DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATIONINAGLOBALPATENTINGENVIR - PDF document

DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATIONINAGLOBALPATENTINGENVIRONMENT?ACross-CountryAnalysisofPharmaceuticalPatentProtection,1978–2002YiQian*—Thispaperevaluatestheeffectsofpatentprotectiononphar-maceuticalinnovationsfor26countriesthatestablishedpharmaceuticalpatentlawsduring1978–2002.Controllingforcountrycharacteristicsthroughmatchedsamplingtechniquestoestablishtwopropercomparisonsetsamong92sampledcountriesandthroughcountry-pairxed-effectsregressions,thisstudyyieldsrobustresults.Nationalpatentprotectionalonedoesnotstimulatedomesticinnovation,asestimatedbychangesincitation-weightedU.S.patentawards,domesticR&D,andpharmaceuticalindustryexports.However,domesticinnovationacceleratesincountrieswithhigherlevelsofeconomicdevelopment,educationalattainment,andeconomicfreedom.Additionally,thereappearstobeanoptimallevelofintellectualpropertyrightsregulationabovewhichfurtherenhancementreducesinnovativeactivities.I.IntroductionINCEthe1980s,intellectualpropertyrights(IPR)protectionhasbecomemuchmoreextensiveascoun-triesatvariousstagesofdevelopmentbegantoimple-mentorextendtheirnationalpatentrights.ThequestionofwhethernationalIPRlegislationcouldstimulateenoughinnovationtojustifytheeconomic,political,andsocialcostsassociatedwithitsimplementationanden-forcementstillprovokesheateddebate,andhaspertinentpolicyimplications.Thispaperstudieswhetheranation’simplementationofpharmaceuticalpatentsstimulatesdo-mesticpharmaceuticalR&Dexpendituresandinnova-tions,asmeasuredbyU.S.patentsawardedtoresidentsofthatcountry.Becausewecannotobservecounterfac-tualoutcomesofpatentprotectionwithincountries,in-ternationalcomparisonsprovidevaluableleveragefortestingthehypothesis.Thepassageofnationalpharma-ceuticalpatentlawsinanumberofcountriesinthe1980sand1990screatedanaturalexperimenttotesttheeco-nomicimpactofpatents.Iidentiedthesecountriesandcomparablecontrolcountrieswithdataavailablefrom1978to2002.Thisstudyseekstoovercomedataandmethodologicalconstraintsthathaveconnedpreviousresearchpredominantlytosingle-countryanalyseswithinconclusiveresults(Pazderka,1999andMcFetridge,1996versusScherer&Weisburst,1995andChallu,Themainndingsofthisstudyarethatinthegroupofsampledcountriestheimplementationofpatentlawsbyitselfdoesnotpromptlystimulatedomesticinnovation.However,patentlawsinnationswithhighlevelsofdevel-opment,education,andeconomicfreedomdostimulateinnovation.Thisstudyalsoprovidesnovelempiricalsup-portforthetheorythattherelationshipbetweeninnovationandtheIPRstrengthhasan“invertedU”shape(Gallini,1992;Horwitz&Lai,1996).AnoptimallevelofIPRappearstoexist,abovewhichadditionalstrengtheningac-tuallytendstodiscourageinnovation.OneoftherationalesforpatentprotectionisthatgrantingexclusiverightstoinnovatorswillenablethemtoreapthebenetsandrecoupthecostsofR&Dinvestments,increasingtheirincentivestoinnovate.TheactualeffectofIPRoninnovation,however,remainsoneofthemostcontroversialquestionsintheeconomicsoftechnology.Secrecywasfoundtobemuchmoreimportantthanpatentsforprotectingintellectualproperty,ina1994surveyof1,478Americanmanufacturingrms(Cohen,Nelson,&Walsh,2000).Patentsmayevenbecounter-productive,incurringadditionalapplicationcostsandpromotinglitigationandwastefulattemptstoinventaroundpatents(Jaffe&Lerner,2004).Patentlawscouldalsodelayspillovereffectsinsequentialinnovations,whereeachinnovationisbuiltuponitspredecessors,byfosteringhighlicensingfeesandracesforlicensing(Scotchmer&Green,1990).WhileanegativecorrelationbetweentighteningIPRandinnovationwasfoundem-piricallyinBessenandMaskin(2000)andSakakibaraandBranstetter(1999),itwasnotsupportedinKortumandLerner(1998).AlthoughaseriesofsurveysconductedintheUnitedStates(Manseld,Schwartz,&Wagner,1981;Levinetal.,1987)andSwitzerland(Harabi,1997)uniformlyestablishtheimportanceofpatentsforpharmaceuticalinnovationsrelativetootherindustries,itisnotclearhowmuchpatentprotectionisoptimal.TheoreticalmodelspredictthatmorenationalpatentprotectionindevelopingcountriesmaynotaddmuchtoR&Dinvestmentincentives,giventheexistingReceivedforpublicationNovember19,2002.RevisionacceptedforpublicationMay5,2006.*KelloggSchoolofManagement,NorthwesternUniversity.IwouldliketogivemyspecialthankstoProfessorsRichardCaves,JoshLerner,andDonaldRubinfortheirconstantadviceandencouragement;toLeeBranstetter,GaryChamberlain,RichardCooper,DaleJorgenson,MarkusMobius,ArielPakes,ZorinaKhan,PierreSauve,DanBenjamin,DavinChor,GopalGaruda,JenniferHill,PingMa,JesseShapiro,KiaSong,HuiXie,StanleyWatt,andNBERconferencesparticipantsforcommentsonvariousdrafts;toDrs.ArvindSubramanian,JayashreeWatal,GaryHufbauer,andKeithMaskusfortheiradviceandreferencesatearlystagesofmyresearch;toAndreiShleifer,PaulBeamish,FritzFoley,andmyfriendMichaelKatzforprovidingdata;andtoProfessorDaronAcemogluandthethreeanonymousrefereesfortheirinvaluableAtleastfortydevelopingcountrieslackedpharmaceuticalproductpatentprotectionasofthelate1980s.Bytheendof1999,however,onlysixteenWorldTradeOrganization(WTO)membercountriesexcludedpharmaceuticalsfromnationalpatentprotection.“Inventingaroundapatent”occurswhenimitatorsattempttoavoidpatentprotectionandlicensingrulesbymakingsmallmodicationsontheoriginalinnovation.Thedisclosureoftechnicaldetailsrequiredinpatentshelpsthisactivity.TheReviewofEconomicsandStatistics,August2007,89(3):436–4532007bythePresidentandFellowsofHarvardCollegeandtheMassachusettsInstituteofTechnology worldintellectualpropertyregime(Chin&Grossman,1990;Deardorff,1992;Helpman,1993).Counterargumentsassertthatadditionalpatentlawsprovideafavorablelocalenvironmentforpursuitofinnovationsbydomesticinven-tors,whohaversthandknowledgeofcountry-specicdiseases.Nationalpatentlawswouldalsoinducedomesticinvestorstoswitchfromimitativeactivitiestoinnovativeones.Theimportanceofpatentprotectioninadevelopingcountryisleastcontroversialfortreatmentsofdiseasesfoundonlyinthatcountry.Inaddition,onecountry’sna-tionalpatentcannotalwaysprovideenoughmarketincen-tiveforinnovatorstodevoteresearchresourcesforthatcountry.Thisimpliesthatagroupofcountrieswithsimilartherapeuticneedscouldimplementpatentlawstogether,whichisexactlythepositionthePharmaceuticalResearchandManufacturingAssociationofAmericaheldduringtheTRIPsnegotiations.AlthoughAcemogluandLinn(2003)foundthatanincreaseinthepotentialmarketsizeforadrugcategorymayaugmentthenumberofnewdrugsapprovedbytheFoodandDrugAdministrationintheUnitedStates,thereisnoevidencesofarthatresearchinvestmentsorinnovationsintropical-diseasedrugshaveincreasedsignif-icantlyafterdevelopingcountriesimplementedpatentlaws(Lanjouw&Cockburn,2000).Costsofpatentimplementa-tionhavebeenidentied,includingthecaptureofnationalpatentmonopolyrightsmainlybyforeigners(Lanjouw,1998;Maskus,2000),andthelegaladministrationandlitigationcostsofapatentsystem(UNCTAD,1996;Love,2001).CumbersomelegalsystemsmaydiscourageR&DVerylittleresearchhasbeendonefornon-OECDcoun-tries,mainlybecauseofthedifcultyincollectingdata.Furthermore,somepreviousstudiesofIPRindevelopingcountriespooleddataacrossindustries.Thisisproblematicbecausepatentsmayhavedifferenteffectsindifferentindustries(Levinetal.,1987).Instandardeconomicanal-yses,theeffectofpatentprotectiononinnovationisesti-matedsimplybycomparingthelevelofinnovationincountrieswithandwithoutpatentlaws,controllingforcountrycharacteristicsusuallybyOLS.Thisstudymakesseveralcontributionstotheliterature.First,itgoesbeyondcross-sectionstudiesbyassemblingapanelofdataconsist-ingof92countriesovertwodecades.Thepanelstructureenablesmetobettercontrolforthelatentinnovativepoten-tialofacountryandtoimprovetheprecisionandreducethebiasinthepatenteffectestimator.Thepanelhastheadvan-tageoftestingpre-andpost-patentinnovationchangeswithinthesamecountry.Inaddition,thisstudyonmultiplecountriesmakestheresultsgeneralizableandovercomesthelimitationsintheprevioussingle-countryanalyses.Asecondcontributionofmystudyistoemploynonpara-metricmatchingmethods,whicheasilyaccommodatearichsetofcontrolcovariatesthatarecorrelatedwithacountry’slatentinnovativepotentialandpatentimplementation.Al-thoughcountrieswithpatentprotection,manyofthemdevelopednations,tendtohavehigherinnovationlevels,patentprotectionneednotbethecausalfactor.Countrieswithstrongpatentprotectionmaysimplyhaveagreatercapacityforinnovation.Thefactorsthatdetermineacoun-try’sinnovativepotentialarelikelycorrelatedwithnationalpatentpolicies.Failingtocontrolforthiscorrelationwouldyieldbiasedestimatesofpatenteffects.Previousresearchhasindeedfoundcorrelationsbetweennationalpatentleg-islationandmarketopenness,laggedR&Dexpenditure,GDPpercapita(Acemogluetal.2004),economicgrowth(Evenson,1990),andthelegaloriginofacountry’scom-merciallaws(Lerner,2000b).Itspharmaceuticalindustry’scharacteristicsshouldalsoaffectanation’sdecisiontoimplementpatentlawsanditsinnovativepotential(Kaufer,1989).Controllingforthesecovariatessubstantiallymiti-gatesthepotentialforomittedvariablebias.OLSfacesadegrees-of-freedomproblembecauseobservationswithmissingdataareautomaticallydroppedinestimation,whileMahalanobismatchingovercomesthisdifculty.Thelastcontributionliesintheprocedureforappropri-atelycontrollingtheseobservedcountrycharacteristics.Anadditiveregressionequationappliedtotheentiresampleessentiallycontrolscovariatesbyforcingthesamelinearrelationshiponcountriesfromthecontrolandtreatedgroups.Becausepatentandno-patentcountriesdiffersub-stantiallyincertaincharacteristics(tables2,3),standardlinearregression,whichassumesthesamelinearrelation-shipbetweeneachcontrolvariableandtheoutcomevariableforallobservations,impliesanextendedextrapolationacrosscountrygroupsandthereforemakestheresultantOLSestimatesextremelysensitivetoregressionspecica-PreliminaryanalysesshowthatOLScoefcientsjumpfrom7.34to3.74whenusingrawpatentsasre-sponsevariable,andfrom1.86to0.85whenusinglogofcitation-weightedpatentsasresponsevariable,butallsta-tisticallyinsignicant.Matchingmethodshavebeenshowntoreducetheseconfoundingvariablebiases(Rosenbaum&Rubin,1984;Heckmanetal.,1996),bybalancingtherelevantpretreatmentcountrycharacteristicsofthecontrolandtreatedgroups.Thisstudycontrolscovariatesthroughatwo-stepprocess:therststepidentiespairsofcountrieswithsimilarcharacteristicsthroughMahalanobismatching,andthesecondstepperformspairwiseeconometricanalysesonthematchedpairs.Fixed-effectsregressionsonmatchedcountrypairscontrolthoroughlyforunobservedcountryIfthereisonlyonecontrolvariable,thenthisproblemcanpossiblybeovercomebyaddingasetofhigher-ordertermsofthecontrolvariableandinteractiontermsofthecontrolandthepatentindicatorvariablesuntilthelinearassumptionissatised.However,thisdoesnotworkifthereislimitedoverlapinthecovariatedistributions.Moreover,whenthereisalargesetofvariablestobecontrolled,asisthecasehere,addingtermsforeachcontrolvariabletakesawayalreadylimiteddegreesoffreedomandbecomesunfeasible.Evenwithalargesample,itwouldbedifculttomodelresponsesurfacesinhighdimensions,becauseitisespeciallydifculttoassesslinearityinahigh-dimensionalcovariatespace.DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?437 Therestofthepaperisstructuredasfollows.SectionIIdescribesthestudydesignandtheconstructionofthedataset(supplementedbythedataappendix).Themethodolo-giesaredescribedinsectionIII.SectionIVpresentsthemainempiricalresults.Finally,sectionVsummarizesthemainapproachandresultsandmakesrecommendationsforpoliciesandfuturestudies.II.DataA.DesignoftheStudyInansweringwhethernationalpatentsindevelopingcountriesstimulatedomesticinnovations,theidealex-perimentwouldrandomlyassignasetofcountriesinthepoolofno-patentcountriestoimplementnationalpatentlawsatacertaintime.Theeffectsofpatentsoninnova-tioncouldbetestedsimplybycomparingmeaninnova-tionlevelsofthepatentandno-patentcountries.Unfor-tunately,inrealitypatentreformsdifferacrossnationsintermsofboththetimeandthedegreeofimplementation.Table1liststheyearwheneachsampledcountryimple-mentedpharmaceuticalpatents.Theyweregroupedinfourperiods:1978–82,1983–85,1986–90,1991–95.Mostnewlypatent-grantingcountrieshadprovidedsomedegreeofprotectionforpharmaceuticalprocessesbeforetheyformallyintroducedproductpatents.AnindicatorPATMODisconstructedtodistinguishthesecountries(PATMOD1)fromthosethatimplementedpatentsforthersttime(PATMODSecondly,althoughthedecisiontoadoptthepatentprivilegecanberegardedasexogenoustotheextentthatmanyofthemaremadeunderpersistentpressuresfromtheWest,somevoluntarydecision-makingisinvolved.Thatisdemonstratedbythedifferencesinscopetimingoflegislation.Countrycharacteristicsmayaffectboththeinnovationoutcomeandthedecisiontoimple-mentnationalpatents.Thisendogeneityimpliesthattheconventionalmethodofregressingtheoutcomevariableonthepatentindicatorandcountrycovariateswouldverylikelyproduceabiasedestimateoftheeffect.Thereisnosoundinstrumentalvariabletoaddressthisconcern.In-stead,thisstudyappliesamatchingmethodtoformcountrypairsforwhichpatenttreatmentcanbeconsid-eredrandomlyassignedwithineachpair.Idenetreat-mentastheimplementationofnationalpharmaceuticalpatentlaws(dummyvariablePAT1)duringtherefer-enceperiod,andthecontrolasnochangeinpatentlawPAT0).Twocontrolgroupsaredenedinordertomakethemostuseofthesamplesizeavailable.Onecontrolgroupconsistsofcountriesthatneverhadpatentprotectionpriortothenextreferenceperiod,theotherofcountriesthathadpatentprotectionbeforethereferenceperiod.Eachnew-patentcountryispairedwithonecoun-tryfromthenever-patentgroupandanothercountryfromthealways-patentgroup.Fixed-effectsregressionanaly-sesarethencarriedoutseparatelyonthesetwosetsofmatchedpairs,whereOLSlinearassumptionsarebettersatisedduetothemorebalancedcovariates.Thetwo-waycomparisonbasedonthesetwocontrolgroupscanhelptodetectthepotentialbiasarisingfromthecountry-specicfactorsinthematchedpairs.Thisiscomputationallyequivalenttoregressingtheinnovationoutcomevariableonthebinaryvariableofpatentprotection.Controllingforallothercountrycovariatesisunnecessaryforobtaininganunbiasedestimatebecauseofrandomization,butitassistsinobtaininggreaterprecisionoftheexperimentalestimates.BecausetheOECDR&Dexpendituredata(innovationoutcomevariables)areonlyavailableupuntil1999,thecountriesthatswitchedpatentlawsin1999willnotbeexamined.Somecountrieshesitateaboutwhethertopatent.Thevaluesforthecontrolvariablesinthepreviousseveralyearscouldbeasimportantasthevaluesintheyearbeforelegislationinaffectingacountry’sdecisions.Inaddition,somecontrolvariables,suchasaverageyearsofschoolingandeconomicfreedomindicesareavailableonlyatve-yearintervals.Ithereforeusetheaveragedvaluesoverthepreviousthreetoveyears.Thisalsohelpstosmoothoutanyoutliervaluesinaparticularyear.Thelaggedone-yearvaluesarealsousedinrobustnesschecks,andresultsaresimilar.TheonlyexceptionsareBrazil,China,Chile,Korea,Indonesia,Mexico,Peru,Romania,Taiwan,Thailand,andTurkey.Forinstance,Chinarstimplementednationalpatentlawsin1983,butexcludedthepharmaceuticalsectoruntil1992.AIRSANDIMINGOFATCHEDWITHTHE YearofPatentPeriod1(1978–82)1983DenmarkNorwaySwedenPeriod2(1983–85)1986TaiwanHungaryHongKong1987CanadaNorwayNetherlands1986KoreaThailandSingapore1987AustriaFinlandAustraliaPeriod3(1986–90)1993BrazilArgentinaKorea1991ChileUruguayPanama1991ChinaIndiaKorea1992SpainArgentinaBelgium1995FinlandSloveniaAustralia1992GreecePolandSingapore1992HungaryRomaniaIsrael1992–1993IndonesiaEgyptPhilippines1991MexicoArgentinaKorea1992NorwaySloveniaAustralia1992PortugalRomaniaHongKong1992–1993ThailandColombiaPhilippinesPeriod4(1991–95)1996BoliviaParaguayZimbabwe1996ColombiaEgyptPhilippines1997GhanaJordanKenya1996IcelandSloveniaLuxembourg1996PeruGuatemalaAlgeria1996or1997TurkeyIranSouthAfrica1997RomaniaBulgariaChile1996EcuadorTunisiaElSalvador1996VenezuelaCostaRicaChileThistableprovidesinformationaboutthetimingofreformsinthecountriesthatstartedtoimplementtheirdomesticpharmaceuticalpatentprotectionsduring1980–1997,andthecorrespondingmatchedcountriesthatneverhadpharmaceuticalpatentlawsinthecontemporaryperiodoralwayshadpharmaceuticalpatentlawsevenpriortothecontemporaryperiod.Patentimplementationyearsareobtainedbyreferencingthelegaldocumentsofeachcountry’sintellectualpropertyrightsofce.THEREVIEWOFECONOMICSANDSTATISTICS B.SelectionoftheOutcomeVariablesandConstructionoftheControlVariablesFollowingpreviouscross-countryanalysesofinventiveactivities,thisstudyusescitation-weightedU.S.pharma-ceuticalpatentawards(listedbyapplicationdates)asthemaininnovationproxy.Patentdataarelistedbycountryofresidenceoftherstlistedinnovator.Thisinformationprovidesauniformbaseforcomparison,becauseU.S.patentlawhastreatedapplicationsfromdifferentcountriesinanondiscriminatoryandconsistentmannersincethePapersbyEvenson,Griliches,Pakes,andothers(Griliches,1984)suggestthatpatentcountsandR&Dex-penditurearehighlycorrelatedincrosssection.StudiesalsoshowconcordancesintheshiftsofR&Dexpenditureandpatentinglevel(Kaufer,1989).IntheOECDdataset,containingreliableR&Ddata,thecorrelationbetweenthedifferenceinthelogofU.S.patentsawardedtoacountryinconsecutiveyearsandthatofitsR&Dexpendituresisaround0.8.Thus,“toarstapproximation,onecanusepatentdataasanindicatoroftechnologicalactivityinparallelwithorinlieuofR&Ddata”(Griliches,1984,p.14).Manseld(1986)ndsinhis1981–1983surveyof100U.S.rmsthataround82%ofpatentableinventionsinpharmaceuticalswerepatented.Althoughpatentsawardedaregoodindicatorsofinnova-tions,thevalueofinnovationsisnotaccuratelymeasuredbypatentcounts,becauseoftheexistenceofasymmetricin-formationbetweeninnovatorsandpatentofces.Thecita-tionweightscouldservetoovercomesuchproblems(Hall,Jaffe,&Trajtenberg,2001).TheNBERpatentdatabasecontainsthenumberofcitationsmadetoeachpatentgrantedbytheU.S.patentofcefrom1960to2002.FollowingTrajtenberg(1990),Icalculatethecitation-weightedpatentcountsbysumming(1overallthepharmaceuticalpatentsawardedtoacountryinagivenyear,isthecitationmadetopatentNotallforeigninnovationsarepatentedintheUnitedStates.Theinnovator’sdecisiontoapplyforapatentintheUnitedStatesdependsonmanyfactors,includinggeo-graphicdistancefromtheUnitedStates,marketpotentialoftheirinventionintheUnitedStates,andsoforth.Becausethesefactorsarealsorelatedtotradebetweenaninnovator’scountryofresidenceandtheUnitedStates,theirbiasmightbecorrectedbycontrollingforpharmaceuticalexportstotheUnitedStates.GiventhattheUnitedStatesmarketistheworld’slargest,innovationsofmorethanlocalsignicancetendtobepatentedintheUnitedStatesiftheyarepatentedatall(Scherer&Weisburst,1995).Sincethecostandstandardofpatentlingarehigh,U.S.patentdataareexpectedtocaptureonlysubstantialinnovations,anditservesasanaturalselectionofonlyimportantinnovations,appropriateforthisresearch.Thisalsohelpstomakethelevelsofinnovationcomparableacrossyears,becauseanygivencountry’simportantinnovationswouldbepatentedintheUnitedStatesbothbeforeandaftertheimplementationofdomesticpatents.Thecountry-specicpropensitytopatentintheUnitedStatesisfurthercontrolledbytheconstructionofan“innovativepotential”variable,acate-goricalvariableindicatingatwhatlevelthecountryhasbeenawardedU.S.patentsinallindustriesexceptpharma-Inaddition,thisstudyteststhechangeinU.S.patentsduetonationalpatentlaws,insteadofabsolutenumbersofpatents.OtheradvantagesandconcernsofusingtheU.S.patentdataandthecorrespondingrobustnesschecksarediscussedinappendixI.B(appendicesI.B–IIIareavailableathttp://www.nber.org/nyiqian/patentappend.pdf).PharmaceuticalR&Dexpenditures,analternativemea-sureofinnovation,areavailableonlyfor23OECDcoun-tries.Ithereforeapplyregressionmodelstoonlythesecountrieswhentestingtherelationshipbetweenpatentim-plementationandR&Dexpenditures(sectionIIIBb).DataonR&DpersonnelfortheOECDcountriesarealsousedasanotherproxyforinnovation.Forrobustnesschecks,IimputedR&Dvaluesforthenon-OECDcountriesusingaregressionmodelasdescribedinappendixI.C,andcarriedoutanalysesontheseimputedvalues.Pharmaceuticalex-portstotheUnitedStatesareusedasanalinnovationmeasureintheappendix.C.ControlVariablesforLatentInnovativePotentialInordertoobtainunbiasedestimatedcoefcientsofthekeyindependentvariables(PATPATMODasspeciedinsectionIIA),onewouldhopetocontrolforcountries’differentinnovativepotentials.EconomistshavespeculatedwidelyoncountrycharacteristicsthatmightrelatetolatentU.S.pharmaceuticalpatentlawhasnotchangedmuchintheperiodIamexamining,exceptfortwomodicationsin1995.TheU.S.extendedthedurationofpharmaceuticalpatentsfromseventeenyearstotwentyyearsandmodiedtheinterference(whenmultiplepatentapplicationsmakesimilaroridenticalclaims)rule.After1995,innovativeactivitiesinforeigncountriesalsocountasvalidevidenceforestablishingtherstinventiondateincasesofinterference.Althoughtheprobabilityofhavinginterferencecasesisverylow—therewerethirteeninterferencecasesdeclaredoutof2,000applicationsin1999(USPTO,2000,interviewofPaulHarrison)—thethreatoflitigationcanstillhaveimportantimplica-tions.ThismaypartlyexplainthegenerallyincreasingtrendofforeignpatentingintheUnitedStates.Fortunately,thischangedoesnotinuencetheinternationalcomparativeanalysesthisstudycarriesout,becauseitaffectsinnovationsfromallforeigncountriesequally.Putnam(1996)showsthataround63.9%ofinternationalpatentsin1975(thosepatentsledinatleasttwocountries)arepatentedintheUnitedStates.Basedonthedatacollectedinthisstudy,thenumberofU.S.patentawardsisaboutthreetimesthatoftheEuropeanPatentOfce(EPO)patentapplicationsonaverageforaparticularcountryandyear.TheratioofnumberofpatentsawardedintheUnitedStatestothatinthedomesticcountryiscalculatedtobeintherangeof0.83and11,usingtheU.S.PatentandTrademarkOfce(USPTO)andWorldIntellectualProp-ertyOrganization(WIPO)dataforaparticularcountryandyear.TheU.S.patentcountsarebiggerthanthedomesticpatentcountsinsomecases,possiblybecauseittakeslongertoprocesspatentapplicationsinthesecountriesthanintheUnitedStates.Theipsideisthatthisestimatelosesinformationonnewinnovationsthatmightbelocallysuccessful.Thepharmaceuticalpatentsareexcludedtoavoidinterferencewiththepharmaceuticalpatentsoutcomevariable.DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?439 innovativepotentialandthedecisiontoimplementdomesticpatentlaw.GDP,GDPpercapitaPPP,GDPgrowthrates,andeducationalattainmentaretheobviousconfoundingcharacteristics.WhileIleavethedetailsofthecontrolcovariatestotheappendix,thissectionmotivatessomenonobviouscontrolvariableconstructions.Ameasureofeconomicfreedomisincludedtoindicateacountry’smarketandtradefreedomsandlegalandnancialsystems’developments.AsAcemoglu,Johnson,andRob-inson(2004)havepointedout,economicinstitutionsinu-encetheinvestmentsofphysicalandhumancapitalandtechnologies,theorganizationofproduction,andinturninnovativeactivities.Theeconomicfreedomindexisalsocorrelatedwithacountry’spatentimplementationdecision,sinceprotectionofIPRrightsisacomponentformaintain-ingmarketorder.Inrobustnesschecks,Ifurthercontrolfortheoriginofthecountry’slegalsystem.Lerner(2000b)foundthatcountrieswithBritishandFrenchcommerciallawlegaloriginsaremorelikelytohavenationalpatentlaws,andBritishlegaloriginisalsofoundtobeconduciveforappropriatingthereturnsoninvestment(LaPortaetal.,Relevantpharmaceuticalindustrycharacteristicsdifferacrosscountriesandcouldaffectboththeoutcomesofinnovationandthedecisiontoimplementpatentlegislation,asreectedbythefactthatsomecountriesdecidetoexcludethepharmaceuticalindustryfromtheirnationalpatentlaws.Iusethesector’semploymentleveltonormalizeindustrysizeacrossdifferentcountries.Becausethetransferoftechnologyfromabroadcouldhaveanimpactondomesticinnovationandcountries’decisionsofpatentimplementa-tion,estimatesoftechnologytransferthroughforeigndirectinvestment(FDI)arealsoincluded.ThemostrelevantdatawouldbethetotalFDIreceivedinacountry’spharmaceu-ticalindustry,whichisnotavailable.ByincludingbothU.S.andJapaneseFDI,geographicproximityinFDIlocationscouldbecontrolledfortosomeextent.Itisworthemphasizingthatthisstudyusesthelagged(pre-patentperiod)valuesforallthecontrolcovariates,becausetheyarelikelytobeaffectedbynationalpatentD.InteractionVariablesfortheConditionalImportanceofPatentImplementationCountriesatdifferentdevelopmentlevelsdifferinotherlatentfactorsthatcouldaffecttheirR&DorU.S.patent-lingresponsestodomesticpatentprotection.Onesuchnotablefactoristechnologyinfrastructure.Maskus(2000,p.202)suggeststhatdevelopedcountriesandmanyhigh-incomedevelopingcountrieshavealreadybuiltextensivesystemsforpromotingnationaltechnologicalchange.IinteractlogGDPpercapitaPPPwithPATtotesttheeffectsofpatentimplementationconditionalonacountry’slevelofdevelopment.InteractionvariablesbetweenlogeducationPAT,andbetweenlogeconomicfreedomandPAT,respectively,arealsoincludedtotestthehypothesisthathumancapitalandopenmarketsarecomplementaryfactorstopatentprotectioninstimulatinginnovation(Maskus,2000).TheinteractiontermofthepricecontrolpolicyandPATteststhelinkagesbetweenpatentprotectionandotherindustrypolicy.TotestthevalidityofthetheorythatthereisanoptimallevelofIPRstrength,Iconstructseveralvariables:thesquaredtermofthelogIPRcompositescore(Ginarte&Park,1997),theinteractiontermofthislogcompositescorewithPAT,aswellasthequintiledummiesoftheIPRscoreinrobustnesschecks.III.MethodologyA.MatchedSamplingGroupingcountrieswithsimilarcharacteristicsaccordingtoasinglecountryvariable,asdonebyGinarteandPark(1997),balancesthecountriesonthisparticularvariablebutdoesnothelptoeliminatebiasesduetodisparitiesinothervariables.Thechallengeistondacompositescorethatencompassesallthecountrycharacteristicsthataredeemedtobeimportantbothfortheprobabilityofimplementingdomesticpatentsandforinnovativeactivityinthecountry.Eachnew-patentcountrycanthenbematchedtoano-patent(oralways-patent)countrybyorderingthevaluesofthiscompositescoreamongalltheno-patent(oralways-patent)countriesandndingthecountrywhosescoreistheclosesttothatofthenew-patentcountrytobematched.ThepropensityscoremethodandtheMahalanobismatchingmethodaretwowaystocalculatethiscompositescore.SuchanonparametricmatchingmethodisusedinsteadofHeck-man’sproceduremainlybecausethedecisiontoimplementpatentingistoocomplicatedandidiosyncratictomodel.Animportantdiagnosticcheckfortheeffectivenessofamatch-ingmethodisthecovariatebalance—thedegreeofsimilar-ityincountrycharacteristicsbetweentheno-patent(oralways-patent)andnew-patentcountries—withinmatchedpairs(Rosenbaum&Rubin,1984).BothpropensityscoresandMahalanobisdistancescanbethoughtofasinstrumentsforcovariatebalance.Aslongasthecountrycharacteristicsaresimilaraftermatching,itdoesnotmatterwhichmethodisadoptedtoachievesuchbalance.Itisalsoimportanttorecognizethatthematchingproceduresdonotinvolvetheoutcomevariableatall,sothatthereisnochanceofbiasingresultsinfavorofonepatentconditionovertheotherduringMatchedsamplingisamethodforselectingunitsfromalargepoolofpotentialcontrolstoformareducedcontrolgroupthathassimilardistributionsofobservedcovariatestoatreatedgroup(Rosenbaum&Rubin,1985).Thisstudyattemptedtocalculatepropensityscores—theprobabilityofpatentlawimplementationbasedonthecountries’characteristics—withpartiallymissingdata.Unfortunately,theseriousmissingdataprob-lemintheinitialdatasetrequiresestimatingalargenumberofparameters,whicharenotsupportedbythesamplesize.ThismethodisappliedinrobustnesscheckswhenallthedataareassembledafterusingtheMahal-anobismethod.THEREVIEWOFECONOMICSANDSTATISTICS ThemainadvantageofusingMahalanobismatchingisitsgreaterexibilityandaccuracyinmatchingindividualcountries,whichresultsincountrypairsthataremostsuitableforpairwisestatisticalanalyses.Italsohelpstogetaroundtheproblemofmissingdatabymatchingintwopasses:therstpassmatchedonthevariableswithnomissingobservations;andthesecondpassmatchedonallthecharacteristicsafterllinginthemissingdatawithvariousnationalabstracts(appendixIII).Thismethodmatchesthepointsinamultidimensionalspaceaccordingtothedistancesbetweentwopoints(Rosenbaum&Rubin,1985).Inthisstudy,thecoordinatesofthemul-tidimensionalspacearethematchingvariables,andthepointstobematchedarethesampledcountries.Thedistancebetweenanytwocountriesiscalculatedasafunctionofthedifferencesinthematchingvariables(appendixIII).TheMahalanobismethodcollapsesthesetofcountrycovariatesintoascalardistancescore.Table1liststhematchedcountries.EachmatchisdonebyndingthecountryinacontrolgroupthathastheminimumMahalanobisdistancetothenew-patentcoun-try.Countriesthatlonghadpatentprotectionhavesta-tisticallysignicantlyhigheraveragelevelsofincomes,GDPpercapitaPPP,pharmaceuticaloutputs,andexportsinthepreviousperiodthanthoseinthenew-patentcountries(table2);thevariablevaluesofthenew-patentcountriesareagainhigherthanthoseoftheno-patentcountries(table3).Comparingthe-statisticsoftheOMPARISONOFATENTAND New-PatentCountriesAlways-PatentCountries-statisticBeforeMatchingAfterMatchingBeforeMatchingAfterMatchingBeforeMatchingAfterMatchingCountry-LevelCovariatesGDP(inbillionsof1995constantUSD)148.99145.60422.80162.94(164.29)(158.50)(1,080.06)(264.18)RealGDPgrowth3.443.972.474.631.65*(2.85)(2.81)(3.90)(3.02)GDPpercapitaPPP7,201.057,102.519,396.018,486.38(4,946.28)(5,264.23)(6,518.03)(5,974.89)EconomicFreedom6.076.216.466.86(1.49)(1.31)(1.85)(1.43)LegalOriginofU.K..12.15.45.44(.33)(.36)(.50)(.51)LegalOriginofFrance.39.48.31.41.92.54(.50)(.51)(.46)(.50)LegalOriginofSocialist.27.11.0902.27**1.68*(.45)(.32)(.28)0LegalOriginScandinavian.12.15.05.041.241.41*(.33)(.36)(.21)(.19)PriceControlIndicator0.520.740.540.70.28.30(0.51)(0.45)(0.50)(0.47)Education6.846.817.357.60(2.21)(2.15)(2.71)(2.17)IPRScore2.472.553.243.12(.85)(.80)(.64)(.54)InnovativePotential2.582.703.193.07(1.30)(1.07)(1.71)(1.00)Industrial-LevelCovariatesEmployment29.9828.0736.0911.871.43*1.11(76.30)(73.60)(60.90)(17.53)Output(inmillionsofUSD)1,094.941,046.104,819.991,136.58(1,283.19)(1,213.58)(10,559.58)(2,317.89)PharmaceuticalExportstotheU.S.7.298.7143.588.374.35***.09(12.37)(13.23)(106.97)(14.70)NumberofsubsidiariesofU.S.MNE5.456.526.746.85(8.50)(8.97)(9.75)(8.56)NumberofsubsidiariesofJapaneseMNE1.301.591.462.41(2.88)(3.12)(4.52)(4.64)Indicatorformissingvariable“Employment”.24.401.76**(.44)(.49)Numberofobservations332617626ThistableprovidesdescriptivestatisticsforthedataobtainedfromWorldTradeAnalyzer,WorldBankandUNIDOIndustrialPropertyStatistics.TheeconomicfreedomindexcomesfromtheFraserInstitute,legalfamiliesfromLaPortaetal.(1996),andtheeducationproxyfromBarroandLee(2000).Thepricecontroldummyequals1ifacountryhaspharmaceuticalpricecontrolpolicy,andisdrawnfromDanzon(1997)andEconomistIntelligenceUnitreports.ThenumbersofsubsidiariesofU.S.andJapaneseMNEareprovidedbyProf.FritzFoleyandPaulBeamish.Iconstructedthe“innovativepotential”categoricalvariableusingthedatafromtheUSPTO.Itequals6iftheU.S.patentawards(inallindustriesexceptthepharmaceuticalsinayear)surpass1,000,5ifpatentcountisunder1,000butgreaterthan100,4ifitisunder100butgreaterthan6,3ifitisbetween6and1,and1ifnopatentisawardedatall.Author’scalculationsfromthesampledataofthefourreferenceperiodspriortonationalpatentimplementation,wherecontrolcovariatesareused.Standarddeviationsareinparentheses.Theindustrial-levelemploymentandoutputvariablesareonlyobservedforallcountriesinthereducedmatchedsample.Thestatisticsarecalculatedfortheobservedvalues.The-statisticsareobtainedbyregressingthecovariateonthepatentimplementationindicatorandaconstantwithinthesubclass.Theyreducedsignicantlyaftermatching,indicatingbettercovariatebalancesacrossthenew-patent(treatment)andalways-patent(control)groups.Signicancelevels:*.10,**.05,***DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?441 covariatesbeforeandaftermatching,weclearlyseethatthecovariatesaremuchmorebalancedaftermatching.B.RegressionModels(a)RegressionsontheEntireMatchedSample.thisstudyattemptstomatchonandcontrolforanextensivelistofvariablesthatarecorrelatedwithacountry’sinnovativepotential,biasesmaystillexistduetoincompletecontrols.Thisproblemisaddressedbyusingapaneldataregressionmethod(Rubin&Thomas,2000).Theformalregressionmodelsareestimatedonthesetwogroupsofmatchedpairs(set1:no-patentandnew-patentpairs,andset2:always-patentandnew-patentpairs)separatelyinadditiontopoolingtogetherthetwocontrolgroups:istheoutcomevariableofeachofpairinthereducedsampleinperiod(oryear)OMPARISONOFATENTAND New-PatentCountriesNo-PatentCountries-statisticBeforeMatchingAfterMatchingBeforeMatchingAfterMatchingBeforeMatchingAfterMatchingCountry-LevelCovariatesGDP(inbillionsof1995constantUSD)148.99145.6076.0467.312.38**2.43**(164.29)(158.50)(112.48)(72.82)RealGDPgrowth3.443.972.972.73.811.64*(2.85)(2.81)(3.33)(2.37)GDPpercapitaPPP7,201.057,102.514,248.355,727.533.21***1.12(4,946.28)(5,264.23)(3,203.33)(3,531.82)EconomicFreedom6.076.215.185.672.90***1.54*(1.49)(1.31)(1.41)(1.30)LegalOriginofU.K..12.15.07.12.92.35(.33)(.36)(.26)(.33)LegalOriginofFrance.39.48.54.50(.50)(.51)(.50)(.51)LegalOriginofSocialist.27.11.29.31(.45)(.32)(.46)(.47)LegalOriginScandinavian.12.15.05.081.24*.81(.33)(.36)(.21)(.27)PriceControlIndicator0.520.740.660.76(0.51)(0.45)(0.48)(0.44)Education6.846.815.496.482.92***.58(2.21)(2.15)(1.83)(1.88)IPRScore2.472.552.022.152.46**1.97**(.85)(.80)(.68)(.58)InnovativePotential2.582.702.022.272.35**1.67*(1.30)(1.07)(.88)(.78)Industrial-LevelCovariatesEmployment29.9828.0722.4015.47.47.84(76.30)(73.60)(43.78)(26.16)Output(inmillionsofUSD)1,094.941,046.10649.34668.262.37**1.38*(1,283.19)(1,213.58)(862.73)(722.29)PharmaceuticalExportstotheU.S.7.298.711.931.682.45**2.67**(12.37)(13.23)(5.22)(3.39)NumberofsubsidiariesofU.S.MNE5.456.523.482.771.49*1.93**(8.50)(8.97)(6.89)(4.31)NumberofsubsidiariesofJapaneseMNE1.301.59.36.312.86***1.90**(2.88)(3.12)(1.38)(1.19)Indicatorformissingvariable“Employment”.24.50(.44)(.50)Numberofobservations332615926ThistableprovidesdescriptivestatisticsforthedataobtainedfromWorldTradeAnalyzer,WorldBankandUNIDOIndustrialPropertyStatistics.TheeconomicfreedomindexcomesfromtheFraserInstitute,legalfamiliesfromLaPortaetal.(1996),andtheeducationproxyfromBarroandLee(2000).Thepricecontroldummyequals1ifacountryhaspharmaceuticalpricecontrolpolicy,andisdrawnfromDanzon(1997)andEconomistIntelligenceUnitreports.ThenumbersofsubsidiariesofU.S.andJapaneseMNEareprovidedbyProf.FritzFoleyandPaulBeamish.Iconstructedthe“innovativepotential”categoricalvariableusingthedatafromtheUSPTO.Itequals6iftheU.S.patentawards(inallindustriesexceptthepharmaceuticalsinayear)surpass1,000,5ifpatentcountisunder1,000butgreaterthan100,4ifitisunder100butgreaterthan6,3ifitisbetween6and1,and1ifnopatentisawardedatall.Source:Author’scalculationsfromthesampledataoffourreferenceperiodspriortopatentimplementation,wherecontrolcovariatesareused.Standarddeviationsarelistedinparentheses.Theindustrial-levelemploymentandoutputvariablesareonlyobservedfor39outofthe85sampledcountries.Thestatisticsarecalculatedfortheobservedvalues.Signicancelevelsarereferencedforeachvariableaccordingtotheirdegreeoffreedom:*.10,**.05,***.01.The-statisticsareobtainedbyregressingthepatentimplementationindicatoroneachcovariateandaconstant.TheyreducedsignicantlyaftermatchTHEREVIEWOFECONOMICSANDSTATISTICS yearsafterpatentimplementation).Asimilardenitionappliestoforperiod(oryear)Fortherstspecication,theoutcomevariableisthecitation-weightedU.S.patentawardsafterthenewpatentInthealternativespecicationsforro-bustnesschecks,IuserawU.S.patentcounts,thecitation-weightedU.S.patentswhosecitationnumbersarehalfofastandarddeviationabovethemean(themaininnovations),andpharmaceuticalexportsastheoutcomevariables.consistsofthepair-speciceffects.PATPATMODasdenedinsectionIIA.isthevectorofinteractionvariablesspeciedinsectionIID.COVARIATESreferstoavectorofcontrolvariables.ThefullmodelcontrolsforlogofpharmaceuticalexportstotheU.S.,logGDPpercapitaPPP,logIPRcompositescoreofGinarteandPark(1997)anditssquaredterm,logaverageyearsofschooling,logeconomicfreedom,logJapaneseandU.S.foreignafliatecounts,loginnovativepotential,adummyforpricecontrolpolicy,andlogpharmaceuticalindustrystandsforthevedummiesforeachoftheperiods(forexample,takesonvalue1ifpairismatchedwhenexaminingtheperiod,andtheotherfourperioddummiesforpairhavevalue0).TheregressionresidualisdenotedbyInrobustnesschecks,logGDP,logGDPgrowth,andUKlegaloriginareaddedascontrolcovariates.(b)RegressionsontheSampleofOECDCountries.testingtheimpactsofnationalpatentprotectiononR&Dincentives,Imainlyusedthesampleof23countrieswhoseactualpharmaceuticalR&Dexpendituresareobserved.Asapreliminarystep,IcarriedoutaseriesofdynamicpanelregressionsofOECDcountries’R&Dinlogsonthecontrolcovariates,perioddummies,andthecountryxedeffects,yieldingstatisticallyinsignicantcoefcientsonthepatentchangeindicatorvariable(0.59withstandarddeviation0.34).Ithenadheredtothepreviouslyoutlinedstudydesigntoavoidrelyingonimplicitassumptionsoflinearityim-posedbyOLSregressionmodels.DuetothesmallsamplesizeoftheOECDcountries,Iemployedonlyonecontrolgroup—countriesthatdidnotchangetheirpharmaceuticalpatentlawsduringaparticularperiod—toformabasisofcomparisonforthenew-patentcountries.Ialsodidnotdenepairshere.Table4showsthatthereareafewoutliercountrieswhoseincome,pharmaceuticaloutput,andem-ploymentlevelsaremuchhigherthanthemajorityofothercountries:theUnitedKingdom,France,Germany,Italy,Japan,andtheUnitedStates.Theseoutlierobservationswereremovedfromthesample,signicantlyimprovingthecovariatebalances(table4).Onewouldideallyliketorunseparateregressionsforeachperiod.However,thisisnotpossiblegiventhatonlyafewOECDcountrieschangeddomesticpatentlawsineachperiod.Ithereforestackedtheobservationsofthefourperiodstoformapanel,andthepanelmethodwasagainapplied,buttakingtheperiod-speciceffectsasthexedeffects.istheconstantterm.aresimilartothoseinmodel1.isthepost-patentR&Dexpendituresofcountryofperiodinthesample.(c)RobustnessRegressions.Inordertoobtainrobustresults,regressionsbasedonthreedifferentlikelihoodfunc-tionsareemployed:thenormalregression,theleastabsolutevalue(LAV)regression,andtheHuberregression.Mini-mizingBayes(average)riskand(whichminimizesmaximumrisk)aretwomainapproachesinclassicaldeci-siontheory.TheHuberregressiondoesexactlythelatterbyassumingaleastfavorabledistributionforthedatawithinaclassofdistributions,andobtainstheMLEfromthisHuberTheLAVandHuberregressionsareparticu-larlygoodatavoidinginuencesduetooutliersintheoutcomevariable.Ifthedistributionisexactlynormal,thenleastsquareisthemaximumlikelihood,andthestandardnormalregressionprovidesbetterestimatesthantheotherregressions.AllthreeregressionsarethereforeperformedinTheregressionmodelwiththenumberofU.S.patentsafternationalpatentlegislationasresponsevariableandtheU.S.patentcountsofthepreviousyear(orperiod)ontherightsideisineffecttakingthedifferenceofthetwoU.S.patentawardsvariablesandtestingtheeffectsofpatentimplementationontheincrementofU.S.patentawards.Limitedbythesamplesizeandbythefactthatsomecountrieswithsimilarcharacteristicstendtoimplementnationalpatentlawsinthesameperiod,somecountriesarematchedtomorethanoneothercountryandthereforeappearmorethanonceinthesample.Theobservationsinmodel1arethennotentirelyindependentofeachother;thisbiasesthe-statisticsupward.Analternativemodelisusedtotestrobustness:Model2resemblesmodel2,exceptthatissimplytheconstantterminsteadofthelinearcombinationofpairdummiesinmodel2.Itisemployedsothateachcountryappearsintheregressiononlyonce.However,thelimitationofmodel2isthatitassumesacommonlinearrelationshipbetweentheoutcomevariableandthecontrolvariablesfortheentirecovariatespace,whichmaynotactuallyapplyforthecovariatesacrossdifferentpairs.Thismodelisonlyusefulforrobustnesstests.LAVregression,alsoknownastheminimumL1-normregression,assumesaLaplacedistributionforthedata,andobtainstheMLEthroughminimizingtheabsolutevalueofthedeviationofdatapointsfromtheTheHuberdistributionisatypeofcontaminatednormaldistribution,wherearandomvariable(theoutcomevariable,forinstance)isdrawnfromastandardnormaldistributionwithprobability(1)andfromanalternativedistribution(suchasCauchyorLaplace)withprobability.Intheregressionanalysesofthisstudy,thisparameterisdeterminedthroughseveraliterationstondthevaluethatismostsuitabletodescribethedataDONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?443 ordertoobtainrobustresults.Thesethreetypesofregres-sionsallassumelinearitybetweentheoutcomevariableandeachcontrolvariable.Thisassumptionistestedwithscatterplotsbetweenthelogoutcomevariableandeachofthelogcountrycovariates.IV.ResultsOLSprovidessomepreliminarytestsforwhetherpatentlawsstimulatedomesticinnovation.IregresstheU.S.phar-maceuticalpatentawardsonanindicatorvariablethattakesonvalue1ifacountryhadapharmaceuticalpatentlawinand0otherwise,theU.S.patentawardsforallproductsexceptpharmaceuticals,andcountrydummies.Theresultingcoefcientonthepatentimplementationin-dicatorisstatisticallyinsignicant(0.10withstandardde-viation0.41).TheOLSregressionestimatesrelyheavilyontheassumedlinearrelationsbetweentheresponsevariableandtheindependentvariables.Thematchedsamplingcom-binedwithpairwiseregressionmethodologyeffectivelycontrolsforawiderangeofvariablesindicatinginnovativepotentialandeffectivelyreducessensitivitytothelinearityassumption.Inthissection,IpresenttheregressionresultsonthematchedpairsobtainedfromtheMahalanobismatch-inginsectionIVA.IthendiscusstheresultsobtainedusingR&DoutcomesintheOECDcountriesinsectionIVB.SectionIVCsummarizesthekeyrobustnesschecksforpotentialbiasesduetosmallsamplesandomittedvariables.A.UsingCitation-WeightedU.S.PatentsasInnovationProxyIcarriedoutregressionanalysesonthetwosetsofmatchedpairsfollowingmodels1to2speciedinsectionIIIB.IrsttesttheeffectsofnationalpatentlawonthechangeinthelogofU.S.pharmaceuticalpatentsawardedtodomesticinnovatorsafterthepatentlegislation(model1).Iusethelogofcitation-weightedU.S.patentawardsforaparticularyearaftertheestablishmentofthenationallawsastheresponsevariable.Specically,threeyears(abbrevi-OMPARISONOFTHEOUNTRIESANDOUNTRIESWITHHANGESINTHEOECDC New-PatentCountriesNo-ChangeCountries-statisticsCompleteSampleBalancedSampleCompleteSampleBalancedSampleCompleteSampleBalancedSamplePricecontrol0.240.240.110.171.83**0.90(0.43)(0.43)(0.32)(0.38)U.S.subsidiarycounts7.247.2414.6110.05(10.87)(10.87)(12.10)(9.46)Japanesesubsidiarycounts0.120.121.910.58(0.33)(0.33)(5.53)(0.87)EmploymentinPharms11.5311.5351.8911.61(10.89)(10.89)(53.81)(9.50)Outputinpharms1,228.231,228.238,128.281,501.44(1,427.97)(1,427.97)(13,348.67)(1,459.25)GDPgrowth2.932.932.502.481.84**1.86**(1.46)(1.46)(1.24)(1.32)Economicfreedom6.766.767.367.28(0.90)(0.90)(1.28)(1.23)GDP179.46179.46983.70209.85(145.15)(145.15)(1,561.53)(143.47)GDPpercapitaPPP12,061.5512,061.5513,689.3913,320.63(4,365.54)(4,365.54)(4,991.05)(4,874.86)UKLegalFamily0.070.070.310.30(0.26)(0.26)(0.46)(0.46)FrenchLegalFamily0.480.480.340.341.70**1.64*(0.51)(0.51)(0.47)(0.47)SocialistLegalFamily0000(0)(0)(0)(0)GermanLegalFamily0.020.020.140.04(0.15)(0.15)(0.35)(0.19)ScandinavianLegalFamily0.430.430.210.332.70***1.21(0.50)(0.50)(0.41)(0.47)AverageYearsofSchooling8.038.038.808.85(2.23)(2.23)(1.78)(1.77)IPRindex2.772.773.603.40(0.70)(0.70)(0.62)(0.65)InnovativePotential3.333.333.923.57(0.98)(.98)(0.51)(0.79)No.ofobs.4242304196ThistableprovidesdescriptivestatisticsforthedataobtainedfromWorldTradeAnalyzer,WorldBankandUNIDOIndustrialPropertyStatistics.TheeconomicfreedomindexcomesfromtheFraserInstitute,legalfamiliesfromLaPortaetal.(1996),andtheeducationproxyfromBarroandLee(2000).Thepricecontroldummyequals1ifacountryhaspharmaceuticalpricecontrolpolicy,andisdrawnfromDanzon(1997)andEconomistIntelligenceUnitreports.ThenumbersofsubsidiariesofU.S.andJapaneseMNEareprovidedbyProf.FritzFoleyandPaulBeamish.Iconstructedthe“innovativepotential”categoricalvariableusingthedatafromtheUSPTO.Itequals6iftheU.S.patentawards(inallindustriesexceptthepharmaceuticalsinayear)surpass1,000,5ifpatentcountisunder1,000butgreaterthan100,4ifitisunder100butgreaterthan6,3ifitisbetween6and1,and1ifnopatentisawardedatall.ThebalancedsampledoesnotincludeU.S.,U.K.,Japan,Germany,France,andItaly.Source:Author’scalculationfromthesampledataofOECDcountriesinthefourreferenceperiodspriortonationalpatentimplementation,wherecontrolcovariatesareused.Thestatisticsarecalculatedfortheobservedvalues.The-statisticsisobtainedbyregressingthecovariateonthepatentimplementationindicatorandaconstantwithinthesubclass.Signicancelevels:.10,**.05,***THEREVIEWOFECONOMICSANDSTATISTICS atedasthree-yearforward),four-yearforward,uptoten-yearforward,arethemarkersusedasalternativeyearsfortheoutcomevariablesinaseriesofregressions.TheaverageofU.S.patentawardsvetotenyearspost-reformisalsousedbecausetheU.S.patentawardsmaybesubjecttoyear-to-yeaructuations.Ireportresultsusingtheindivid-ualyearspecicationsaswell,becauseaverageddatacanerodeimportanttrends.Yeardummiesarecontrolledforinthecorrespondingregressionstoaccountforthedifferencesincitationsduetoyeartruncations(Halletal.,2001).AseeminglyunrelatedregressionprocedureisadoptedtoobtainGLSestimatorsinalltheseregressionspecications,andnocoefcientonPATPATMODisfoundtobestatisticallysignicantatthe5%level.Table5liststheregressionresultsforthenew-patentandno-patentmatchedcountrypairs,usinglogten-yearforwardcitation-weightedpatents,thedifferencebetweenlogten-yearforwardandlogbase-yearcitation-weightedpatents,andtheaverageoflogve-toten-yearforwardcitation-weightedpatentsasalter-nativedependentvariablespecications.Table6recordstheestimationsfromthealways-andnew-patentcountrypairs,withthesamesetofalternativespecications.Neitheracountry’spharmaceuticalpatentlawnoritsmodicationofinitialprocessprotectionhasastatisticallysignicantim-pactonthepatentoutcome(rows1and2intables5and6).Althoughpatentimplementationalonedoesnotsigni-cantlyimpactthenumberofpatentsreceivedfromtheUSPTO,itnonethelessmayhaveeffectsconditionalonaOUNTSINATENTAND ResponseVariables(logcitation-weightedUSPindifferentyears)10thyears10thyearsForward10thyear-baseyearForward10thyear-baseyearForward5to10yearsaverageForward5to10yearsaverage.22.76.61.42(.30)(1.80)(.42)(.34)(.37)(.33).06.16(.64)(.58)(.60)(.51)(.63)(.51)logGDPpcPPP.68**1.14***1.39*(.25)(.34)(.91)logFreedom.181.86**1.91(.48)(.81)(1.75)logEducation.38**.10.56**(.19)(.32)(.25)logIPRscore.68.10(.50)(.76)(.45)PriceControl(1.72)(.74)(.81)LogGDPpercapitaPPP.54**.92*.56*.79.33*.63(.22)(.58)(.33)(.83)(.19)(.72)LogEconomicFreedom.64.621.50*.711.37*.17(.76)(1.37)(.91)(.82)(.74)(1.44)LogEducation.55**1.08.281.95**1.10**1.89*(.21)(.80)(.80)(.85)(.41)(1.01)LogIPRScore.02.24.40.59.25.39(.73)(.81)(.58)(.79)(.87)(.90)LogIPRsquared(.10)(.09)(.08)PriceControl(.33)(.52)(.52)(.82)(.34)(.72)LogInnovativePotential.10.35.91*1.43**1.86**.39(.45)(.57)(.52)(.64)(.63)(.58)LogLabor(.13)(.13)(.14)(.22)(.17)(.16)LogPharmaceuticalExportstotheU.S..32*.21.03.03.41**.24(.17)(.25)(.28)(.43)(.18)(.18)LogNumberofU.S.MNEsubsidiaries.08.04.03.22.53**(.16)(.27)(.26)(.44)(.20)(.24)LogNumberofJapanesesubsidiaries.03.10.07.32(.25)(.40)(.43)(.71)(.19)(.36)Logbase-yearcitation-weightedpatents.43**.30**.47*.33(.14)(.14)(.25)(.23)Pair(Year)xedeffectsYYYYYY#ofObs.525252525252-square.94.95.87.87.92.95Regressionresultsfromdifferentregressionspecicationsaretabulatedindifferentcolumns.Thedependentvariableofeachspecicationislistedineachcolumnheader.ThePATImplementationdummyisadummyequalto1intheyearsstartingfromdomesticpatentimplementationidentiedintable1.ThePATMODdummyisadummyequalto1ifthecountryhadpharmaceuticalprocesspatentspriortopatentimplementationaslistedinsectionIIA.MacrodataobtainedfromWorldTradeAnalyzer,andUNIDOIndustrialPropertyStatistics.TheeconomicfreedomindexcomesfromtheFraserInstitute,legalfamiliesfromLaPortaetal.(1996),andtheeducationproxyfromBarroandLee(2000).Thepricecontroldummyequalsto1ifacountryhaspharmaceuticalpricecontrolpolicy,andisdrawnfromDanzon(1997)andEconomistIntelligenceUnitreports.ThenumbersofsubsidiariesofU.S.andJapaneseMNEareprovidedbyProf.FritzFoleyandPaulBeamish.Iconstructedthe“innovativepotential”categoricalvariableusingthedatafromtheUSPTO.Itequals6iftheU.S.patentawards(inallindustriesexceptthepharmaceuticalsinayear)surpass1,000,5ifpatentcountisunder1,000butgreaterthan100,4ifitisunder100butgreaterthan6,3ifitisbetween6and1,and1ifnopatentawards.PATreferstotheinteractionvariableofPATdummyandthecovariate,whereislogofGDPpercapitaPPP,economicfreedomindex,education,orIPRindex,orthepricecontroldummy.Heteroskedasticity-consistentstandarderrorsthatcorrectforclusteringatthepairlevelappearinparentheses.DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?445 country’sinnovativepotential.Toexploresucheffects,variablesinteractingthepatentimplementationindicatorwithcountrycharacteristicsareincludedintheregressionsinthesecondroundofanalyses,followingmodel2insectionIIIBa.Columns2,4,and6intables5and6showthatpatentprotectiondemonstratessomeimportancecon-ditionalonacountry’sdevelopmentlevels,economicfree-dom,andeducation.Inowdiscusstheresultsonthesethreeinteractioneffectsonebyone.Intheregressionsusingthecitation-weightedU.S.pat-entsasoutcomevariables,theGDPpercapitaPPPandPATinteractionvariablebearspositivesigns,andthecoefcientsarestatisticallysignicantatthe5%or10%levels(row3inTables5and6).Thissuggeststhatpatentsareimportantforinnovationconditionalonacountry’sdevelopmentlevel.Amoredevelopedcountrywithpharmaceuticalpatentsislikelytohavemoreinnovationscomparedtoasimilarlydevelopedcountrywithoutpatents,oralessdevelopedcountrywithpatents.Thestatisticallysignicantcoefcientonthisinteractiontermindicatesthatthepatent-reformeffectonthelogcitation-weightedU.S.patentsawardedtenyearsafterthereformdoubles,onaverage,withaunitincreaseinlogGDPpercapitaPPP(coefcient0.68incolumn2oftable5:e1.97).ThecoefcientsonthelogGDPpercapitaPPPtermitselfarepositiveinallregressionsandsignicantinsomespecications(tables5and6).Normconclusionscanbeestablishedaboutthecondi-tionalimportanceofpatentprotectiongivenacountry’sOUNTSINATENTAND ResponseVariables(logcitation-weightedUSPindifferentyears)10thyears10thyearsForward10thyear-baseyearForward10thyear-baseyearForward5to10yearsaverageForward5to10yearsaverageImplementation.12.16.04.42(.23)(.37)(.50)(.53)(.25)(.41)(.33)(.31)(.64)(.61)(.27)(.31)logGDPpcPPP.40*.87**.49**(.22)(.32)(.24)logFreedom1.34*2.15*.34(.87)(1.13)(.90)logEducation.16.39**.06(.14)(.19)(.11)logIPRscore(.76)(.94)(.43)PriceControl(.47)(.87)(.51)LogGDPpercapitaPPP.54**.35*.57*.46.24.40(.25)(.21)(.29)(.62)(.30)(.34)LogEconomicFreedom.43.351.892.63**.10.30(.76)(1.05)(1.42)(1.20)(.56)(.73)LogEducation1.66**1.98*.221.691.20*.64**(.72)(.99)(1.01)(1.11)(.68)(.28)LogIPRScore.03.70.121.83.30.95(.25)(.64)(.60)(1.51)(.27)(.62)LogIPRsquared(.02)(.09)(.04)PriceControl(.28)(.35)(.55)(.96)(.27)(.43)LogInnovativePotential.40.781.62**2.50**1.21**1.33**(.38)(.69)(.82)(1.03)(.48)(.66)LogLabor.13.15.42.18.10.12(.14)(.20)(.46)(.43)(.23)(.22)LogPharmaceuticalExportstotheU.S..11.12.12.03.18.20*(.13)(.11)(.47)(.28)(.11)(.12)LogNumberofU.S.MNEsubsidiaries.38.08(.10)(.18)(.38)(.34)(.21)(.24)LogNumberofJapanesesubsidiaries.47**.44.003.27.18(.16)(.20)(.31)(.50)(.17)(.20)Logbase-yearcitation-weightedpatents.10.11.77***.81***(.08)(.15)(.09)(.13)Pair(Year)xedeffectsYYYYYY#ofObs.525252525252-square.96.97.76.88.97.97Regressionresultsfromdifferentregressionspecicationsaretabulatedindifferentcolumns.Thedependentvariableofeachspecicationislistedineachcolumnheader.ThePATImplementationdummyisadummyequalto1intheyearsstartingfromdomesticpatentimplementationidentiedintable1.ThePATMODdummyisadummyequalto1ifthecountryhadpharmaceuticalprocesspatentspriortopatentimplementationaslistedinSectionIIA.MacrodataobtainedfromWorldTradeAnalyzer,andUNIDOIndustrialPropertyStatistics.TheeconomicfreedomindexcomesfromtheFraserInstitute,legalfamiliesfromLaPortaetal.(1996),andtheeducationproxyfromBarroandLee(2000).Thepricecontroldummyequalsto1ifacountryhaspharmaceuticalpricecontrolpolicy,andisdrawnfromDanzon(1997)andEconomistIntelligenceUnitreports.ThenumbersofsubsidiariesofU.S.andJapaneseMNEareprovidedbyProf.FritzFoleyandPaulBeamish.Iconstructedthe“innovativepotential”categoricalvariableusingthedatafromtheUSPTO.Itequals6iftheU.S.patentawards(inallindustriesexceptthepharmaceuticalsinayear)surpass1,000,5ifpatentcountisunder1,000butgreaterthan100,4ifitisunder100butgreaterthan6,3ifitisbetween6and1,and1ifnopatentawards.referstotheinteractionvariableofPATdummyandthecovariate,whereislogofGDPpercapitaPPP,economicfreedomindex,education,orIPRindex,orthepricecontroldummy.Heteroskedasticity-consistentstandarderrorsthatcorrectforclusteringatthepairlevelappearinparentheses.THEREVIEWOFECONOMICSANDSTATISTICS educationattainmentoreconomicfreedomforthecitation-weightedpatentoutcomes,duetothevaryingsignicancelevelsincoefcientsonthetwointeractionvariablesacrossvariousregressionspecications.However,allthesecoef-cientsarepositive,andmanyarestatisticallysignicantatthe5%or10%levels(rows4and5intables5and6).Inaddition,thepositivecoefcientsonthemaineffectsoflogeducationandfreedomarestatisticallysignicantinmanyspecicationsaswell.Allthesedemonstratetheimportanceofeducationattainmentandeconomicfreedomforana-tion’sinnovationlevel.ThepositivecoefcientsonthelogIPRscoreandthenegativecoefcientsonitssquaredterm(tables5and6)shedlightsonthetheorypioneeredbyGallini(1992)thattherelationshipbetweenpatentstrengthandinnovationadoptsan“invertedU”shape.TheinteractionoflogIPRandthepatentreformdummyalsotakesonnegativecoef-cientsinmostspecicationsandsignicantatthe5%levelinpredictingthecitation-weightedpatentoutcometenyearsafterpatentimplementationsinthenew-patentandalways-patentgroups(table6).Theseresultsrelatetothetheoriesthatcompetitionsometimescaninduceinnovation(Aghionetal.,2002;Qian,2005).Thevariableinteractingthepricecontroldummywiththepatentreformdummytakesonnegativecoefcientsthroughoutallspecications,althoughnotstatisticallysignicant.Togetherwiththenegativeco-efcientsonthepricecontroldummy,theresultsindicateanegativerelationshipbetweenpricecontrolpoliciesandthepatentoutcomes.Thisndingisinagreementwithothers’(Grabowski&Vernon,1992;Danzon,1997)thatpricecontrolpolicytendstoimpairdomesticinnovation.Inmostoftheregressionequationsintables5and6,thecoefcientsonthe“innovativepotential”variablearepos-itiveandstatisticallysignicantatthe5%or10%level.Thisillustratestheimportanceofacountry’sinnovativepotentialinexplainingtheinnovationdifferencesbetweencountries.Whilethetrueinnovativepotentialofacountryisobviouslynotdirectlyobserved,apparently,itcanbepartlycapturedinthevariableconstructed.Thecountry’sbase-yearpharmaceuticalexportstotheUnitedStateshavepositivecoefcients,statisticallysignif-icantinseveralregressions.TheincentivetopatentintheUnitedStatesislargelydeterminedbythemarketpotentialinnovatorsseeandseekinthatcountry.Inaddition,becausetradebetweenthetwocountriesisafunctionofgeographicdistance,linguisticdifferences,andothervariablesthatcouldaffectforeigninnovators’propensityforpatentingintheUnitedStates,controllingfortradevalueshelpstocontrolfortheseindirectvariables.Infact,innovatorsinagivencountryaremorelikelytoseekU.S.patentsif,historically,theircountryhasexportedmorepharmaceuti-calstotheU.S.market.Inallthexed-effectsregressions,the-statisticsforthegroupofpairdummies,,arestatisticallysignicantatthe5%level,andmanyofthemaresignicantevenatthe1%level.Thisimportanceofpair-speciceffectsindicatesthemethodologicalsignicanceofthematchingtechnique.Thecoefcientsontheothercontrolvariablesaremostlyinsignicantatthe5%level.B.UsingR&DExpendituresandPersonnelasInnovationProxies(a)ResultsontheMainPatentImplementationDummy.U.S.patentawardscanbeconsideredanestimateofinno-vationoutputs,whileR&Dexpendituresprovideanesti-mateofinnovationinputs.ItislikelythatthestimulusfrompatentprotectioncouldimpactR&DmuchsoonerthanitwouldU.S.patentgrants.Onaverage,thepatent-grantingprocesstakesonetotwoyears(USPTO,2000,interviewofPaulHarrison),notcountingthetimeneededfordrugdevelopment.TheR&Dresponsetodomesticpatentlawsisfoundtobeimmediate(Lo,2004).R&Dexpendituresforoneyearandoneperiodafternationalpatentreforms(theperiodsaredenedinsectionIIAandtable1),andthedifferencebetweenR&DtwoyearsafterpatentreformandR&Dinthebase-yeararespeciedasalternativedependentvariablesintheregressionmodel3intable7.Forrobustnesschecks,one-year,two-year,andone-periodforwardR&Dexpendituresandone-periodforwardR&Dpersonnel(RSE)arespeciedasalternativeoutcomes.Toaddresstheconcernthatpatentimplementationmaybeendogenous,R&Doutcomesofoneortwoyearspriortopatentlegisla-tionarealsoadopted.Allthesespecicationswithdifferentdependentvariablesbutsamesetofindependentvariablesarecarriedoutasasystemofseeminglyunrelatedregres-sions.Inaddition,regressionswithR&Doneperiodfor-wardarecarriedoutseparatelyonthepairsofthenon-OECDcountriesintheappendix.Estimatesofpatents’effectsremainnullforallthesespecications.Row1intable7showsthestatisticallyinsignicantcoefcientsonthepatentdummyinthemainspecications.ThePATMODvariableisdisregardedbecausealltheOECDcountries,exceptTurkey,hadprocesspatentspriortotheirnationalproductpatentlegislation.Mostregressionsyieldnostatisticallysignicantcoef-cientonPAT.TheonlystatisticallysignicantpositivecoefcientonthepatentimplementationindicatorappearsintheHuberregressionofR&Dtwo-yearforwardfortheOECDcountries.ThisndingaloneisinsufcienttorejectthenullhypothesisthatpatentimplementationshavenotgenerallystimulatedR&Dincentives,givenalltheotherR&Dexpenditureisavailablefrom1978to1997formostsampledOECDcountries,with1998dataavailableonlyforvecountries.There-fore,R&Dobservationsofthree-yearforwardaretoosparsetoobtainaccurateestimatesonthemainpatentimplementationvariable.RSEdataistoofewtocarryoutindividualyearspecications.Aresidualplotrevealsanoutlierinthedata:TurkeyexperiencedadropinR&Dexpenditurefrom$10.09millionin1992to$0.74millionin1994.WhilethestandardnormalandLAVregressionswerelikelyskewedbythis“inuential”outlier,HuberregressionsuccessfullytstheoptimallinefortheremainingdatapointsthattendtohavesimilarR&Dtrends.DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?447 insignicantresultsobtainedinaseriesofrobustnesstestsinappendixII.(b)ResultsfromtheInteractionTermsandControlCo-ThevariablesinteractingpatentreformdummywithlogofGDPpercapitaPPP,logeconomicfreedom,andlogeducation,respectively,allcarrypositivecoefcientsandarestatisticallysignicantatthe1%,5%,or10%levels(rows2–4intable7).Themaineffectsofthedevelopmentlevel,education,andfreedomproxiesarealsopositive.Alltheseimplythatpatentlawsmayhavepositiveeffectsconditionalonhighlevelsofdevelopment,education,andeconomicfreedom.Here,itisworthnotingtheconsistentlysignicantpositiveinteractioneffectofeconomicfreedomandpatentreforms.ThisresulthintsatthepossibilitythatinamoreintegratedmarketsuchasthatformedbytheOECDcountries,nationalpatentlawcouldcomplementamembercountry’sopenmarketaccessandfavorabledomesticin-vestmentpoliciestoattractFDIandotherformsofforeigntechnologytransfers,whichisempiricallysupportedinBranstetter,Fisman,andFoley(2006).Patentlawscouldalsohelpdomesticcompaniesassimilatetheseinwardtech-nologytransfersthroughpatentdisclosures.Economicfree-domcanhelpcountrieswithnewpatentsystemstoleveragetheiremergingnationalintellectualpropertyadvantagesbyfacilitatingexportsaswell.Itisinterestingtoseetheconsistentlysignicantpositiveeffectofeducationlevelsinteractedwithpatentreforms,suggestingthatdomesticpatentlawscouldprovideincentivesforthehigh-levelhumancapitaltoengageininnovativeactivities.AstheR&DEXPENDITURESINTHEOECDC ResponseVariables(logR&D)Forward2-baseForward2-baseImplementation.26.51.01.04.43(.18)(1.09)(.09)(2.26)(.04)(.96)logGDPpcPPP1.52***.76**.59*(.52)(.25)(.35)logFreedom.87***.12**.15*(.26)(.06)(.09)logEducation.32**.29***.55**(.13)(.08)(.20)logIPRscore(.11)(.17)(.12)LogGDPpercapitaPPP.74*.65.08.06.82**.51(.43)(.81)(.32)(.35)(.41)(.76)LogEconomicFreedom.09.24.18.14.37***.17(.24)(.32)(.26)(.29)(.08)(.25)LogEducation4.97***.061.46**2.16*.48.19(1.24)(.48)(.63)(1.36)(.35)(.42)LogIPRScore.79**.001.18.16.88.09(.38)(.43)(.27)(.24)(1.77)(.30)LogIPRsquared(.05)(.03)(.06)LogInnovativePotential.04.18.11.10**.09.19(.48)(.47)(.22)(.01)(.13)(.25)LogPharmaceuticalExportstotheU.S..12**.04.05**.04.20***.35(.05)(.06)(.02)(.03)(.04)(.46)LogNumberofU.S.MNEsubsidiaries.06**.04.02.01.21***.60(.03)(.07)(.02)(.03)(.05)(.60)LogNumberofJapaneseMNEsubsidiaries.02.14.10.11.53***.13(.11)(.17)(.07)(.09)(.10)(.14)PriceControlIndicator(.09)(.12)(.05)(.11)(.07)(.11)LogLabor.01.08.16.02.23.07(.09)(.14)(.09)(.07)(.18)(.13)Logbase-yearR&D.86***.97***.98***.99***(.11)(.13)(.05)(.06)Country&periodFEYYYYYY#ofobs.238238238238238238-square.98.98.99.99.48.42Regressionresultsfromdifferentregressionspecicationsaretabulatedindifferentcolumns.Thedependentvariableofeachspecicationislistedineachcolumnheader.ThePATImplementationdummyisadummyequalto1intheyearsstartingfromdomesticpatentimplementationidentiedintable1.MacrodataobtainedfromWorldTradeAnalyzer,andUNIDOIndustrialPropertyStatistics.TheeconomicfreedomindexcomesfromtheFraserInstitute,legalfamiliesfromLaPortaetal.(1996),andtheeducationproxyfromBarroandLee(2000).Thepricecontroldummyequals1ifacountryhaspharmaceuticalpricecontrolpolicy,andisdrawnfromDanzon(1997)andEconomistIntelligenceUnitReports.ThenumbersofsubsidiariesofU.S.andJapaneseMNEareprovidedbyProf.FritzFoleyandPaulBeamish.Iconstructedthe“innovativepotential”categoricalvariableusingthedatafromtheUSPTO.Itequals6iftheU.S.patentawards(inallindustriesexceptthepharmaceuticalsinayear)surpass1,000,5ifpatentcountisunder1,000butgreaterthan100,4ifitisunder100butgreaterthan6,3ifitisbetween6and1,and1ifnopatentawards.referstotheinteractionvariableofPATdummyandthecovariate,wherestandsforlogofGDPpercapitaPPP,economicfreedomindex,education,orIPRindex,orthepricecontroldummy.Heteroskedasticity-consistentstandarderrorsthatcorrectforclusteringatthecountrylevelappearinparentheses.TheinteractionvariableofpatentimplementationdummyandpricecontroldummyisomittedintheregressionduetolackofvarTHEREVIEWOFECONOMICSANDSTATISTICS OECDcountriesofferadvancededucation,theirhumancapitalcanbeproductiveininnovationswheninducedtoswitchoutofimitationsbyimplementingpatents.ThestatisticallysignicantnegativecoefcientsontheinteractionbetweenIPRandPAT(table7,row5)andonthesquaredlogIPRterm(table7,row10)supportndingsintheprevioussectionthattherelationshipbetweenpatentstrengthandinnovationadoptsan“invertedU”shape.MostOECDcountrieshadpharmaceuticalprocesspatentsbeforetheyintroducedproductpatents;itislikelythatacountry’sprocessinnovationswereeffectivelyprotectedifitsinitialnationalIPRprotectionwasstrong.Additionalproductpatentsthenmaynotstimulateinnovation,asScherer(1977)andKumar(1996)suggest.Infact,thestrengtheningofpatentprotectionmayblockdomesticinitiativestoen-gagein“imitative”innovations,while“ingenious”innova-tionmaynotcomeeasilyandquickly.ThisleadstoanoveralldeclineindomesticR&Dactivitiesintheshortrun.AnalternativeexplanationisthatcountrieswithahigherIPRindexaremorelikelytoenforcelawsthatprotectintellectualproperty,suchastradesecretlaws.Domesticinnovatorsmayhavealternativewaystoappropriateprotsfromtheirinnovations.Thisndingiscorroboratedinthetimetrends(gure1)oflogcitation-weightedpatentcountsveyearspre-andpost-patentlawimplementationforthecompletetreatmentsampleofcountries(dashedline)andthesubsetofcountriesthathavestrongIPenforcement,asindicatedbytheIPRindex(solidline).Noticethatpost-implementation,thelogcitation-weightedpatentcountsdroppedmoreforthehighlyenforcedsubsamplethanthecompletesample.Thisplotmainlyusesthewithin-countrypre-andpost-implementationcomparison,andissuscepti-bletoendogeneityproblemsofnationalpatentimplemen-tation.Itonlyservestoprovideavisualcheckofthepatentingtrends.Table7showsthatthecoefcientsontheothercontrolcovariatescarrysignsandstatisticalsignicancessimilartothoseintheprevioussectionwithcitation-weightedpatentsoutcomes.ThelognumbersofU.S.andJapaneseMNEsubsidiariesgainafewsignicantpositivecoefcientsatthe1%or5%levels,providingsomeevidenceforpositiveimpactsofFDIoninnovations.ThevariableinteractingpricecontrolandpatentimplementationisomittedfromtheregressionsforOECDcountriesduetoalackofvariationinthesample.Thepricecontroldummycarriesnegativeco-efcientsthroughoutthevariousspecications.C.CheckingPotentialBiasesduetoSmallSamplesandOmittedVariablesBootstrapsimulationsareappliedtoestimatethesizeofthebiasinthecoefcients’estimatesusingsmallsamples.Inparticular,randomlydraw78countriesamongthetwosetsoftreatment-controlcountrypairsandexecutetheregressionmodelsinsectionIIIBaonthesimulatedboot-strapsamplesasdonewiththeoriginalsample.Irepeatthissimulation10,000times,andobtainadistributionforeachoftheregressioncoefcients.IthencalculatethemeanofRENDSOFOUNTSINTHEOUNTRIESTHATAWSINTHE Time Trend of log Citation Weighted Patents for the New-patent Countries-0.3-0.2-0.1-6-4-20246Year relative to National Patent Implementationlog citation weighted patents relative to that in the year of national patent implementation complete highenforced Thisgureplotsthetrendsfortwogroupsofcountries—thecompletesampleofcountriesinthetreatmentgroup,andthesubsampleoftreatmentcountriesthathashighIPenforcementlevels,asindicatedbyanoverallIPRscore(Ginarte&Park,1997)nolessthan0.66.TheIPRscorestakevaluesof0,0.33,0.66,and1amongallcountries.DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?449 eachcoefcient’sdistributionandcompareitwiththeco-efcientestimatefromtheoriginalstudy.Thedifferencesbetweenthemeancoefcientestimatefrombootstrapsam-plesandthecoefcientestimatefromtheoriginal26pairsistheestimateofthebiasofthiscoefcientestimate.Thebiasesofestimatesforthemainpatentandtheinteractionvariablesareinthe0.01–0.03rangesandwouldnotalter-valuesofthecoefcientsestimatessignicantly,indicat-inganegligiblebiasintheoriginalestimates.Tocheckhowsensitivethepatenteffectestimatesaretootherpotentiallatentvariables,IconductstatisticalanalysesfollowingRosenbaum(1999,2002),andtheestimatesinthisstudyareagainrobusttothesealternativetests.Asensitivityanalysisaddressesthefollowingconcern:aftermatchingontheobservedcountrycharacteristics,howlargemusttheresidualdeparturefromrandomassignmentbebeforethequalitativeconclusionsarealtered.IcalculatetherangeofpointestimatesforthepatentlaweffectsinmystudyandcheckthemagainstthestandardsensitivityrangelistedinRosenbaum(2002).Theestimatesinthisstudyareshowntobeinsensitivetohiddenbiasescausedbyunob-servedcovariates.Detailsareavailableuponrequest.V.ConclusionsThispaperteststheeffectsofnationalpharmaceuticalpatentlegislationondomesticinnovations.Itaddressestwoobstaclesintheeconomicliteratureoftechnologicalchanges:datadeciencyandmethodologicallimitations.Thisstudyconstructsadatabasethatapproachesidealexperimentaldata,giventhelimitationsofobservationalstudies.Intheliteratureoftechnologicaladvances,thisstudyisthersttoadoptthematchedsamplingmethodscombinedwithxed-effectspanelregressions.Thelackofobservedcounterfactualoutcomes—whatwouldhavehappenedinthepresenceorabsenceofnationalpatentlaw—foragivencountrynecessitatesinternationalcom-parisons.Inpreviousstudiesinvolvingdevelopedcoun-tries,thenaturalbenchmarksforcomparisonsareotherdevelopedcountries,notablyOECDcountries.Thechoiceofcontrolgroupsbecomesmuchmoredifcultwhenoneintendstostudycountriesatvariousincomeanddevelopmentallevels,asisthecasehere.OnekeyinnovationofthisstudyistoapplytheMahalanobismatchingmethodtoovercomethedifcultiesofmissingdata,andtomatchcountriesofsimilarcharacteristics.Fixedpair-effectsregressionmodelsonthesampleofmatchedcountriescontrolwellforthevariouscountrycharacteristics—bothobservedandunobserved—whicharecorrelatedwithlatentinnovativepotentialandareimportantforexplaininginnovationoutcomes.Thishelpstoimprovetheprecisionofestimates.Aftercontrollingforalistofcovariatesthatarelikelytoaffectinnovativepotentials,IndnostatisticallysignicantrelationshipbetweennationalpharmaceuticalpatentprotectionandU.S.patentawards(innetandcitation-weightedcounts)ordomesticR&D.Thisempir-icalndingishardlysurprising.Somedevelopingcoun-trieshavealwayshadpatentprotection,yetdomesticallytheydonothaveinnovativepotentialandrelyheavilyonHowever,theinteractionofnationalpatentimplementationwithdevelopmentlevelispositivelyre-latedwithdomesticR&DexpenditureanddomesticallyoriginatedpharmaceuticalpatentsintheUnitedStates.TheinteractionofpatentimplementationandeconomicfreedomandtheinteractionofpatentimplementationandeducationareshowntobepositivelyrelatedtoR&DexpenditureintheOECDcountries.Furthermore,thereappearstobeanoptimallevelofIPRregulationabovewhichenhancedprotectioniseventuallyassociatedwithadeclineininnovativeactivities.Inshort,forcountriesthathaverelativelylowlevelsofdevelopment,educa-tion,andmarketfreedom,anypotentialbenetsfromadditionalinnovationdependultimatelyondomesticmacroeconomicfactorsandrequireasubstantialtime-Itisalsopossiblethatthelackofastatisticallysignif-icantincreaseintheU.S.patentawardsafternationalpatentlegislationislinkedtodatalimitations,asdis-cussedindetailinappendixI.B.Patentlawsmayalsoaffectinnovationindimensionsotherthanraisingitsabsolutenumber,inparticular,changingthedirectionofinnovativeactivity(Moser,2003)orincreasingFDI(Branstetter,Fisman,&Foley,2006).Nevertheless,thendingsinthisstudyhaveimportantpolicyimplications.TheyvindicateMaskus’s(2000,p.199)argumentthat“expectationsthatstrongerIPRsalonewillbringtechni-calchangeandgrowtharelikelytobefrustrated.”Coun-trieswithdifferentdegreesofdevelopment,generalin-tellectualpropertystrength,andeconomicfreedomhavevaryinginnovativeresponsestonationalpatentlaw.Mostofthesecountrycharacteristicsindeedgohandinhand.Kumar(1996)ndsapositiverelationshipbetweentheR&DintensityofU.S.afliatesandthestrengthofnationalIPRonlyindevelopedcountries,notdevelopingones.Manydevelopedcountries,includingGermanyandSwitzerland,hadopposednationalpatentlegislationwhentheyweretechnologyimportersinordertotakeadvantageoffreelyaccessibleforeigntechnologies.Evenson(1990)arguesthatcountrieshavenointerestinstrongIPRuntiltheybecomesignicanttechnologyexporters.AlthoughtheTRIPsallowforadjustmentitisunlikelythatdevelopingcountrieswilltrans-formfrommere“technologyimporters”toevenmoderate“technologyexporters”withinthisshorttimeframe.Forinstance,datashowthatFrenchWestAfricaneverappliedforanypharmaceuticalpatentfromtheEPOorUSPTOduring1978–1999,despiteitswell-establishednationalpatentlaws.Developingcountrieshadagraceperiodofveyearsandtheleastdevelopedhadtenyears.THEREVIEWOFECONOMICSANDSTATISTICS A.DataSourcesPharmaceuticalPatentIndexConstructionBaxter,WorldPatentLawandPractice(London:Sweet&Maxwell,Hanellin,Elizabeth,PatentsThroughouttheWorld,4thed.,ed.AlanJ.Jacobs,updatedbyE.Hanellin(Deereld,IL:ClarkBoardmanCallaghan,1978).Scherer,andWatal,“Post-TRIPSOptionsforAccesstoPatentedMedi-cinesinDevelopingCountries,”manuscript(2001).WIPO,“Existence,ScopeandFormofGenerallyInternationallyAcceptedandAppliedStandards/NormsfortheProtectionofIntellectualProperty,”DocumentpreparedbytheInternationalBureauofWIPOfortheMultilateralTradeNegotiations,TheUruguayRound,MTN.GNG/NG11/W/24/Rev.1.SpecialDistribution.Pro-videdbyMs.JayashreeWatal(1988). “ConsultativeMeetingofDevelopingCountriesontheHarmoni-zationofPatentLaws,”HC/CM/INF/1Rev.(Geneva:WIPO,Yusuf,Abdulqawi,“IntellectualPropertyProtectionintheCountriesofInternationalJournalofTechnologyManagement,(1995),269–292.OtherVariablesBarro,Robert,andJong-WhaLee,“InternationalDataonEducationalAttainmentUpdatesandImplications,”NBERworkingpapersW7911(2000).Beamish,Paul,AndrewDelios,andDonaldLecraw,JapaneseMultina-tionalsintheGlobalEconomy(Cheltenham,U.K.;Northampton,MA:EdwardElgar,1997).BureauofEconomicAnalysis,“U.S.MultinationalEnterprisesForeignAfliateCounts1980–99,”FileprovidedbyDr.FritzFoley.Burger,Delien,SouthAfricaYearbook,GovernmentCommunicationandInformationSystem(1999).CentralStatisticsOfce,StatisticalAbstractIreland,GovernmentofIreland(1999).CouncilofLaborAffairs,YearbookofLaborStatistics,CouncilofLaborAffairs,ExecutiveYuan,December1990(1989,1996).EconomicDailyNews,EconomicYearbookoftheRepublicofChina(1985,1990,1995).Euromonitor,InternationalMarketingDataandStatistics(London:Eu-romonitorPublications,2001).EuropeanCommission,EurostatYearbook:AStatisticalEyeonEurope,Data1988–1998,EuropeanCommunities(2000).IMF,InternationalFinancialStatistics,CD-ROM(2000).Menem,Carlos,R.Fernandez,R.Frigerio,andH.Montero,YearbookoftheArgentineRepublic,MinistryofEconomy,PublicWorksandServices,NationalInstituteofStatisticsandCensusesAnalyticalResearchers,ScientistsandEngineersdatabaseProvidedbyMr.DominiqueGuellecatOECD,France BasicScienceandTechnologyStatistics OECDHealthData2000—AComparativeAnalysisof29Coun-CD-ROM(2000c). OECDSTANDatabaseforIndustrialAnalysis1978–1997 “PharmaceuticalPoliciesinOECDCountries:ReconcilingSocialandIndustrialGoals,”LabourMarketsandSocialPolicy—Occa-sionalPapers,http://www.ocde.org/els/social/docs.htm(2000e). ResearchandDevelopmentExpenditureinIndustry1977–1998(2000f).http://www.oecd.org/dsti/sti/stat-ana/index.htm.Pricecontrolresearch,http://www.warner-lambert.com/pricecontrolstudy/Reddy,Marlita,StatisticalAbstractoftheWorld,GaleResearch(1996).SingaporeDepartmentofStatistics,YearbookofStatistics,Singapore,DepartmentofStatistics,MinistryofTrade&Industry,RepublicofSingapore(1997).StatisticsCanada,WorldTradeAnalyzer,CD-ROM(2000).StatisticalOfceoftheEuropeanCommunities,EurostatDatabases(NewCronos)CD-ROM.Summers,Robert,andAlanHeston,“ThePennWorldTable(Mark5):AnExpandedSetofInternationalComparisons,1950–1988,”Quar-terlyJournalofEconomics106:2(May1991),327–368.UnitedNationsIndustrialDevelopmentOrganization,IndustrialStatisticsDatabase4-digitISIC,CD-ROM(2000).UnitedNations,IndustrialStatisticsYearbook(1980–1997annual). StatisticalYearbookforAsiaandthePacic,StatisticsDivision,EconomicandSocialCommissionforAsiaandthePacic.Bankok,Thailand(1999).U.S.PatentandTrademarkOfce,“PatentingTrendintheU.S.1999,”CD-ROM,USPTO,InformationProductsDivision,TechnologyAssessmentandForecastBranch(2001).Wilkie,J.,E.Aleman,andJ.Ortega,StatisticalAbstractofLatinAmerica,36,UCLALatinAmericanCenterPublications,UniversityofCalifornia,LosAngeles(2000).WorldBank,WorldDevelopmentIndicators,InternationalBankforRe-constructionandDevelopment(2000).B.LiteratureAcemoglu,Daron,andJoshuaLinn,“MarketSizeinInnovation:TheoryandEvidencefromthePharmaceuticalIndustry,”NBERworkingpaperw10038(2003).Acemoglu,Daron,SimonJohnson,andJamesRobinson,“InstitutionsastheFundamentalCauseofLong-RunGrowth,”NBERworkingpaperw10481(2004).Aghion,Philippe,N.Bloom,R.Blundell,R.Grifth,andP.Howitt,“CompetitionandInnovation:AnInvertedURelationship,”Har-vardworkingpaper(2002).Bessen,James,andEricMaskin,“SequentialInnovation,Patents,andImitation,”DepartmentofEconomics,MassachusettsInstituteofTechnologyworkingpaperno.00-01(2000).Branstetter,Lee,RaymondFisman,andFritzFoley,“DoStrongerIntel-lectualPropertyRightsIncreaseInternationalTechnologyTrans-QuarterlyJournalofEconomics121:1(2006),321–349.Challu,Pablo,“EffectsofMonopolisticPatentingofMedicineinItalySince1978,”InternationalJournalofTechnologyManagement(1995),237–251.Chin,Judith,andGeneGrossman,“IntellectualPropertyRightsandNorth-SouthTrade”(pp.90–107),inRonaldW.JonesandAnneO.Krueger(Eds.),ThePoliticalEconomyofInternationalTrade:EssaysinHonorofRobertE.Baldwin(Cambridge,MA:BasilBlackwell,1990).Cohen,Wesley,RichardR.Nelson,andJohnP.Walsh,“ProtectingTheirIntellectualAssets,”NBERworkingpaper7552(2000).Danzon,Patricia,PharmaceuticalPriceRegulation(Washington,DC:AmericanEnterpriseInstitute,1997).Deardorff,Alan,“WelfareEffectsofGlobalPatentProtection,”59(1992),36–51.Evenson,RobertE.,“SurveyofEmpiricalStudies,”inWolfgangE.Siebeck(Ed.),StrengtheningProtectionofIntellectualPropertyinDevelopingCountries(Washington,DC:WorldBankdiscussionpaperno.112,1990).Gallini,Nancy,“PatentPolicyandCostlyImitation,”RandJournalof23(1992),52–63.Ginarte,Juan,andWalterPark,“DeterminantsofPatentRights:ACross-NationalStudy,”ResearchPolicy26(1997),283–301.Grabowski,HenryG.,andJohnM.Vernon,“BrandLoyalty,Entry,andPriceCompetitioninPharmaceuticalsafterthe1984DrugAct,”JournalofLaw&Economics35:2(1992),331–350.Griliches,Zvi,R&D,Patents,andProductivity(UniversityofChicagoPress,1984).Hall,B.H.,A.B.Jaffe,andM.Trajtenberg,“TheNBERPatentCitationDataFile:Lessons,InsightsandMethodologicalTools,”workingpaper8498(2001).Harabi,Najib,“AppropriabilityofTechnicalInnovations:AnEmpiricalResearchPolicy24:6(1997),981–992.Heckman,J.J.,H.Ichimura,J.Simith,andP.Todd,“SourcesofSelectionBiasinEvaluatingSocialPrograms,”ProceedingoftheNationalAcademyofSciencesoftheUnitedStatesofAmericaV93(1996),Helpman,Elhanan,“Innovation,Imitation,andIntellectualProperty61(1993),1247–1280.DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?451 Horowitz,Andrew,andEdwinLai,“PatentLengthandtheRateofInternationalEconomicReview37(1996),785–801.Jaffe,Adam,andJoshLerner,InnovationandItsDiscontentsUniversityPress,2004).Johnson,Richard,andDeanWichern,AppliedMultivariateStatisticaled.3(Prentice-Hall,1992).Kaufer,Erich,TheEconomicsofthePatentSystem(HarwoodAcademicPublishers,1989).Kortum,Samuel,andJoshLerner,“StrongerProtectionorTechnologicalRevolution:WhatIs,BehindtheRecentSurgeinPatenting?”Carnegie-RochesterConferenceSeriesonPublicPolicy48(1998),Kumar,Nagesh,“IntellectualPropertyProtection,MarketOrientationandLocationofOverseasR&DActivitiesbyMultinationalEnter-WorldDevelopment24(1996),673–688.Lanjouw,Jean,“TheIntroductionofPharmaceuticalProductPatentsinIndia,”NBERW6366(1998).Lanjouw,Jean,andIainCockburn,“DoPatentsMatter?EmpiricalEvi-denceAfterGATT,”NBERworkingpaperW7495(2000).LaPorta,Rafael,FlorencioLopez-de-Silanes,AndreiShleifer,andRobertVishney,“LawandFinance,”NBERworkingpaperW5661(1996).Lerner,Josh,“150YearsofPatentOfcePractice,”NBERworkingpaperW7477(2000a). “150YearsofPatentProtection,”NBERworkingpaperW7478Levin,R.,A.Klevorick,R.Nelson,andS.Winter,“AppropriatingtheReturnsfromIndustrialResearchandDevelopment,”BrookingsPapersonEconomicActivity3(1987),783–820.Lo,Shih-Tse,“StrengtheningIntellectualPropertyRights:Experiencefromthe1986TaiwanesePatentReforms,”UniversityofCalifor-nia,LosAngeles,mimeograph(2004).Love,James,“AccesstoMedicineandCompliancewiththeWTOTRIPSAccord,”paperfortheUnitedNationsDevelopmentProgrammeConsumerProjectonTechnology,http://www.cptech.org(2001).Manseld,E.,“PatentsandInnovation:AnEmpiricalStudy,”TheInstituteofManagementSciences(1986).Manseld,Edwin,MarkSchwartz,andSamuelWagner,“ImitationCostsandPatents:AnEmpiricalStudy,”TheEconomicJournal(December1981),907–918.Maskus,Keith,IntellectualPropertyRightsintheGlobalEconomy,InstituteforInternationalEconomics(2000).McFetridge,DonaldG.,“IntellectualProperty,TechnologyDiffusionandGrowth,”CarletonUniversitymimeograph(1996).Moser,Petra,“HowDoPatentLawsInuenceInnovation?EvidenceFromNineteenth-CenturyWorldFairs,”NBERworkingpaper9909(2003).Pazderka,Bohumir,“PatentProtectionandPharmaceuticalR&DSpend-inginCanada,”CanadianPublicPolicy25:1(1999),29–46.Putnam,Jonathan,“TheValueofInternationalPatentRights,”unpub-lisheddissertation,YaleUniversity(1996).Qian,Yi,“DoAdditionalNationalPatentLawsStimulateDomesticInnovationinanInternationalPatentingEnvironment?”Harvardworkingpaper,http://kuznets.fas.harvard.edu/ “PricingandMarketingImpactsofEntrybyCounterfeitersandImitators,”Harvardworkingpaper(2005).Rosenbaum,P.(1999)“UsingQuantileAveragesinMatchedObserva-tionalStudies,”AppliedStatistics48(1999),63–78. ObservationalStudies,2nded.(NewYork:Springer,2002).Rosenbaum,P.R.,andD.B.Rubin,“ReducingBiasinObservationalStudiesUsingSubclassicationonthePropensityScore,”oftheAmericanStatisticalAssociation79(1984),516–524. “ConstructingaControlGroupUsingMultivariateMatchedSam-plingMethodsthatIncorporatethePropensityScore,”JournaloftheAmericanStatisticalAssociation39(1985),33–38.Rubin,Donald,“TheUseofMatchedSamplingandRegressionAdjust-menttoRemoveBiasinObservationalStudies,”(1973),184–203.Rubin,Donald,andNealThomas,“CombiningPropensityScoreMatch-ingwithAdditionalAdjustmentsforPrognosticCovariates,”Jour-naloftheAmericanStatisticalAssociation95(2000),573–585.Sakakibara,Mariko,andLeeBranstetter,“DoStrongerPatentsInduceMoreInnovation?Evidencefromthe1988JapanesePatentLawReforms,”NBERworkingpaper7066(1999).Scherer,F.M.,“TheEconomicEffectsofCompulsoryPatentLicensing,”NewYorkUniversityGraduateSchoolforBusinessAdministrationScherer,F.M.,andS.Weisburst,“EconomicEffectsofStrengtheningPharmaceuticalPatentProtectioninItaly,”InternationalReviewofIndustrialPropertyandCopyrightLaw26(1995),1009–1024.Scotchmer,Suzanne,andJerryGreen,“NoveltyandDisclosureinPatentLaw,”RandJournalofEconomics21(1990),131–146.Trajtenberg,M.,“APennyforYourQuotes:PatentCitationsandtheValueofInnovations,”TheRandJournalofEconomics21:1(1990),UnitedNationsConferenceonTradeandDevelopment,TheTRIPsAgree-mentandDevelopingCountries(Geneva:UNCTAD,1996).DATAAPPENDIXCountrycovariatesPATandPATMOD—Thesetwoindicatorvariablesareconstructedfortheperiodsinthisstudybycross-referencingseveraldifferentsources.AmongthemaretheGinarteandPark(1997)patentcoverageindex,theWIPOdocumentsonharmonizingpatentlaws,thecountryreportsand“Super301list”publishedbytheU.S.DepartmentofCommerce,andthecompiledpatentlawsfolioattheHarvardLawlibrary.TheGinarteandParkindexcoversseventyofthecountriesinmysample.Iassignedtheindicatorvaluesfortheremaining22countriesbylookinguptheothersourcesmentionedabove.Allsourcesarelistedinthereferencessection.Inafewcases,wheredifferentsourcesprovideconictinginformationonwhenaparticularcountrystartedimplementingpharmaceuticalpatents,Igainedclaricationbycontactingindividualpatentofces.GDP,realGDPgrowth,andGDPpercapitaPPP—Thesethreevariablesareavailablefor85outofthe92samplingcountries.TheyarefromtheWorldDevelopmentIndicator(WDI)databasepublishedbytheWorldBank(2000).Thisdatabasecontainsdataforover200countriesfrom1960to2000,althoughtheGDPpercapitaPPPdataareonlyavailablefrom1975onwards.Averageyearsofschoolingfortotalpopulation—Thiseducationat-tainmentvariablehasdatafor65samplingcountries(Barro&Lee,2000).Dataareavailableatve-yearintervalsin1960–1999.Economicfreedom—EstimatedusingtheFraserInstitutecompositeindex,whichtakesintoaccountanumberofgovernmentpoliciesandopennessfactors.Acountrywithmoreeconomicfreedomfacilitatesmoretradeandothereconomicexchanges,andthereforefacilitatesmoreLegalfamily—Thisfamilyofdummiesidentifythelegaloriginofthecompanylaworcommercialcodeofeachcountry.TheveoriginsareEnglishcommonlaw,Frenchcommercialcode,Germancommercialcode,Scandinaviancommercialcode,andSocialist/Communistlaws(LaPortaetal.,1996).Innovativepotential—Thevariabletakesvalue6iftheU.S.patentawards(inallindustriesexceptthepharmaceuticalsinayear)surpass1,000;5ifpatentcountisunder1,000butgreaterthan100;4ifitisunder100butgreaterthan6;3ifitisbetween6and1;and1ifnopatentisawardedatall.Pharmaceuticalindustryemploymentandoutput—ExtractedfromtheIndustrialStatisticsCD-ROMpublishedbytheUnitedNationsIndustrialDevelopmentOrganization(UNIDO).Thedatabaseliststhesevariablesbycountryandbyindustry(asclassiedbythefour-digitInternationalStandardIndustrialClassicationcodes),coveringtheyears1978–1999.Thepharmaceuticalindustryislistedunderthecode3522.Comprehen-siveasitis,thedatabasestillhasmanymissingvalues.Theemploymentandoutputvariablesareobservedforonlyftyofthesamplingcountries.DataforthesetwovariablesareaugmentedwiththeOECDStructuralAnalysis(STAN)database,theUNIndustrialStatisticsYearbook,andnationalstatisticalabstractsofsomecountries.IndustryoutputdatafromtheseextrasourcesareconvertedtoU.S.dollarsusingtheannualex-changeratespublishedintheInternationalFinancialStatistics(IFS).TheFordetailedcomponentsoftheindex,pleaseseehttp://www.ThesethresholdvaluesaretakenbytabulatingthequartilesofthevariableontotalU.S.patentsinallotherindustries.THEREVIEWOFECONOMICSANDSTATISTICS compatibilityofthedatafromthesedifferentsourcesisveriedusingarandomsampleofcountrieswheredataareavailableinallthesesources.USFDI—Becauseofcondentialityformultinationalenterprises,thedetailedassetandR&Ddataattheforeignsubsidiariesinthepharmaceu-ticalindustryarenotreleased.Instead,Dr.FritzFoleyattheBEAkindlyreleasedtheU.S.foreignsubsidiarycountslistedbycountry,andhealsoprovidedmewiththeinformationthatthecorrelationsbetweenthesesubsidiarycountsandassets(andR&Doutlays)whencomputedyearbyyearliebetween0.724and0.934.Thus,thesubsidiarycountscanactasanestimateforthetechnologytransferfromtheUnitedStatestothedifferentcountries.JapaneseFDI—Similarly,IobtainedtheJapaneseforeignsubsidiarycountsdatainthepharmaceuticalindustryfromProfessorPaulBeamishattheUniversityofWesternOntario.BecausetheUnitedStatesandJapaneseR&Dspendingpersubsidiarycanbequitedifferent,Idecidedtokeepthetwocountsasseparatevariablesratherthanmergingthemtogetherintheregressionspecications.Thesetwovariablesarefullyobservedforallthe92samplingcountries.Country’spharmaceuticalexportstotheUnitedStates—TheWorldTradeAnalyzerdatabaseproducedbyStatisticsCanadaprovidesadatasourcefortheimportsoftheUnitedStateslistedbycountryandbyindustryfrom1980to1996.ThedatabaseliststhedifferentindustriesaccordingtotheStandardInternationalTradeClassication(SITC),underwhichcode54correspondstothemedicinalandpharmaceuticalproducts.Sincethedatabasecoversonlythemanufacturingindustries,thepharma-ceuticalexportsrefertothemanufacturedproducts.ThedataforthepharmaceuticalexportstotheUnitedStates(or,equivalently,U.S.im-ports)areavailableforallthesamplingcountries.Pricecontrol—Thisvariableisconstructedbycross-referencingsev-eralsources,includingthecountryreportspublishedbytheU.S.Depart-mentofCommerce,theEconomistIntelligenceUnit(EIU)database,thepricecontrolcomponentoftheFraserInstituteeconomicfreedomindex,theOECDreportforthepharmaceuticalindustry,Danzon(1997),andotherdocumentssearchedinGoogle.IPRstrengthscore—Itisanunweightedsumofvecomponentindices,includingthedomesticpatentabilityofsevenproductcategories,themembershipacountryhasininternationalagreements,thedurationofnationalpatentprotection,protectionoflossesfromcompulsorylicensing,revocationofpatents,andsoon,andtheenforcementevaluationofapatentsystem(Ginarte&Park,1997).Theinclusionofthisvariablealsohelpstocontrolfortheenforcementofnationalpharmaceuticalpatentlaws,correctingpossibleloopholesthatthesimplepatentcoverageindi-catordoesnotaccountfor—thelackofenforcementofnationalpatentlawsinsomecountries.—Itisacategoricalvariablethattakesonvalues1to5forthecorrespondingquintilesoftheIPRstrengthscore.FivedummyvariablesforeachquintileofIPRarethengeneratedbytabulatingtheAppendicesI.B,I.C,II,andIIIareincludedintheworking-paperversionandarepostedonmyWebpage:http://www.nber.org/DONATIONALPATENTLAWSSTIMULATEDOMESTICINNOVATION?453