innovation, patents, productivity in firmsbhhall/papers/bhh13_industry_canada... · innovation,...

Innovation,patents,andproductivityinfirms

BronwynH.HallUniversityofMaastrichtandUC

Berkeley

April2013 IndustryCanada 1

Abigtopic beginwithdefinitions Innovation:theimplementationofaneworsignificantly

improvedproduct(goodorservice),orprocess,anewmarketingmethod,oraneworganisational methodinbusinesspractices,workplaceorganisation orexternalrelations.(OsloManual,OECD2005,thirdedition,p.46)

Patents:therighttoexcludeothersfrompracticinganinventionforalimitedamountoftime,inreturnfordisclosingtheinventiontothepublic.

Productivity: theamountofoutputthatcanbeobtainedfromagivensetofinputs. TFP =totalfactorproductivity(controlsforallinputs) Laborproductivity=outputperpersonorpersonhour


Stylizedpictureofinnovationintheprivatesector


Diffusionprocess

Marketstructureandindustry;appropriabilityenvironment

Knowledgecapitalcreatedbyinnovationinvestment

R&Dandotherinnovationinvestments

Outcomes:Productivity,Profitability,

andEconomicGrowth

Firmsizeandmarketshare,diversification,andexperience

Demandpull(taste,marketsize)

PhysicalcapitalWorkerskills

Patents,otherIPR

Innovationoutput

Innovative sales

Techopportunity(sciencebase)

Focusofthispresentationontheboxesinred

Introduction Whatarethemechanismsconnectinginnovationand

patents? Innovationbasedonnovelinventionencouragedbythepatent

system(limitedmonopoly) =>patentsmayserveasindicatorsofcertainkindsof

innovation

Whatarethemechanismsconnectinginnovationandproductivity? Improvementswithinexistingfirms

Creationofnewgoods&services,leadingtoincreaseddemandforfirmsproducts

Processandorganizationalinnovationleadingtoefficiencygainsinproduction

Entryofmoreefficientfirmsandfirmsontechnologyfrontier ExitoflessefficientfirmsApril2013 IndustryCanada 4

Twoquestions

1. Whatistherelationshipbetweeninnovationandtotalfactorproductivity(TFP)?

2. Doesthepatentsystemincreaseinnovativeactivity?

This talk reviews what economists know about theanswers to these two questions.

Note: topic is broad and omissions inevitable.


Problem1:Measuringinnovation

LargeliteratureusingR&D(capital)asaproxyforinnovationinput Hall,Mairesse,Mohnen2010 survey,interalia

Smallerliteratureusingpatentsasaproxyforintermediateinnovationoutput

Bothmeasureshaveweaknesses,especiallyoutsidemanfacturingsector.

Nowwehavemoredirectmeasures dotheyhelp?


Innovationsurveyscontain..

Dataoninnovation: Productorprocessnewtofirm/market(yes/no) Shareofsalesduringpast3yearsfromnewproducts Latersurveyshaveexpendituresonvariouskindsofinnovationinvestments(answerscanbemissingornoisy)

Dataonproductivity: Oftensalesperworker(laborproductivity) SometimesTFP(adjustedforchangesincapital) Issuesarisingfromdeflationandlevelofaggregation

NexttwofiguresaddUSdatatographsproducedbyOECDinMeasuringInnovation



WhereisCanada?Difficulttosayexactly,butprocess~17%;product~1825%

Thereisaroughrelationshipbetweenshareofinnovatingfirmsandaggregatelaborproductivity


Problem2:MeasuringTFP

Twoapproaches: Growthaccounting usesharesgoingtolabor,capital,etc.toestimatetheirproductivity,subtractfromoutputtogetresidual

Regression outputonlabor,capital,materials,etc toobtaincoefficientestimates;computeresidual

Measurementissues: Qualityadjustmentforinput variables affectsallocationofproductivitygains

Usualdatagivessalesdeflatedbyindustryleveldeflator,butinnovativeactivitymayaffectfirmoutputquality(andpossiblymarketpower)


Reviewingtheevidence

FocushereonmicroevidenceusingCDMmodel(CreponDuguetMairesse1998) SomenonR&DfirmsinnovateandsomeR&Dfirmsdonotinnovate(duringa3yearperiod)

Dataisusuallycrosssectional,sosimultaneitybetweenR&D,innovation,andproductivity

Modelattemptstoaccommodatethesefeaturesofthedata


AbriefoverviewoftheCDMmodel

Threeblocksofequations1. equationsexplainingtheR&Ddecisionandthe

amountofR&Dperformed2. Innovationoutputequations(KPF)withR&Das

input3. Productivityequation,inwhichinnovation

outputindicatorsappearasexplanatoryvariables

Estimationisrecursiveusingsingleequationblocks,orsimultaneous(nofeedback)


WhathavewelearnedfromapplyingtheCDMmodeltoCISdata?

estimatedfor~1520countries confirmedhighratesofreturntoR&Dfoundinearlierstudies

Likepatents,innovationoutputstatisticsaremuchmorevariable(noisier)thanR&D,andR&Dtendstopredictproductivitybetter

Backupslidessummarizeresultsintables.


TFPoninnovativesalesshare

Robustlypositive,supportstheviewthatproductinnovationshiftsthefirmsdemandcurveout Elasticitiesrangefrom0.04to0.29withatypicalstandarderrorof0.03

Kintensiveandhitechfirmshavehigherelasticities(equalizedratesofreturn)

Coefficientofprocessinnovationdummyusuallyinsignificantornegative,suggestingmarketpower profitsmayincrease,butrevenueproductivitydoesnot(seealgebrainbackupslides)


TFPlevelresultswithdummies

Productdummysupportsinnovationsalesshareresult,althoughnoisier.

Processinnovation: Withproductinnovationincluded,processinnovationoftennegativeorzero

Withoutproductinnovation,processinnovationpositiveforproductivity

Substantialcorrelationbetweenproductandprocessinnovation,especiallywhentheyareinstrumented. Possiblymisleadingresultsinsomeoftheliterature


Aggregation

Howdoesindividualfirmrelationshipaggregateuptomacroeconomy?

Foster,Haltiwanger,andSyverson (2008)distinguishrevenueandquantityproductivity,butincludeexit&entryeffects Revenueproductivitytendstounderstatethecontributionofentrantstoproductivitygrowth(becausetheirpricesarelower)

Demandvariationisamoreimportantdeterminantoffirmsurvivalthanefficiencyinproduction(consistentwithproductivityimpacts)


Entryandexit Aghionetal.(2009);Gorodnichenkoetal.(2010)

Competitionandentryencouragesinnovationunlessthesectorisveryfarbehind

Djankov(2010)survey: strongerentryregulationand/orhigherentrycostsareassociatedwith

fewernewfirms,greaterexistingfirmsizeandgrowth,lowerTFP,lessinvestment,andhigherprofits

Bartelsmanetal.(2010),andsubsequentresearch: Sizeproductivitymorehighlycorrelatedwithinindustryifregulationis

efficient EvidenceonEasternEuropeanconvergence

Fullsetoflinksbetweeninnovation,competition,entry,andproductivitygrowthnotyetexplored.


Addingpatentstothepicture

Dopatentsprovideanincentiveforinnovativeactivity?

Simpleeconomicview: Tradeofflimitedtermrighttoexclude(monopoly)inreturnforincentivetoinnovate(andrevealtheinnovation)

Goodforinnovation Badforcompetition

ButApril2013 IndustryCanada 18

Patentsystemasviewedbyatwohandedeconomist

Effects on Positive NegativeInnovation creates an incentive

for R&D and innovation investments

impedes the combination of new ideas & inventions; raises transaction costs;inhibits cumulative invention

Competition facilitates entry of new or small firms with limited assets; enables vertical disintegration

creates short-term monopolies, which may become long-term in network industries

IndustryCanada 19April2013

Patentsmayinhibitinnovation

Thepatentthicket problemofcontractingwhenmanyinputsareessential Hightransactioncostsleadtobreakdown Negotiationsfailduetoholduppotential Discouragesentry(increasessunkcosts)

Largenumbersofpatentsinagivenarea,impossibilityofadequatesearch Expostholdupbypatentholderaftercostsaresunk Givenlitigationcosts,eveninvalidpatentscanbeenforced

Increasestheriskofinnovation


Patentsmayhelpcompetition

Increasedynamiccompetitionbyfacilitatingentry Usefulforsecuringfinancinginknowledgeintensiveindustries(wheretherearefewtangibleassets)

Downside searchforsalvagevaluehasledtoPAE/NPEactivityintheUS,madeprofitablebyholduppotential

Canleadtocompetitionenhancingverticaldisintegrationbyfacilitatingtradeintechnology(specialization;interfacestandardization) Chemicals Arora,Fosfuri,Gambardella Semiconductordesignfirms Hall&Ziedonis


Whendopatentsencourageinnovation?

Theory Whenoneproduct=onepatent Whenoneproduct=manypatents? notclear Whenoneinventionbuildsonanother? notclear

Empiricalevidence Historicalinvestigationsofchangesinpatentsystems Firmsurveys Crosscountryrehgressions


Historicalevidence 19thcentury(variationacrossEurope/US)

Moser(2005) littleeffectonoverallinnovation,butchangeinfocus

Lerner(2001) increaseinpatentingbyforeignersbutnoincreasebyfirmswithincountryorinBritain(thatis,noincreaseininnovation)

20thcentury ParkandGinarte 60countries,196090.StrengthofIPR

(includingcoverageofpharmaceuticals)positiveforR&Dindevelopedcountries

Branstetter&Sakakibara increasingpatentscopeinJapan(1988)didnotincreaseR&D

Baldwinetal Canadianinnovationsurvey.Innovationcausespatenting,butpatentingdoesnotseemtoincreaseinnovation


Surveyevidence IndustrialR&DmanagersintheUS

Yalesurvey(Levin,Klevorick,Nelson,andWinter1983) CarnegieMellonsurvey(Cohen,Nelson,andWalsh1994)

EUinnovationsurveys 1993CISforNorway,Germany,Luxembourg,theNetherlands,

Belgium,Denmark,andIreland 2,849R&Dperformingfirms(reportedinArundel2001)

patentsnot themostimportantmeansofsecuringreturnstoinnovation Only~10%ofrespondentsratethemfirstorsecond Exceptions:pharmaceuticals,specialtychemicals,medicalinstruments,autoparts


Ausefultaxonomy discreteproductindustries

food,textiles,chemicalsincludingoilandplastics,pharmaceuticals,metals,andmetalproducts

patentsusedtoexclude,andsometimesforlicensing;alsotopreventlitigation

complexproducttechnologies machinery,computers,software,electricalequipment,electronic

components,instruments,andtransportationequipment patentsusedinnegotiations(crosslicensingandother),andto

preventlitigation Ingeneral,patentsmoreimportantforappropriabilityin

discreteproductindustries Strategicuses(crosslicensing,negotiations)greaterin

complexproductindustries


Summary

1. Whatistherelationshipbetweeninnovationandtotalfactorproductivity(TFP)?

Positiveforthemostpart,buttheavailableinnovationmeasuresareverynoisy,sopreciseanswersarenotpossible.

2. Doesthepatentsystemincreaseinnovativeactivity? Theroleofpatentsinencouraginginnovationisambiguous

Positiveonbalanceindiscreteproductindustries Neutralornegativeincomplexproductindustries BUTconsiderableheterogeneitywithinindustry

Patentsmayactuallyhelpcompetitioniftheyfacilitateentryorleapfrogging


BACKUPSLIDES


Productivity

Conventionalsetup:

q =logvalueadded(sometimesjustoutput)c =logtangiblecapitall =loglaborinputait =TFP(totalfactorproductivity)Coefficients,measuredasshares(growthaccounting)orbyregression(ashere)


it it it itq a c l i entity t time ,

Revenueproductivity

Iffirmshavemarketpowerandidiosyncraticprices,weobserverealrevenuer,notoutputq, with r=p+q(allinlogs)

Adddemand:qit =pit , >1),revenuefallswithincreasedoutput,althoughprofitmayrise


1it it it itr a c l

( )

Addinginnovation

Addtwoterms:kit intheproductionfunction,>0kit inthedemandfunction,>0

Thisyieldsthefollowingrevenuefunction:

Productimprovement(/)alwayspositiveProcessimprovement((+1)/)canbenegative


1 1it it it it itr a c l k

( )

SomepapersestimatingtheCDMmodelandvariants

Crepon,Duguet,Mairesse1998 Duguet2006 Loofetal2001 Janzetal2003 LoofandHeshmati2003 CriscuoloandHaskel2003 HuergoandJaumandreu2004 Benavente2006

Jefferson,Baietal2006 LoofandHeshmati2006 VanLeeuwenandKlomp2006 Parisietal2006 Griffithetal2006 Mairesseetal2009 Polderetal2009 MairesseandRobin2010 Halletal2011


Levelvsgrowth

CDMisintermsofproductivitylevelsprimarilyfordataavailabilityreasons

Itseemsmorenaturaltothinkofinnovativeactivityasaffectingproductivitygrowth

Someworkalongtheselines,butmatchingacrosssurveysusuallyleavesaveryselectedsample,possiblynotrepresentative


ProductivityinnovationrelationshipinTFPlevels


Sample TimeperiodElasticitywithrespecttoinnovsalesshare

Processinnovationdummy

Chileanmfgsector 19951998 0.18(0.11)*ChineseR&Ddoingmfgsector 19951999 0.035(0.002)***Dutchmfgsector 19941996 0.13(0.03)*** 1.3(0.5)***Finnishmfgsector 19941996 0.09(0.06) 0.03(0.06)Frenchmfgsector 19861990 0.07(0.02)***FrenchHitechmfg# 19982000 0.23(0.15)* 0.06(0.02)***FrenchLowtechmfg# 19982000 0.05(0.02)*** 0.10(0.04)***GermanKintensivemfgsector 19982000 0.27(0.10)*** 0.14(0.07)**Irishfirms# 20042008 0.11(0.02)*** 0.33(0.08)***Norwegianmfgsector 19951997 0.26(0.06)*** 0.01(0.04)SwedishKintensivemfgsector 19982000 0.29(0.08)*** 0.03(0.12)Swedishmfgsector 19941996 0.15(0.04)*** 0.15(0.04)***Swedishmfgsector 19961998 0.12(0.04)*** 0.07(0.03)***Swedishservicesector 19961998 0.09(0.05)* 0.07(0.05)

Source:author's summaryfromAppendixTable1.#Innovativesales shareandprocess innovationincludedseparatelyintheproductionfunction.

ProductivityinnovationrelationshipinTFPlevels


Sample Timeperiod Productinnovationdummy


Argentinianmfgsector 19982000 0.22(0.15)Brazilianmfgsector 19982000 0.22(0.04***Estonianmfgsector 19982000 0.17(0.08)** 0.03(0.09)Estonianmfgsector 20022004 0.03(0.04) 0.18(0.05)***Frenchmfgsector 19982000 0.08(0.03)**Frenchmfgsector 19982000 0.06(0.02)*** 0.07(0.03)**Frenchmfgsector 19982000 0.05(0.09) 0.41(0.12)***Frenchmfgsector 20022004 0.08(0.13) 0.45(0.16)***Frenchservicesector 20022004 0.27(0.52) 0.27(0.45)Germanmfgsector 19982000 0.05(0.03) 0.02(0.05)Irishfirms# 20042008 0.45(0.08)*** 0.33(0.08)***Italianmfgsector 19952003 0.69(0.15)*** 0.43(0.13)***ItalianmfgsectorSMEs 19952003 0.60(0.09)*** 0.19(0.27)Mexicanmfgsector 19982000 0.31(0.09)**Spanishmfgsector 20022004 0.16(0.05)***Spanishmfgsector 19982000 0.18(0.03)*** 0.04(0.04)Swissmfgsector 19982000 0.06(0.02)***UKmfgsector 19982000 0.06(0.02)*** 0.03(0.04)

ProductivityinnovationrelationshipinTFPgrowthrates


Sample Timeperiod ElasticitywrtInnovsalesshare

Productinnovationdummy


Argentinianmfgsector 19922001 0.09(0.08) 0.18(0.08)**Dutchmfgsector 19941998 0.009(0.001)*** 1.2(0.7)*Dutchmfgsector 19961998 0.0002***#Frenchmfgsector 19861990 0.022(0.004)***Germanmfgsector 20002003 0.04(0.02)** 0.14(0.08)*@Italianmfgsector 19921997 0.12(0.09) 0.04(0.12)Spanishmfgsector 19901998 0.015(0.004)***Swedishmfgsector 19961998 0.07(0.03)**Swedishservicesector 19961998 0.08(0.03)***UKmfgsector 19941996 0.02(0.02) 0.02(0.01)*UKmfgsector 19982000 0.07(0.03)** 0.04(0.02)**

#elasticitywithrespecttoinnovationexpenditurepersales.@elasticitywithrespecttocostreductionperemployee.

Source:author's summaryfromAppendixTable1.

innovation, patents, productivity in firmsbhhall/papers/bhh13_industry_canada... · innovation,...

Documents