

(5)FullAll

(6)FullAll

(7)FullBinary

(8)FullBinary

Accepted ArticleSample

InstrumentNADSAQreturnsObservationsR2

FilingyearFEIndustryFE

ControlvariablesF-statistic

0.704***0.763***0.690***0.723***0.381***0.400***0.106***0.111***(0.102)(0.106)(0.128)(0.132)(0.080)(0.081)(0.022)(0.022)1,8010.138yesyesno47.79

1,8010.149yesyesyes52.03

1,4580.082yesyesno28.9

1,4580.089yesyesyes29.9

1,8010.127yesyesno22.63

1,8010.136yesyesyes24.13

1,8010.124yesyesno24.16

1,8010.134yesyesyes25.99

40

This article is protected by copyright. All rights reserved.

TableIII

NASDAQDropsandFirmCharacteristics

Tablepresentsdi?erencesin?rmcharacteristicsandinnovationperformancebetweenIPO?lersthatexperienceaNASDAQdropandother?lersinthesameyear.A?rmissaidtohaveexperiencedaNASDAQdropifitstwomonthNASDAQreturnsfollowingtheIPO?lingisatthebottomofthedistributionofallIPO?lersinthesameyear.Incolumn(1)Bottom10%referstoall?rmsthatexperiencethelowest10%NASDAQreturnsofallIPO?lerswithinayear,andincolumn(2)Top90%referstotheremaining?rms.Incolumn(4)Bottom25%referstoall?rmsthatexperiencethelowest25%NASDAQreturnswithinayear,andincolumn(5)Top75%referstoallremaining?rms.Allvariablesarede?nedintheAppendix.*,**,and***indicatethatdi?erencesinmeansarestatisticallysigni?cantatthe10%,5%,and1%level,respectively.

Accepted Article(1)(2)(3)(4)(5)(6)

NASDAQReturnsThreshold:BottomTopDi?erenceBottomTopDi?erence

10?%u%

Pre-FilingFinancialInformationTotalAssets3.163.060.1043.133.050.077R&D/Assets0.260.260.0070.260.26?0.002NetIncome/Assets?0.32?0.330.008?0.34?0.33?0.014Cash/Assets0.320.280.0360.300.290.013Sales/Assets0.860.89?0.0240.850.90?0.051IPOCharacterisitics

LeadUnderwriterRankingFirmAgeatFilingVCBacked

8.2211.870.468.0911.810.490.1240.068?0.0298.1911.100.498.0812.050.500.110?0.946?0.011

Pre-FilingPatentCharacteristicsCitations13.38ScaledCitations1.81NumberofPatents8.53ScaledNumberofPatents3.21ScaledGenerality1.11ScaledOriginality1.03ScaledBestPatent4.06

12.48

1.877.922.881.121.074.260.905?0.0700.6030.330?0.020?0.039?0.19712.631.926.972.671.141.064.4512.571.858.322.991.121.074.170.0640.072?1.354?0.3260.023?0.0170.277

41

This article is protected by copyright. All rights reserved.

TableIVPlaceboTest

Thistablereportsaplacebotesttoassessthevalidityoftheinstrumentalvariableexclusionre-striction.Thedependentvariableistheaveragescaledcitationsinthe?veyearsaftertheIPO?ling.ReturnsfollowingIPO?lingarethetwo-monthNASDAQreturnscalculatedfromtheIPO?lingdate.ReturnsfollowingIPOoutcomearethetwo-monthNASDAQreturnscalculatedfromeitherthedateoftheequityissuanceorthedateoftheIPO?lingwithdrawal.WhenthedateofIPO?lingwithdrawalisnotavailable,270dayssubsequenttothelastamendmentoftheIPO?lingisassumed(Lerner(1994)).ReturnsinyearbeforeIPO?lingarethetwo-monthNASDAQreturnscalculatedfromayearbeforetheIPO?ling.ReturnsinyearafterIPO?lingarethetwo-monthNASDAQreturnscalculatedfromayearaftertheIPO?ling.Thevariablesincludedintheregressionsarepre-?lingaveragescaledcitations,pre-?lingnumberofpatents,VCBacked,Pioneer,EarlyFollower,andthethree-monthNASDAQreturnsbeforetheIPO?ling.Allvariablesarede?nedintheAppendix.ThemodelisestimatedusingOLS,androbuststandarderrorsarereportedinparentheses.*,**,and***indicatethatthecoe?cientisstatisticallysigni?cantatthe10%,5%,and1%level,respectively.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Accepted ArticleDependentVariableReturnsfollowingIPO?lingReturnsfollowingIPOoutcomeReturnsinyearbeforeIPO?ling

ScaledScaledScaledScaledScaledScaledScaledCitationsCitationsCitationsCitationsCitationsCitationsCitations?0.498**(0.239)

0.207(0.251)

0.201(0.254)

?0.482**?0.495**?0.509**(0.237)(0.237)(0.241)0.162(0.248)

0.193(0.252)

0.0060.037(0.096)(0.094)10790.239yesyesyes

10790.242yesyesyes

10790.242yesyesyes

10790.242yesyesyes

ReturnsinyearafterIPO?lingObservationsR2

FilingyearFEIndustryFE

Controlvariables

10790.242yesyesyes

10790.240yesyesyes

10790.239yesyesyes

42

This article is protected by copyright. All rights reserved.

TableVReducedForm

Thistablereportsdi?erencesinthe?ve-yearinnovationperformancefollowingtheIPO?lingbe-tween?lersthatexperienceaNASDAQdropandother?lersinthesameyear.A?rmissaidtohaveexperiencedaNASDAQdropifthetwo-monthNASDAQreturnsaftertheIPO?lingarewithinthebottom25%ofall?lersinthesameyear.IPOisadummyvariableequaltooneifa?rmcompleteitsIPO?ling,andzerootherwise.Allvariablesarede?nedintheAppendix.*,**,and***indicatethatthedi?erenceinmeansisstatisticallysignicantatthe10%,5%,and1%level,respectively.

Accepted ArticleNASDAQDropMeanMedianS.D.IPO0.741.000.44ScaledCitations1.120.881.21ScaledNumberofPatents5.561.9112.42ScaledGenerality1.101.100.67ScaledOriginality1.091.090.39ScaledBestPatent3.612.104.66NoNASDAQDrop

MeanMedianS.D.Di?erence0.851.000.36?0.111***0.990.810.860.134**5.911.4916.64?0.3511.101.090.67?0.0051.041.060.430.047*2.891.943.130.721***

43

This article is protected by copyright. All rights reserved.

TableVI

InnovationNovelty

Thistablereportsthee?ectofanIPOoninnovationnovelty.Thedependentvariableistheaveragescaledcitationsinthe?veyearsaftertheIPO?ling.IPOisadummyvariableequaltooneifa?rmcompletetheIPO?ling,andzerootherwise.NASDAQReturnsisthetwo-monthNASDAQreturnscalculatedfromtheIPO?lingdate.Controlvariablesincludedintheregressionsare:pre-?lingaveragescaledcitations,pre-?lingaveragescalednumberofpatentsperyear,VCBacked,Pioneer,EarlyFollower,andthethree-monthNASDAQreturnsbeforetheIPO?ling.Allvariablesarede?nedintheAppendix.Incolumns(1)and(2)themodelisestimatedusingOLS,andincolumn(3)itisestimatedusing2SLS.Incolumn(4)theinstrumentalvariablesapproachisestimatedusingaquasi-maximumlikelihoodPoissonmodel.Inallspeci?cations,marginale?ectsarereported.MagnitudeistheratioofIPOcoe?cienttothepre-?lingaverageaverageofscaledcitations.Robuststandarderrorsarereportedinparentheses.Thestandarderrorsincolumn(4)arecorrectedusingthedeltamethod.*,**,and***indicatethatthecoe?cientisstatisticallysigni?cantatthe10%,5%,and1%level,respectively.

(1)ScaledCitationsOLS?0.019(0.069)

?0.498**(0.239)

-1.02%1,0790.239yesyesyes

-1,0790.242yesyesyes

-43.51%1,0790.128yesyesyes

-52.41%1,0790.148yesyesyes

(2)ScaledCitationsOLS

(3)ScaledCitations2SLS?0.831**(0.409)

(4)ScaledCitationsPoisson?0.980**(0.427)

Accepted ArticleDependentVariableModelIPO

NASDAQReturnsMagnitudeObservationsR2

FilingyearFEIndustryFE

Controlvariables

44

This article is protected by copyright. All rights reserved.

TableVII

FundamentalNatureofResearch

Thistablereportsthee?ectofanIPOonthefundamentalnatureofresearch.Incolumns(1)to(3)thedependentvariableistheaveragescaledoriginalityinthe?veyearsaftertheIPO?ling,andincolumns(4)to(6)itisaveragescaledgenerality.IPOisadummyvariableequalstooneifa?rmcompletestheIPO?ling,andzerootherwise.NASDAQReturnsisthetwo-monthNASDAQreturnscalculatedfromtheIPO?lingdate.Incolumns(1)to(3)Icontrolforthepre-?lingaveragescaledoriginality,andincolumns(4)to(6)Icontrolforthecorrespondinggeneralitymeasure.Additionalcontrolvariablesare:pre-?lingaveragescaledcitations,pre-?lingaveragescaledpatentsperyear,VCBacked,Pioneer,EarlyFollower,andthethree-monthNASDAQreturnsbeforetheIPO?ling.Allvariablesarede?nedintheAppendix.ThemodelisestimatedusingOLSincolumns(1),(2),(4),and(5)andusing2SLSincolumns(3)and(6).MagnitudeistheratioofIPOcoe?cienttothepre-?lingaverageofscaledoriginalityorscaledgenerality.Robuststandarderrorsarereportedinparentheses.*,**,and***indicatethatthecoe?cientisstatisticallysigni?cantatthe10%,5%,and1%level,respectively.

(1)(2)(3)(4)(5)(6)

DependentVariableScaledScaledScaledScaledScaledScaled

OriginalityOriginalityOriginalityGeneralityGeneralityGenerality

ModelOLSOLS2SLSOLSOLS2SLSIPO

NASDAQReturnsMagnitudeObservationsR2

FilingyearFEIndustryFE

Controlvariables

-0.10%1,0790.231yesyesyes?0.006(0.010)

?0.081**(0.036)

-1,0790.234yesyesyes

-13%1,0790.102yesyesyes

0%1,0790.226yesyesyes

?0.137**?0.001(0.068)(0.016)

?0.050

(0.051)

-1,0790.226yesyesyes

-8%1,0790.206yesyesyes?0.087(0.092)

Accepted Article45

This article is protected by copyright. All rights reserved.

TableVIIIInnovationScale

Thistablereportsthee?ectofanIPOoninnovationscale.Thedependentvariableistheaveragescalednumberofpatentsperyearinthe?veyearsaftertheIPO?ling.IPOisadummyvariableequalstooneifa?rmcompletestheIPO?ling,andzerootherwise.NASDAQReturnsisthetwo-monthNASDAQreturnscalculatedfromtheIPO?lingdate.Controlvariablesincludedinregressionsare:pre-?lingaveragescaledcitations,pre-?lingaveragescalednumberofpatentsperyear,VCBacked,Pioneer,EarlyFollower,andthethree-monthNASDAQreturnsbeforetheIPO?ling.Allvariablesarede?nedintheAppendix.ThemodelisestimatedusingOLSincolumns(1)and(2),and2SLSincolumn(3).Columns(4)estimatesthespeci?cationusingaquasimaximumlikelihoodPoissonmodel.Inallspeci?cations,marginale?ectsarereported.MagnitudeistheratiooftheIPOcoe?cienttothepre-?lingscalednumberofpatentsperyear.Robuststandarderrorsarereportedinparentheses.*,**,and***indicatethatthecoe?cientisstatisticallysigni?cantatthe10%,5%,and1%level,respectively.

(1)ScaledPatentsOLS0.268***(0.066)

(2)ScaledPatentsOLS

(3)ScaledPatents2SLS0.200(0.474)

(4)ScaledPatentsPoisson0.002(0.662)

Accepted ArticleDependentVariableModelIPONASDAQReturnsMagnitudeObservationsR2

FilingyearFEIndustryFE

ControlVariables

0.127(0.305)

37.75%1,8010.184yesyesyes

1,8010.178yesyesyes

28.17%1,8010.184yesyesyes

0.28%1,8010.168yesyesyes

46

This article is protected by copyright. All rights reserved.

TableIX

InventorMobilityandInnovativeProductivity

Thistablereportsthee?ectsofanIPOoninventors’mobilityandinnovativeactivity.Inventorsareclassi?edintothreecategories:stayers,leavers,andnewcomers,asde?nedinthetext.Incolumn(1)thesampleisrestrictedtostayersandthedependentvariableistheaveragescaledcitationsaftertheIPO?ling.Incolumns(2)and(3),thesampleincludesstayersandleavers,andthedependentvariableequalsoneiftheinventorleavesthe?rmorgeneratesaspin-o?,respectively.Anassigneeisconsideredaspin-o?ifthenumberofappliedpatentsbeforetheleaver’spatentiszero.Incolumn(4)thesampleincludesstayersandnewcomers,andthedependentvariableequalsoneiftheinventorjoinsthe?rm.IPOisadummyvariableequaltooneifa?rmcompletestheIPO?ling,andzerootherwise.Theinstrumentisthetwo-monthNASDAQreturnscalculatedfromtheIPO?lingdate.Inallspeci?cationsIcontrolfortheinventor’spre-IPO?lingaveragescaledcitationsandscalednumberofpatents.Additionalcontrolvariablesare:VCBacked,Pioneer,EarlyFollower,andthethree-monthNASDAQreturnsbeforetheIPO?ling.Allvariablesarede?nedintheAppendix.Allmodelsareestimatedusing2SLS.MagnitudeistheratiooftheIPOcoe?cienttothepre-?lingaverageofscaledcitations.Robuststandarderrorsarereportedinparentheses.*,**,and***indicatethatthecoe?cientisstatisticallysigni?cantatthe10%,5%,and1%level,respectively.

Accepted Article(1)(2)(3)DependentVariableCitationsofLikelihoodLikelihood

StayersofLeavingofaSpin-o?

Model2SLS2SLS2SLSIPO-1.094**0.183***0.095**(0.457)(0.062)(0.048)

Magnitude-47.94%-Observations665787738773R20.2030.0170.03FilingyearFEyesyesyesIndustryFEyesyesyesControlVariablesyesyesyes(4)LikelihoodofHiring2SLS0.388***(0.078)-116780.058yesyesyes

47

This article is protected by copyright. All rights reserved.

TableX

AcquisitionofExternalTechnologies

Thistablereportsthee?ectsofanIPOonthenumberofexternalpatentsacquiredthroughmergersandacquisitions.Thedependentvariableistheaveragenumberofexternalpatentsacquiredperyearinthe?veyearsaftertheIPO?ling.IPOisadummyvariableequaltooneifa?rmcompletestheIPO?ling,andzerootherwise.NASDAQReturnsisthetwo-monthNASDAQreturnscalculatedfromtheIPO?lingdate.Controlvariablesincludedinregressionsare:pre-?lingaveragescaledcitations,pre-?lingaveragescalednumberofpatentsperyear,VCBacked,Pioneer,EarlyFollower,andthethree-monthNASDAQreturnsbeforetheIPO?ling.Allvariablesarede?nedintheAppendix.ThemodelisestimatedusingOLSincolumns(1)and(2),andusing2SLSincolumn(3)and(4).Column(4)alsocontrolsforindustryacquisitionintensity,de?nedasthetotalvolumeofacquisitionswithinanindustrynormalizedbythetotalmarketvalueofall?rmsinthatindustry.Robuststandarderrorsarereportedinparentheses.*,**,and***indicatethatthecoe?cientisstatisticallysigni?cantatthe10%,5%,and1%level,respectively.

Accepted ArticleDependentVariableModelIPONASDAQReturnsObservationsR2

FilingyearFEIndustryFE

Controlvariables

AcquisitionPropensity

(1)ExternalPatentsOLS0.630***(0.154)

(2)ExternalPatentsOLS

(3)(4)ExternalExternalPatentsPatents2SLS2SLS2.603**2.489*(1.277)(1.288)

1,8010.366yesyesyesno

1.636**(0.785)1,8010.365yesyesyesno

1,8010.347yesyesyesno1,8010.350yesyesyesyes

48

This article is protected by copyright. All rights reserved.

Accepted ArticleFigure1.QualityofinnovationaroundtheIPOevent.The?gurepresentschangesinpatentquality,asmeasuredbyScaledCitations,intheyearsaroundtheIPO.Theestimatesandcon?denceintervalsaretakenfromthefollowingspeci?cation:

Yit=β0+

k=5??k=?3k=0

γkEventYeari,k+τi+μt+εi,t.

TheunitofobservationisatthepatentlevelandthedependentvariableisScaledCitations.EventYeari,kisadummyvariableindicatingtherelativeyeararoundtheIPOinwhichapatentapplicationwassubmitted(yearzeroistheyearoftheIPOandtheomittedcategory).Thespeci?cationisestimatedusingOLSandincludes?rm?xede?ects(τi)andyear?xede?ects(μt).Standarderrorsareclusteredatthe?rmlevel.TheestimatesarereportedintheInternetAppendix.

49

This article is protected by copyright. All rights reserved.

本文来源：https://www.bwwdw.com/article/7cew.html