Innovation and Entrepreneurship in Renewable Energy

Transcription

1 Innovation and Entrepreneurship in Renewable Energy Ramana Nanda Harvard Business School Ken Younge Purdue University Lee Fleming University of California, Berkeley College of Engineering University of California, Berkeley Fung Technical Report No July 18, Blum Hall #5580 Berkeley, CA (510)

2 The Coleman Fung Institute for Engineering Leadership, launched in January 2010, prepares engineers and scientists from students to seasoned professionals with the multidisciplinary skills to lead enterprises of all scales, in industry, government and the nonprofit sector. Headquartered in UC Berkeley s College of Engineering and built on the foundation laid by the College s Center for Entrepreneurship & Technology, the Fung Institute combines leadership coursework in technology innovation and management with intensive study in an area of industry specialization. This integrated knowledge cultivates leaders who can make insightful decisions with the confidence that comes from a synthesized understanding of technological, marketplace and operational implications. Lee Fleming, Faculty Director, Fung Institute Advisory Board Coleman Fung Founder and Chairman, OpenLink Financial Charles Giancarlo Managing Director, Silver Lake Partners Donald R. Proctor Senior Vice President, Office of the Chairman and CEO, Cisco In Sik Rhee General Partner, Rembrandt Venture Partners Fung Management Lee Fleming Faculty Director Ikhlaq Sidhu Chief Scientist and CET Faculty Director Robert Gleeson Executive Director Ken Singer Managing Director, CET Copyright 2013, by the author(s). All rights reserved. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission. 130 Blum Hall #5580 Berkeley, CA (510)

3 InnovationandEntrepreneurshipinRenewableEnergy RamanaNanda(HBS),KenYounge(Purdue) andleefleming(ucberkeley) 1 July2013 Abstract Wedocumentthreefactsrelatedtoinnovationandentrepreneurshipinrenewableenergy.First,we comparepatentingbyventurebackedstartupsandincumbentfirms,usingdatafromtheuspatentand TrademarkOffice.Usingavarietyofmeasures,wefindthatVCbackedstartupsareengagedinmore novelandmorehighlycitedinnovations,comparedtoincumbentfirms.incumbentfirmsalsohavea highershareofpatentsthatarecompletelyuncitedorselfcited,suggestingthatincumbentsaremore likelytoengageinincrementalinnovationcomparedtovcbackedstartups.second,wedocumenta rising share of patenting by startups that coincided with the surge in venture capital finance for renewableenergytechnologiesintheearly2000s.wealsoshowthattheavailabilityofventurecapital financeforrenewableenergyhasfallendramaticallyinrecentyears,withimplicationsfortherateand trajectory of innovation in this sector.finally, we highlight a number of structural factors about renewableenergythatmakeithardtoattractsustainedfinancingfromventurecapitalinvestorsand suggestpoliciesthatmightfacilitateinnovationandentrepreneurshipinrenewableenergy. 1 WethankGuidoBuenstorf,RonnieChaterji,JasonDavis,ChuckEesley,MazharIslam,MattMarxandthe participantsatthenberpreconferenceandstanfordsocialscienceandtechnologyseminarforhelpful comments.dandipaoloprovidedexcellentresearchassistanceandguanchengliprogrammedthegeographical visualizations.forsupportwethankthedivisionoffacultyresearchandfundingatharvardbusinessschool,the FungInstituteforEngineeringLeadershipatUCBerkeley,andtheNationalScienceFoundation( ).

4 1. Introduction Theglobaldemandforenergyisprojectedtoalmosttripleoverthenextseveraldecades.Estimates suggestthatagrowingworldpopulation,combinedwithrisinglivingstandardswillleadglobalenergy consumptiontoreachabout350,000twhin2050fromthe2010levelof130,000twh.toputthis increaseinperspective,itwillrequiretheequivalentofsettingup750largecoalburningpowerplants peryearfor40yearsinordertomeettotheincreaseddemandforenergyinthecomingdecades. Figure1providesabreakdownofthesourcesthatproducedtheenergyconsumedin2010basedon datafromthebpstatisticalreviewofworldenergy,2012.ithighlightsthat87%oftheenergywas producedfrom conventionalenergy,namelycoal,oilandnaturalgas.ontheotherhand,solar, Wind,Biomass,Hydroandotherrenewablesaccountedforamere8%ofglobalenergyproducedin Inadditiontothechallengesofmeetingthegrowingenergyneedsoftheworld spopulationwith conventionalsourcesofenergy,theimplicationsofcontinueddependenceonfossilfuelsarebelievedto beparticularlystarkforclimatechange.theshalegasrevolutionintheusinrecentyearshasimplieda reduceddependenceoncoal.nevertheless,thebenchmarkoftryingtoachieve zeroemissions has ledtheusandseveraleuropeancountriestofocusmoreintenselyonpromotinginnovationin renewableenergytechnologiesinrecentyears.whilethereisnoclearwinningalternativeatpresent, thereisalsoagrowingbeliefthatprogresswillcomefromradicalinnovationsthatwillallowustomake thejumpfromthestatusquo,whetheritisinrenewableenergyorothermoreconventionalsourcesof energyproduction.weexaminethetechnologicalandorganizationalsourcesofsuchinnovationin renewableenergy,withaneyetowardsunderstandingthefactorsthathavepromotedprogressinthe past. VenturecapitalhasbeenakeysourceoffinanceforcommercializingradicalinnovationsintheUnited States,particularlyoverthelastthreedecades(KortumandLerner,2000;GompersandLerner,2002; SamilaandSorenson,2011).Theemergenceofnewindustriessuchassemiconductors,biotechnology andtheinternet,aswellastheintroductionofseveralinnovationsacrossaspectrumofsectorssuchas healthcare,itandnewmaterials,havebeendriveninlargepartbytheavailabilityofventurecapitalfor newstartups.akeyattributeofventurebackedinnovationintheushasbeentheabilityofprivate capitalmarketstofinanceawidevarietyofapproachesinaspecificarea,asopposedtochoosinga specificwinner.sinceitishardtoknow,exante,whichtechnologicaltrajectorywillbesuccessfulex post,itseemsvitalthatinordertomakerapidtechnologicalprogress,weneedtoproceedby conductingnumerous economicexperiments intheenergysector(rosenberg,1994;stern2003;kerr, 2 TheBPstudyonlyreportsdataoncommerciallytradedfuels,includingrenewableenergythatiscommercially traded.theinternationalenergyagency(iea)estimatesaslightlyhighershareofrenewablesbasedonestimates oftheuseofwoodchips,peatandotherbiomassusedindevelopingcountriesthatisnotcommerciallytraded. Evenso,theirestimateofrenewablesincludinghydroelectricityis13%,comparedtothe8%estimatedbyBP.

5 NandaandRhodesKropf,2013).Thismakesventurecapitalanidealcandidatetoplayarolein financingradicalinnovationinrenewableenergytechnologies. Infact,venturecapitalfinancingforrenewableenergystartupsrosedramaticallyinthemid2000safter beingconsistentlylowinthepreviousdecades.figure2adocumentsthatbetween2006and2008, severalbilliondollarswerechanneledintostartupsfocusedonsolar,windandbiofuels.inthelastfew years,however,venturecapitalinvestmentinrenewableenergytechnologieshasplummeted,fallingas ashareofoverallvcinvestmentandevenwithincleantech,shiftingawayfromrenewableenergy productiontoinvestmentsinenergyefficiency,software,andstorage(asseeninfigures2band2c). Inthischapter,weinvestigatetheroleofventurecapitalinrenewableenergyinnovationbycomparing thepatentingactivityofvcbackedstartupswithothertypesoforganizationsengagedinrenewable energyinnovation.wenotonlyexaminepatentingrates,butalsothecharacteristicsofthepatents beingfiledbythedifferenttypesoforganizations.understandingthesefactorswillhelpdeterminethe extenttowhichfallingvcinvestmentinrenewableenergyshouldbeseenasacauseforconcernas opposedtobeingeasilysubstitutablebyinnovationbyotherssuchaslargeincumbentfirms. WeaddressthesequestionsbyusingpatentdatafromtheU.S.PatentandTrademarkOffice(USPTO) overthethirtyyearperiodfrom1980through2009.wefindthatlargeincumbentfirmshave dominatedpatentinginrenewableenergyforseveraldecades.forexample,thetop20firmsaccounted for50%oftherenewableenergypatentsandthetop50firmsaccountfornearly70%ofsuchpatents filedattheusptointheearly1990s.innovationbecamemorewidespreadin2000swhenpatentingby VCbackedfirmsgrew,butthetop20firmsstillaccountedfor40%ofthepatentingactivityin2009. However,despiteaccountingforthelargestshareofpatents,wefindincumbentsaremorelikelytofile patentsthatareeithercompletelyuncitedorareselfcited,suggestingagreaterfocusonincremental orprocessinnovation.furthermore,theyarelesslikelytohaveextremelyinfluentialpatents,thatwe defineasbeinginthetop10percentilesofforwardcitationsinagiventechnologyareaandgivenyear. Finally,wecreateameasureofnoveltyusingtextualanalysisofthepatentdocumentsthatdoesnot dependoncitations.thisindependentmeasurealsosuggeststhatonaverage,incumbentsfirmshave beenengagedinlessnovelpatentingthanventurecapitalbackedstartups,evenmoresointheperiod whenvcfundingforstartupsincreaseddramatically. GiventheimportantrolethatVCbackedstartupsseemtoplayinrenewableenergyinnovationinthe lastdecade,thedramaticfallinfinancingavailableforsuchfirmssuggeststhatitwillhaveimplications forthenatureofinnovationwemayseegoingforwardinthissector.wehighlightthatatleastsomeof thedeclineinvcfinancingforrenewableenergyisrelatedtostructuralfactorsthatmakeitveryhard forvcstofundsuchstartups.wethereforeelaborateontheaspectsoftheinstitutionalenvironment thathavemadesustainedexperimentationbyvcsextremelydifficultinrenewableenergy,andsuggest policiesthatmaybeeffectiveinhelpingtochannelmoreriskcapitaltowardsinnovationand entrepreneurshipinrenewableenergytechnologies. Therestofthechapterisstructuredasfollows.InSection2,weoutlinethedatausedforouranalysis. Section3providesadetaileddescriptionofourmainresultsonthedifferencesininnovationacross

6 incumbentandventurecapitalbackedfirms.insection4,wediscussthechallengesfacedbyventure capitalinvestorsinsustainingthefinancingofrenewableenergystartups.section5offerspoliciesthat maymakeiteasiertofinancesuchstartupsandsection6concludes. 2. Data 2.1. Sampleselectioncriteria Ourfocusinthischapterisonpatentinginsectorsrelatedtorenewableenergyproduction,namely solar,wind,biofuels,hydroelectricpowerandgeothermaltechnologies.beforemovingtoadescription ofthedata,however,wefirstoutlinethecriteriaforselectingoursample. Ourapproachwastodefineasetoftechnologiesthatwouldfirst,allowustobuildacomprehensiveand welldelineateddatasetofpatentingactivitywithinthechosentechnology,andsecond,enableusto comparethecharacteristicsofinnovationbetweenventurecapitalbackedstartupsandotherfirms engagedininnovation. Thisledustoleaveoutsometechnologiesthatareoftenassociatedwithcleanenergyproductionbut arenotrenewableenergy.forexample,althoughnaturalgashasalowercarbonfootprintthanoiland coal,itisdifficulttobreakoutinnovationsrelatedtoenergyproductioninthisarea,asopposedtoother businessespursuedbyoilandgascompanies. 3 Ontheotherhand,wehavealsoleftoutother clean tech sectorsthatreceivevcfinancebutarenotenergyproduction.forexample,venturecapitalhas beeninvolvedinfinancinganumberofinnovationsinsoftwarerelatedtosmartgridandenergy efficiency.theseinnovationsareextremelydifficulttoisolateinasystematicmannerfromother softwarepatentsthatstartupscouldbeworkingon(e.g.agpssoftwarethathelpsroutetrucksina mannerthatconservesfuelishardtodistinguishfromothergpspatents,evenwhenmanually classifyingpatents).ourfocus,therefore,isonrenewableenergyproductiontechnologies.although ourscopeisnarrowerthaneither energyproduction or cleantech,ourhopeisthatourtradeoff buysusgreaterconfidenceindefiningaclearandconsistentsetoftechnologieswithinwhichwecan characterizeboththetrendsinpatentingovertime,andthedifferencesinthenatureofpatenting acrossthevariousorganizationalforms Datausedtocreatethesample Wecreatedoursampleusingthreesteps.First,weworkedwithaprivateresearchfirm,IPCheckups,to defineasetofrenewableenergypatentsattheu.s.patentandtrademarkoffice(uspto)ineachofthe energyproductionsectors,ofsolar,wind,biofuels,hydroandgeothermal.ipcheckupshasparticular expertiseincleanenergy,includingadatabaseofcleantechnologypatentsfiledattheusandforeign 3 Similarly,althoughnuclearenergyisnot renewable,wehavecollecteddataonitgiventheeasewithwhichwe couldclassifypatentsinthistechnologyandthefactthatafewnuclearenergystartupshaverecentlybeenbacked byventurecapital(e.g.seesahlman,nanda,lassiterandmcquade,2012).however,thelevelofpatentingbyvc backedstartupsinnuclearisextremelylowinoursample,sowehavechosennottoreporttheresultsofnuclearin ourmainestimations.however,ourresultsareunaffectedbytheinclusionorexclusionofnuclearenergy.

7 patentoffices(weconsideronlypatentsfiledwiththeuspto).theyprovideduswithasampleof 17,090renewableenergypatentswhoseapplicationdateswerebetweenJanuary1980andDecember 2009acrossthe5subsectorslistedabove. 4 Second,wedevelopedaproceduretovalidateandextendthesamplefromIPcheckupsinorderto ensurethatthesamplewascomprehensive.specifically,weusedthepatentsfromipcheckupsasa trainingset,andappliedtheliblinearmachineclassifieralgorithm(fanetal.,2008),tosearchthrough everypatenttitleandabstractintheuniverseofapprovedutilitypatentsattheusptowithapplication datesbetweenjanuary1980anddecember2009.themachineclassifieralgorithmaimedtoidentify otherpatents(basedontheirtitlesandabstract)thatlooked similar tothoseinthetrainingset providedbyipcheckups.theassumptionbehindthisapproachisthatipcheckupsmayhavemissed patentsatrandom,butwouldnothaveasystematicbiasinthetypesofpatentstheydidnotprovideus. Inthiscase,thealgorithmwouldbeabletosearchefficientlyamongtheover4.3millionpatentsinthe universeofpatentsforotherswithsimilartitlesandabstractsthatmayhavebeenoverlookedbyip Checkups.Theclassifierreturnedanadditional31,712patentsforconsideration. Finally,wecontractedwithIPCheckupstohaveaPh.D.expertincleantechnologiesmanuallyreview eachofthecandidatepatentsidentifiedbythemachineclassifierandselectappropriateonesfor inclusionintothefinalsample.anadditional5,779patentswereselectedforinclusion,resultingina finalsamplesizeof22,869patents. Webelievethatthis3stepprocessoutlinedabovehasproducedoneofthemostcomprehensive samplesofpatentslookingspecificallyatrenewableenergy.giventhesystematic,andreplicable approachusedbythemachinelearningsample,webelievethismethodwillallowsubsequent researcherstoeasilyupdatethesample,aswellasapplysimilartechniquestoidentifypatentsinothers sectorsthatsharethepropertywithrenewableenergyofnoteasilydemarcatedbyspecifictechnology classesattheuspto. Havingthusidentifiedourfiveprimarycategoriesofcleantechpatentsbytechnologytype,wefurther categorizedeachpatentintooneoffourorganizationaltypes:academia&government,vcbacked startups,nonvcbackedfirmsandunassigned.unassignedpatentswerethosewithnoassignee providedinthepatentapplication.thesehavetypicallybeenassumedtobeindependentinventors,but mayalsobecorporatepatentswithjustamissingassigneefield.asweshowinthefollowingsection, unassignedpatentsseemsignificantlydifferentintermsoftheircharacteristics.whilewedoreport someanalysesthatincludeunassignedpatents,themajorityofouranalysesfocusoncomparisons betweenvcbackedstartups,nonvcbackedfirmsandinventorsinacademicinstitutionsor governmentlabs.weclassifiedfirmsasventurecapitalbacked,iftheassigneenameandlocation correspondedwithfirmsineitherthecleantechi3orthebloombergnewenergyfinancedatabaseof 4 AlthoughtheUSPTOdatagoesasfaras2012,wetruncatethesampleattheendof2009toallowforouranalysis offorwardcitations.

8 venturecapitalbackedfinancings. 5 Toclassifyassigneesasuniversityorgovernment,weusedatext matchingprocessfollowedbymanualreviewtoidentifyacademicinstitutions(assigneeswithwords suchas university, universitaet, ecole, regents, etc.)andgovernmentalorganizations(assignees withwordssuchas DepartmentofEnergy, UnitedStatesArmy, LawrenceLivermore, Bundesrepublik, etc.) Ourresidualcategory,therefore,isthecategoryofassigneesthatarenotVCbackedandnotfrom academicinstitutionsorthegovernment.theresidualcategorycanthereforebethoughtofas incumbentfirms(keepinginmindthequalificationsdescribedabove).asfaraspossible,wemanually matchedsubsidiariestotheirparent sname,sothatforexample,allknownsubsidiariesofgewere classifiedasge.whilethiscategorizationisimperfect,caseswherewemissedmatchingasubsidiaryto aparentwilltendtobiasustowardsfindinglessconcentrationinpatentingandourfindingsshouldbe seenasalowerboundtothetruelevelofconcentrationacrossorganizationsinvolvedinrenewable energypatenting. 3. Results 3.1. PatentingRatesinRenewableEnergy Webeginbyprovidinganoverviewofthepatentinglandscapeinrenewableenergytechnologies.Table 1aandTable1bprovideabreakdownofthetotalnumberofpatentsusedinoursample,brokendown byorganizationalformandtechnologyarea.table1areportsthebreakdownfortheentiresample, whiletable1breportstheresultsforinventorswhoarebasedintheus. 6 AscanbeseenfromTable1, about85%ofthepatentsinoursampleareaccountedforbysolar,windandbiofuels.incumbentfirms accountfornearlytwothirdsofthepatentsinthedatasetandabout55%ofthepatentsfiledbyus basedinventors.whiletable1providesasenseofthemostimportanttechnologiesandorganizational forms,itdoesnotgiveasenseofsharesovertimeorthequalityoftheinnovations.weturntothese next. 5 BothdatabaseshavemorecomprehensivecoverageofventurecapitalfinancingsincleanenergythanThompson VentureEconomicsandDowJonesVentureSource,thatarethetwodatabasestypicallyusedforstudieson venturecapitalbackedstartups. 6 SinceoursamplelooksonlyatpatentsattheUSPTO, foreigninventors arethosewholiveoutsidetheusand havechosentopatenttheirinventionsintheunitedstates.ofcourse,therearelikelytobesignificantnumberof renewableenergyinventionsbyforeigninventorsthatarenotpatentedattheuspto.forexample,anumberof patentsrelatedtosolaringermanyarenotpatentedintheus.however,giventhattheusissuchanimportant market,ourprioristhatimportantpatentswouldinfactbepatentedintheusinadditiontoothercountries. Anecdotalevidencesuggeststhatthisisindeedthecase.Nevertheless,thestructureofoursampledoesnotallow ustomakesubstantiveconclusionsaboutusvs.foreignpatents,orspeaktodifferingtrendsinpatentingbetween USandForeigninventorsinrenewableenergyovertime.

9 Figures3Aand3BreporttheabsoluteandrelativeamountofrenewableenergypatentingattheUSPTO. Theyshowthatrenewableenergypatentsfelloverthe1980s,bothinabsoluteandrelativeterms. Whilethepatentingrateincreasedslightlyinthe1990s,itroseconsiderablyinthe2000s,increasingata disproportionateraterelativetooverallpatentingactivityattheuspto.infact,boththenumberof patentsfiledperyearandtheshareofpatentsfiledintheusptoapproximatelydoubledoverthe10 yearperiodfrom Figure4showsthatthemaindriveroftheincreasewaspatentingbyUSbasedinventors.Infact,there wasasharpbreakinthetrendofpatentingbyusbasedinvestorsrelativetoforeigninventorsaround 2004.Figure5documentsthatVCbackedstartupsincreasedtheirproportionalshareofpatentingby USinventorsthemostoverthisperiod,increasingtheshareofpatentingfromunder5%in2000to about20%ofthepatentsfiledin2009. DespitethesharpincreaseinpatentingbyVCbackedstartups,however,patentinginrenewableenergy stillremainsconcentratedinarelativelysmallnumberoffirms.figure6documentstheshareoftotal patentsfiledbyusinventorsworkingateitherincumbentsorventurecapitalbackedfirmsthatare attributedtothe10,20and50mostactivelypatentingfirmsineachyear.ascanbeseeninfigure6, thetop20firmsaccountedforabouthalfofalltherenewableenergypatentsfiledbyfirmsintheearly late1980sandearly1990s.althoughtheconcentrationhasfallenfromthatpeak,itisstillover40%in Table2providesmoredetailbylistingthemostactiveUSbasedassigneespatentinginrenewable energyinrecentyearsandthenumberofpatentsassociatedwiththese.specifically,itfocusesonthe assigneeswithatleastfivepatentsbetween2005and2009,ineachofthetechnologies.ascanbeseen fromtable2,largeincumbentsaccountforthedisproportionateshareoftheoverallpatenting.firms suchasge,dupont,chevron,exxonmobil,appliedmaterialsareamongthemostactivefirmspatenting inrenewableenergy.however,anumberofvcbackedfirmsarealsoonthislist.forexample, Solopower,KonarkaTechnologies,Stion,Nanosolar,Solyndra,Miasole,TwinCreekTechnologiesand SolariaareallVCbackedfirms,sothat8ofthetop20assigneeswithUSbasedinventorspatentingin solarbetween2005and2009werevcbackedstartups.similarly,amyris,kiorandceresinbiofuels, ClipperWindPowerandFloDesignWindTurbinesinWind,OceanPowerTechnologiesandVerdantin Hydroareallventurecapitalbackedfirms.ThislisthighlightshowVCbackedstartupshavegrownto becomemajorcontributorstoinnovationinrenewableenergyinthelastfewyears. Appendix1providesamoredetailedlistofthetopassigneesfromVCbackedstartups,incumbentsand academia/government,includingbothusandforeigninventorspatentingattheusptoandoverthe period giventhatthelistincludesassigneeswithmanyforeigninventors,otherfamiliar namessuchasvestas,sanyo,sharp,gamesaandschottagarenowalsoamongtheleadingassignees involvedinrenewableenergyinnovation.

10 3.2.CharacteristicsofPatentingbyIncumbentvs.VCbackedFirms Wenextcomparethecharacteristicsofthepatentsfiledbythedifferenttypesoforganizations.Our firststepistoexaminethecitationstothepatentsthattheyfile.sincecitationstendtohaveahighly skeweddistribution,wereporttheresultsfromcountmodels.table3reportstheresultsfromnegative binomialregressions,wherethedependentvariableisthecountofcitationsreceivedforeachpatent. Althoughweincludetechnologyandyearfixedeffectstoaccountforfixeddifferencesinpatenting propensitiesacrosstechnologiesandtoaccountforcohortdifferencesinthenumberofcitations,we neverthelessalsoaccountforthefactthatpatentsin1980wouldhavereceivedmorecitationsthan thosein1995bylookingatthecumulativecitationsreceivedbypatentsfiveyearsfromtheyearof application.ourmeasureofcitationsexcludesselfcitations,soweexaminetheinfluenceofthepatents onotherassignees. Columns13ofTable3reportresultsonbothUSandforeigninventors,whileColumns46restrictthe sampletousbasedinventors.weuseacademicandgovernmentpatentsasourreferencegroupas theyarelikelytohaveremainedthemoststableovertheentireperiod.table3showssomeinteresting patterns.first,asnotedaboveandconsistentwithpriorfindings(singhandfleming,2010),unassigned patentsseemtobefarlessinfluentialthanpatentswithassignees,bothinthefullsampleandforus basedinventors.wheninterpretedasincidencerateratios,column1intable3impliesthatunassigned patentsareassociatedwitha75%lowercitationratethanacademicandgovernmentpatents.second, patentsfiledbyincumbentfirmsareslightlymoreinfluentialthanacademicpatents,butonlymarginally so.theeconomicmagnitudeissmallanditisimpreciselyestimated.incumbentsareassociatedwitha citationratethatis1.1timesthatofuniversityandgovernmentpatents.ontheotherhand,patents filedbyvcbackedfirmsaremuchmorelikelytoreceivesubsequentcitations.theeconomic magnitudesarelarge.thecoefficientsimplythatvcbackedstartupsareassociatedwithacitationrate thatis1.9timesthatofuniversityandgovernmentpatents.inaddition,achi2testforthedifferencein thecoefficientbetweencitationstovcbackedfirmsandincumbentsshowsthatthedifferencesare statisticallysignificant.thepvaluesliewellbelow0.05andasidefromcolumn4,arelessthat0.01in eachcase. Thedifferenceintheoveralllevelofcitationscouldcomefromtwodifferentfronts.First,itispossible thatvcbackedfirmshavefewermarginal,oruncitedpatents,sothatthedifferencestemsfromthe lefttailofthecitationdistributionbeingbetter.second,itispossiblethatvcbackedfirmsaremore likelytohavehighlycitedpatents,sothatevenifthelefttailofthedistributionisnobetter,the intensivemarginofcitationsishigher,includingathickerrighttail.toprobethesepossible explanations,weexamineboththeshareofpatentswithatleastonecitationandtheshareofpatents thatarehighlycited. Table4reportstheresultsfromOLSregressionswherethedependentvariabletakesavalueofoneif thepatentreceivedatleastonecitation.again,unassignedpatentsarefarlesslikelytoreceiveasingle citation.thecoefficientsimplya3538percentagepointlowerchanceofbeingcitedrelativeto academicpatents,onabaselineofa50%citationprobability.bothvcbackedstartupsandincumbents havepatentsthataremorelikelytoreceivecitationsthanpatentsbyinventorsinuniversityand

11 governmentlabs.thisofcourse,couldbeduetothebasicnatureofacademicandgovernmentr&d. WhencomparingVCsandincumbents,however,wefindthatVCshavean1114percentagepoint higherlikelihoodofbeingcitedrelativetoacademiclabs,comparedtoa57percentagepointhigher probabilityforincumbents.thesedifferencesarestatisticallysignificant,suggestingthatonaverage, VCbackedpatentsarelesslikelytobemarginalandmorelikelytoinfluencefutureR&D. Table5reportstheresultsfromOLSregressionswherethedependentvariableisequaltooneifthe patentwashighlycited.specifically,wedefineapatentasbeinghighlycitedifthecitationsforthat patentareinthetop10percentof5yearforwardcitationsforpatentsinthattechnologyandyear. Table5showsthatunassignedpatentsaremuchlesslikelytohaveahighlycitedpatent.Sincethe baselineprobabilityisbydefinitionabout10%,thecoefficientsonunassignedpatentsincolumns1and 4ofTable5pointoutthatthechanceofansuchapatentbeinghighlyinfluentialisessentiallyzero.On theotherhand,vcbackedfirmsarealmosttwiceaslikelyasacademicpatentstobehighlycited. Incumbentfirmshavenostatisticallysignificantdifferenceinhighlycitedpatentsintheoverallsample, andaslightlyhigherchanceamongusbasedinventors.however,importantly,thedifferenceinthe chanceofbeinghighlycitedbetweenvcbackedfirmsandincumbentsisbothstatisticallyand economicallysignificant. Thusfarouranalysishassuggestedthatrenewableenergyinnovationbyincumbentfirmstendstobe lessinfluential.innovationbyincumbentsislesslikelytobecitedatallandwhenitis,itislesslikelyto behighlycited.theseresultsareconsistentwiththeliteraturethathasdocumentedthatincumbent firmshavedifferentgoals,searchprocess,competenciesandopportunitycoststhatleadthemtowards moreincrementalinnovation(tushmanandanderson,1986;hendersonandclark,1990;tripsasand Gavetti,2000;RosenkopfandNerkar,2001;AkcigitandKerr,2011),althoughthesepapershavenot directlycomparedinnovationbyincumbentswiththatbyvcbackedstartups. Toprobeourresultsfurther,weturnnexttodirectlyexaminetheextenttowhichincumbentspursue moreincrementalinnovation,byexaminingthedegreetowhichtheycitetheirownpriorworkrelative toothertypesoforganizations.followingsorensenandstuart(2000),wehypothesizethatiffirmsare citingtheirownpatentsatadisproportionaterate,thentheymaybeengagedinmore exploitation ratherthan exploration (March,1991).Wethereforestudytheextenttowhichinventorsinthe differentorganizationalsettingstendtocitethemselves. Table6reportstheresultsfromnegativebinomialregressionswherethedependentvariableisthe countoftheselfcitationsafocalpatentmakes,whereaselfcitationisdefinedascitingapatentfrom thesameassignee.theregressionscontrolforthetotalnumberofcitationsthepatentmade,and technologyandpatentapplicationyearfixedeffects.ascanbeseenfromtable6,vcbackedfirmsare nomorelikelytocitethemselvesthanacademiclabs.althoughthecoefficientisinfactnegative,itis impreciselyestimated.ontheotherhand,thecoefficientonincumbentfirmsimpliesthattheyare50% morelikelytocitethemselvescomparedtoacademiclabs.again,thedifferencebetweenvcbacked firmsandincumbentsisstatisticallysignificant,suggestingthatpartofthereasonthatincumbentshave lessinfluentialinnovationsisthattheyareengagedinmoreincrementalr&dthanvcbackedstartups.

12 Onepossiblereasonfornotbeingcitedatallandforcitingone sownworkcouldalsobethatfirmsare engagedinextremelynovelinnovationsthathavenotyetyieldedcitations.thiscouldbeparticularly trueinnascenttechnologiessuchasrenewableenergy.inaddition,sincepatentingactivityis concentratedinafewincumbentfirms,itispossiblethatsomeofthehigherselfcitationispurelydue tothefactthatthepriorarttobecitedismorelikelytobethatofincumbentsorthatvcbackedfirms donothavemanypriorpatentstocite. Inordertoaddresstheseconcerns,weuseanewmeasureofnoveltythatisnotbasedoncitation measure.instead,wedrawonatextualanalysisofpatentapplicationstolookatthesimilarityofpatent claimsanddescriptionsforpatentsinagiventechnologyarea.intuitively,ourdefinitionissuchthat patentswithgreatertextualsimilaritytoneighboringpatentsareconsideredtobelessnovel.our measureofnoveltyshouldbeparticularlyusefulinthecontextofsciencebasedpatenting,where technicaltermsaremoreuniqueandthereforemorelikelytosignaldifferencesinthecharacteristicsof innovation,andformorerecenttimeperiods,whereinitialforwardcitationsmaybeanoisypredictorof ultimateoutcomes.themeasurealsoavoidsproblemswithcitationbasedmeasures,wherecitation patternscansufferfromselectionbiases.amoredetaileddescriptionofthemeasureisoutlinedin Appendix2(seealsoYounge,forthcoming,andUllmanandRajaraman2011,pp.9293). AscanbeseenfromTable7,ournoveltymeasureisquiteconsistentwiththeothercitationbased measuresofpatenting.first,ithighlightsthatinadditiontounassignedpatentsnotreceivingmany citations,theyarealsolessnovelthanpatentsbeingdevelopedinacademicandgovernmentlabs.the regressionshighlightthatthenoveltyofthepatentsforvcbackedstartupsisnodifferentfromthatof academiclabs.however,incumbentfirmshaveasignificantlylowerlevelofnovelty.althoughthe differencebetweenthenoveltyofpatentingbyincumbentandvcbackedfirmsisnotsignificantforthe overallsample,itisclosetobeingsignificantatthe10%levelforusbasedinventors,particularlyinthe latterpartofoursample.ourresultsthereforesuggestthatincumbentfirmshavebeenengagedinless novelandexploratoryinnovationthanvcs,inparticularintheus. 4. VentureCapitalFinancingofRenewableEnergyStartups ThusfarwehavedocumentedthatVCbackedstartupshaveincreasedtheirshareofpatentingmost substantiallyoverthepastdecadeandthatthesestartupsseemtobeassociatedwithmoreradicaland novelinnovationthanthatbyincumbentfirms.thetimingofgrowthinrenewableenergypatentingby VCbackedfirmsiscloselyassociatedwithventurecapitaldollarsflowingintorenewableenergy startups.in2002,only43cleanenergystartupsreceivedvcfundingintheus,raisingacombinedtotal of$230m.in2008,over200cleanenergystartupsraised$4.1bninventurecapitalintheus. 7 Infact, cleanenergyinvestmentsaccountedforabout15%ofthetotaldollarsinvestedbyvcsintheusin 2008,ofwhichamajoritywenttorenewableenergytechnologies. 7 Source:ErnstandYoung,NationalVentureCapitalAssociationPressReleases.

13 AlthoughourworkcannotdistinguishwhetherVCsleadstartupstoengageinmoreradicalinnovation, orarejustabletoselectmoreradicalinnovationsthantheincumbentstendtofund,itdoeshighlight thatventurecapitalfinancingseemstohaveassociatedwithamorenovelandhighimpactinnovationin renewableenergy,particularlyinthelate2000s(conti,thursbyandthursby,2012). 8 Thisseems importantgiventheneedforthewidespreadexperimentationrequiredtomakeprogressinproviding lowcost,cleanenergythatwillsupportdevelopingwithoutincurringmassivecostsintermsofclimate change.totheextentthattheshiftinventurecapitalfinanceawayfromsuchtechnologiesisdueto structuralfactors,thissuggeststhatitwillhaveanoticeableimpactonthetypeofinnovationbeing undertakeninrenewableenergy(eitherthroughthetreatmentortheselectioneffectofventurecapital investment). Needlesstosay,anumberoffactorsarelikelyresponsiblefortherapiddeclineinVCfinancingfor renewableenergystartups.theeconomiccollapsein2009hadachillingeffectonallventurecapital investment,includingcleanenergy.inaddition,improvementsinhydrofrackingtechnologythat openeduplargereservesofnaturalgasloweredthecostofnaturalgasconsiderablyandchangedthe economicsofrenewableenergytechnologiesintermsofthembeingcloseto gridparity.nevertheless, ourdiscussionswithventurecapitalinvestorssuggestthatthereareinfactstructuralfactors,overand abovethesehistoricaldevelopments,thathaveledinvestorstobecomeunwillingtoexperimentwith renewableenergyproductiontechnologies.inthissection,weoutlinethesestructuralfactorsthatvcs seemtobefacing,makingsustainedfundingofentrepreneurshipinrenewableenergydifficult. 4.1 FinancingRiskandCapitalintensityofenergyproduction VCbackedinvestmentsareextremelyrisky,leadingtoanextremelyskeweddistributionofreturns.Hall andwoodward(2010)andsahlman(1990;2010)documentthatabout60%ofvcinvestmentsarelikely togobankruptandthevastmajorityofreturnsaretypicallygeneratedfromabout10%ofthe investmentsthatdoextremelywell.furthermore,kerr,nandaandrhodeskropf(2013)documenthow harditisforvcstopredictwhichstartupsarelikelytobeextremelysuccessfulandwhichwillfail.vcs thereforeinvestinstages,ineffectbuyingaseriesofrealoptions,wheretheinformationgainedfrom aninitialinvestmenteitherjustifiesfurtherfinancingortheexerciseofthevc sabandonmentoptionto shutdowntheinvestment(gompers,1995;bergemannandhege,2005;bergemann,hegeandpeng, 2008;Guler,2007).Hencepropertiesofstartupsthatmaximizetheoptionvalueoftheirinvestments maketheirportfoliomorevaluable.forexample,investmentsthatarecapitalefficient(costofbuying theoptionisless),wherestepupsinvaluewhenpositiveinformationisrevealedarelargerelativeto theinvestment(morediscriminating experiments beingrunwiththemoneythatisinvested)and wheretheinformationabouttheviabilityofaprojectisrevealedinashortperiodoftimeareall propertiesthatmakeinvestmentsmoreattractiveforvcs.sectorssuchasitandsoftware,thathave relativelylowlevelsofcapitalinvestment,andwhereinitialuncertaintyabouttheviabilityofthe technologyisrevealedquickly,areidealsectorsforvcs.thehighreturnsforseveralofthemost 8 Notethatsimplylookingatthetimingofthepatentsandtheinvestmentwillnothelpuntanglethecausalityas VCswillofteninvestinfirmsthathavepromisingtechnologiesintheanticipationthattheywillpatent.

14 successfulvcfirmsarebasedoninformationtechnologyinvestments.aclassicexampleisthatof Google,thathadamarketcapitalizationof$23billionatitsIPO5yearsafteritreceiveditsfirstroundof VCfunding,andhavingraisedabout$40Minventurecapitalalongtheway. Ontheotherhand,theuniteconomicsofenergyproductiontechnologiesneedtobedemonstratedat scale,becauseeveniftheyworkinalab,itishardtopredicthowtheywillworkatscale.thefactthat earlystageinvestorsneedtofinancenotjusttheinitialexplorationaroundthetechnology,butalsothe scaleupofthetechnologyleadstotwochallengesforrenewableenergystartups.first,itimpliesthat theresolutionofuncertaintytakesmuchlonger,asstartupsoftenneedtobuilddemonstrationandfirst commercialplantsbeforeitisclearthatthetechnologyistrulyviable.second,sincethese demonstrationplantsfacetechnology(ratherthanengineering)risk,theyaretooriskytobefinanced throughdebtfinance.vcswhobacksuchstartupsthereforeneedtofinancethecompaniesthrough extremelylongandcapitalintensiveinvestments.thefundsrequiredtoprovecommercialviabilityfor energyproductiontechnologiescanreachseveralhundredmilliondollarsovera510yearperiod, comparedtothetensofmillionsthatvcsaretypicallyusedtoinvestinginanygivenstartup. 9 Thislevelofinvestmentisnotfeasiblefromatypicalventurecapitalfundwithoutseverely compromisingthediversificationoftheventurefirm sportfolio.forexample,investingonly$815min aprojectthatistwiceascapitalintensivehalvesthedollarreturnifthestartupissuccessful.onthe otherhand,investingasufficientamounttoretainalargeshareinasuccessfulexitrequiresmakingfar fewerinvestmentsacrosstheportfolioandhencemakestheportfoliomuchmorerisky.such investmentsarethustypicallytoocapitalintensiveforvcs,giventhesizeandstructuresofmostvc fundstoday. Theinabilitytoraiseeitherdebtorventurecapitalatthedemonstrationandfirstcommercialstagehas ledthisstageofthestartup slifetobeknownasthe valleyofdeath (seefigure6).thefactthat investorsarenowacutelyawareofthisfundinggapbeforethefirmgetstocashflowpositiveleadstoan unravelingoftheentirefinancingchain.thatis,sinceinvestorsforecastthatevenpromisingstartups mayhaveahardtimegettingfinancingwhentheyreachthestageofneedingtobuildademonstration plant,thebenefitsofsinkingcapitalinastartupatthestagebeforemaynotbeworthwhile.thislogic, thatvcsrefertoasfinancingrisk,worksthroughbackwardinductiontothefirstinvestor.thus,a forecastoflimitedfuturefundingmayleadpromisingprojectstonotbefunded,evenifwhenfully funded,theywouldbeviableandnpvpositiveinvestments(e.g.,seenandaandrhodeskropf,2013). Thischallengeisexacerbatedbythefactthatpotentialentrepreneurswhoknowtheenergyspaceoften tendtobefromlargeoilcompaniesorfromutilities.hence,theymakegoodceosforthestagewhen thestartupismoreestablished,buttheytendtoberelativelypoorentrepreneursattheearlystageof thebusinesswherecashislimitedandneedstoberaisedfrequently,thebusinessmodelisnotclear, 9 Forexample,Solyndra,acompanythatmanufacturersphotovoltaicsystemsusingthinfilmtechnology,hashad toraise$970minequityfinanceinadditiontoa$535mloanguaranteefromthedepartmentofenergy,priorto itsplannedipoinmid2010.thisamountofcapitaltoprovecommercialviabilityisanorderofmagnitudegreater thanthe$40$50mthatvcsaretypicallyusedtoinvestingineachcompanytogetthemtoasuccessfulexit.

15 anddecisionsneedtobemadequicklywithlimitedinformationathand.ontheotherhand,thosewith abackgroundofvcbackedentrepreneurshipmaybesuccessfulatrunningsmallitrelated,biotechor semiconductorstartups,butareillpositionedtomanageandgrowenergyproductioncompaniesthat havedifferentbusinessmodelsandchallenges.sincevcsrequireentrepreneurstoplayacentralrolein fundraisingforthelargeamountsofmoneyrequiredforcommercialtestingofthetechnology,in additiontomanaginglargeproductionfacilities,internationalcommoditypricingandanticipatingthe changinggovernmentpolicies,thecombinedskillsetrequiredofsuchceosisonethatisinshortsupply (Kaplan,SensoyandStromberg,2009).Thesefactorsgreatlyincreasetheoperationalriskofcompanies, andhencethiscreatesimportantchallengesforrunningandgrowingstandaloneenergyproduction companiesthatgobeyondtheircapitalintensity. 4.2 Exitopportunities VCshaveinvestedinsomeindustriessuchasbiotechnology,semiconductorsandIT/networksthatalso sharetheattributesofhugeinfrastructureandmanagementrequirementsthatareoutsidethescopeof astartup.however,intheseinstances,vcsbankonanestablishedexitmechanismtohandovertheir earlystageinvestmentsbeforetheyhitthevalleysofdeathincapitalandmanagerialtalent.for example,inthebiotechnologyindustry,thevcmodelhasevolvedsothatpharmaceuticalcompanies stepintobuypromisingstartupsatapointevenbeforecommercialviabilityhasbeenproven.thisisa keypartoftheinnovationecosystemasitbridgesthepotentialvalleyofdeathandtherebyfacilitates precommercialvcinvestmentsinbiotechnology.thepropensityofpharmaceuticalcompaniestobuy promisingstartupsalsofacilitatestheiriposatprecommercialstages,becausepublicinvestorsbelieve thereissufficientcompetitionamongpharmaceuticalfirmsforbiotechnologystartupswithinnovative solutionsthattheywillbeacquiredwellbeforetheyhitthevalleyofdeath.cisco,lucent,hpand JunipernetworksplayanequivalentroleintheIT/networkingindustry. Thusfar,however,energyproducingfirmsandutilitiesthatsupplyelectricitytocustomershavebeen farfromactiveinacquiringpromisingcleanenergystartups.thisbottleneckinthescalingupprocess hasaknockoneffectontheabilityforvcstofundprecommercialtechnologiesinthisspaceaswell.if earlystageventureinvestorsfacetheriskthattheymaybeunabletoraisefollowonfundingorto achieveanexit,evenforstartupswithotherwisegood(butasyetunproven)technologies,theyrunthe dangerofsinkingincreasingamountsofdollarsforlongerperiodsoftimetokeepthestartupalive. Withincumbentfirmsunwillingtobuythesestartupsatprecommercialstages,thetimetoexitforthe typicalstartupismuchlongerthanthethreetofiveyearhorizonthatvcstypicallytarget(thetimeto buildpowerplantsandfactoriesisinherentlylongerthanasoftwaresalescycleandcaneventake longerthanthelifeofavcfund).asshowninfigure6,thisleadsventurecapitaliststowithdrawfrom sectorswheretheycouldhavehelpedwiththeprecommercialfunding,butwheretheyarenotcertain thattheywillbeabletoeitherfundtheprojectthroughthefirstcommercialplant,ortheyarenotsure iftheycanexittheirinvestmentatthatstage(nandaandrhodeskropf2013). Inthecaseofbiotechnologyindustry,aclearexitmechanismwasfacilitatedbythefactthattheFDA developedwellunderstoodandtransparentmetricsforsuccessateachstage.becausethesetof buyersisuniformandthecriteriaforasuccessfulexitateachstagehavebeendevelopedandwell

16 understood,vcscanworkbackwardsandsettheirowninvestmentmilestones.inthisway,the downstreamexitprocesshasimportantconsequencesforthedirectionofupstreaminnovation.the factthatthereisawelldevelopedecosystemwherelargepharmaceuticalfirmsbuypromisingstartups impliesthatthereisgreaterearlystageventurecapitalfundingofsuchfirms.moreover,stockmarket investorsalsohaveanappetiteforsuchfirms,intheknowledgethatpharmaceuticalfirmswillbewilling toacquireapromisingtargetbeforeithitsthebottleneckofmarketinganddistribution.thisinturn createsanotherexitavenueforventureinvestors,fuellingfurtherearlystageactivity.infact,the historyofcapitalintensiveindustriessuchasbiotechnology,communicationsnetworkingand semiconductorssuggeststhatuntiltheincumbentsstartbuyingstartups,theinnovationpipelinedoes nottrulytakeoff. Theextenttowhichlargeenergycompanieswillplayanequivalentroleintheinnovationpipelinefor cleanenergyisnotyetclear.whileenergycompanieshavenotchosentobeactivebuyersofclean technologystartups,itisstillearlyinthelifeofthecleanenergyecosystem,comparabletothe biotechnologyecosysteminthelate1980s.therearesomesignsthatthismaybechanging,withthe mostpromisingdevelopmentsbeingintheriseofanumberofcorporateventurecapitalfundsamong thelargeenergycompanies(nandaandrothenberg,2011). 4.3 GlobalCommoditiesandPolicyRisk AfinalimportantdifferencebetweentherenewableenergyproductionandthetypicalVCbacked startupisthatenergyisacommodity.successinenergycomesfrombeingalowcostproviderrather thanhavinganinnovationthatcanbepricedhighduetothewillingnessofenduserstopay(asisthe caseforbiotechnology).whileincumbentsinotherindustriescompetewitheachothertoacquire startupsinordertomeetenduserdemand,theenduserintheenergymarketcannotdistinguish electronsproducedfromcoal,thesunorthewind(unlessthegovernmentpricesthecostofcarbon appropriately).intheabsenceofappropriatepricesignalsorincentivestoinvestinrenewables, incumbentsarethereforenotpressedtoacquirestartupsinthisspace.inthecaseofbiofuels,the inputstotheirproductionprocessarealsocommodities.energyproducersthereforefacecommodity riskforbothrawmaterialsandendproducts.sincethesemarketscanexhibitsubstantialpricevolatility, itmakesrunningandmanagingthesecompaniesmoredifficult.forexample,secondandthird generationbiofuelstartupsproducingethanolorbiocrudeat$80$90perbarrelwerecompetitivein 2007priortotheglobalrecessionwhenconventionaloilpricestopped$100abarrel,butmostwent bustwhenoilpricesplummetedinthesubsequentrecession. Thechallengesofbackingaglobalcommodityproducerarecompoundedbythefactthatenergyand cleanenergyaresectorswithlargeinvolvementbygovernmentsacrosstheworld.giventhatclean energytechnologieshavenotyetachievedgridparity,governmentpolicyisalsocriticalindetermining thepricesofinputsandfinishedproducts.somegovernmentschoosetoeithertaxcarboncontentin conventionalfuelsortobuycleanenergyatapremium.otherschoosetosubsidizecleanenergy companiesthroughdirectgrantsandsubsidiesorthroughtaxbreaks.regardlessofthepolicy,itimplies thattheextenttowhichagivenstartup sproductislikelytobeprofitabledependsgreatlyonwhetherit

17 isincludedinthesubsidyorcredit,theextenttowhichcarbonistaxedorthepricepremiumatwhich thegovernmentbuysthecommodity. Policychangesanduncertaintyarethusmajorfactorshinderingthepotentialinvestmentbyprivate sectorplayersacrossthecleanenergyinvestmentlandscape(bloom,2009).thisisparticularlytrue whentheperiodicityoftheregulatorycycleissmallerthantheinvestmentcyclerequiredfor demonstratingcommercialviability.insuchanevent,nooneiswillingtoinvestinthefirstcommercial plantiftheydonotknowwhattheregulatoryenvironmentisgoingtobebythetimesuccesshasbeen demonstrated(basedontherulesofthepriorregulatoryregime). 5. Possiblesolutions ThesectionabovehastriedtodemonstratethatVCsentermarketswheretheybelievetheycancreate andthenexercisegrowthoptions.vc swereinitiallyattractedtorenewableenergytechnologiesinthe beliefthattherewouldbeashiftindemandfromsocietyforrenewableenergy,andthatshiftwould createlargegrowthoptionsforthemduetoaconstrainedsupplyofsociallydesiredrenewableenergy alternatives.thisbeliefhasnotyetpannedout,andmoreover,thecostofcreatingthegrowthoptions haveturnedouttobemuchgreater,astheinnovationpipelineisnascentandtheabsenceof incumbentsbuyingpromisingtechnologiesmeansvcinvestorsneedtofinancethescaleupbeforethey canexercisetheiroptions.inthissection,wediscusspossiblesolutionsthatmayhelpdriveamore robustfinancingecosystemforrenewableenergytechnologies. 5.1Privatecapitalmarketsolutions Withtheappropriatepricesignalsandincentives,incumbentsmaybeincentivizedtobuypromising technologies,therebyreducingfinancingriskconsiderably,astookplaceasthebiotechnologyindustry matured.wediscusstheseincentivesinthenextsubsection. Intheabsenceofenergycompaniesandutilitiesplayinganimportantroleinthisinnovationpipeline, thealternativeforwidespreadinnovationbystartupsinrenewableenergyproductionwillrequirethe structureofthevcindustrytochangeinkeyways. First,itwillrequiresignificantlylargerfundsthanistypicalforventurecapitalinvestors,inorderto addressthechallengesassociatedwithfinancingriskandthe valleyofdeath.indeedkleinerperkins $500MGreenGrowthFund,andKhoslaVentures $750Mfundareexamplesofsuchatrend 10,butthe sectorwillprobablyrequireevenlargerfundstosupportthescaleuprequiredbyenergyproduction companies.themajorityofventurecapitalinvestorsincleantechnologydonothavededicatedfunds forthissector,andcontinuetoraise$250300mfundsandmayneedtohavefargreaterlevelsof syndication,orpresetpartnershipsacrossvcsinordertosustainthelevelofinvestmentrequiredby 10 KhoslaVenturesalsocreatedadualfundstructure,whereasmallerfundwouldfocusonearlystage experimentswhileasecond,largerfundwouldguaranteefundingforcommercialization,helpingtoreduce financingrisk.

18 thissector.moreover,ifventurecapitalinvestmentintheenergysectoristobesustainedinthe absenceofearlyexitopportunities,itwillrequirearadicalreworkingofthevcfundstructuresand terms. Overcomingthesechallengeswillbecompoundedbyanotherfactorassociatedwiththeemergenceofa newindustry:learningthroughexperimentation.investorsinnewtechnologiesgetfeedbackontheir processofduediligence,thetypesofentrepreneurswhoaremostsuccessful,andanunderstandingof thechallengesfacedbycertaintypesofbusinessmodelsovertheinvestmentcycle(goldfarbetal 2007).Thisisalsoaperiodwhenagenerationofnewentrepreneursarises,driveninpartbythemany firmsthatfailduetoatechnologythatdidnotwork,butwheretheentrepreneursdevelopedagood workingrelationshipwiththeventurecapitalinvestors.alloftheseprocessesdevelopfasterwhenthe cycletimesfor experimentation bythevcsaresmaller.inthecontextofcleanenergy,thefeedbackis muchslower,drivenbythedualstagesofriskandthelongercycletimesofcleanenergy.manyplayers thereforeseeacriticalroleforgovernmentinsupportingthegrowthofthecleanenergyinnovation pipeline. 5.2GovernmentSupport TheUSgovernmenthasplayedanimportantroleinsupportingcleantechnologyinnovationintheUS. However,thevastmajorityofthishasbeenonthe supplyside,throughthedirectsupportofspecific governmentanduniversityprograms,grantstosupportprecommercialfundingofnewstartups throughthearpaeprogramandattemptstobridgethevalleyofdeathforindividualprojectsthrough thedepartmentofenergy sloanguaranteeprogram(roberts,lassiterandnanda2010). Whileclearlyveryhelpfulinattemptingtoaddressthefundinggapsinherentintheenergyinnovation pipeline,akeyaspectofensuringthatthepipelineofnewprojectscontinuetogetfundingfromthe privatesectorwillbetoensurethatthereisavibrantsetofexitopportunitiesforthesestartupsbefore theyhitthevalleyofdeath.whilegovernmentguaranteeddebtwillhelpreducesomeofthisrisk, widespreadexperimentationanddeploymentofnewtechnologiescanonlytakeplaceoncestartups haveaclearpathtobeingacquiredorgoingpubliconthecapitalmarkets.thegovernmentcan thereforedomoreintermsofmakingexitseasier.wenotethreeinterestingideasthathaveemerged throughourdiscussionswithvcsandthatarealsoechoedinthebroadercommunityofinvestors lookingforsolutionstothe valleyofdeath (BloombergNewEnergyFinance,2010). Thefirstareawherethegovernmentcanmakeasignificantcontributionisthroughstable,predictable andlongtermpolicymeasuresaimedatstimulatingdemandforcleanenergy.removinguncertainty aroundpoliciesreducespolicyriskdramaticallyandmakesiteasierfortheprivatecapitalmarketsto plantheirinvestmentsaccordingly.furthermore,intheabsenceofenduserpressuretodrivem&a activity,thegovernmentcancreatethispressurethroughpoliciessuchasfeedintariffs(fits).while FITshavetheirmostdirecteffectonincrementalimprovementsofcommerciallyproventechnologies, solutionssuchasemergingtechnologyauctionsmaybeabletosuccessfullycreatetheappropriate demandfornewtechnologies.

19 Second,thegovernmentcandirectlystimulateM&Aactivityeitherthroughtheregulatorysystemor throughcorporateincentives.forexample,withoutdecoupling,utilitieswillhavenoincentivetoadopt newtechnologiesbeyondanythingthattheyaremandatedtodo.renewableportfoliostandardsas theystandtodaytendtobiasutilitiestowardsadoptingmoremature,currentlycheapertechnologies. Thegovernmentcanalsocreateincentivesforincumbentfirmstoactasfirstadoptersfornew technologies.thesecaneffectivelyhelptobridgethe valleyofdeath,createmoreearlystagefunding anddrivethegrowthofasufficientnumberofstartupfirmstoultimatelycreatelargefirmsthatwill competewitheachothertoacquireforthenextgenerationofstartups. Finally,thegovernmentcancreatepublicprivatepartnershipfundsthatcanhelpeitherwithfirst commercialtestingorasmechanismsthateffectivelycompetewiththeincumbents.creatingthis competitioncanhelpstimulatem&aactivityinthesectorandhencedrivetheinnovationpipeline (BloombergNewEnergyFinance,2010).ThiscouldbedesignedalongthelinesofFernandez,Steinand Lo(2012)whohavebeguntoidentifysimilarchallengesinbiotechnologyandaresuggestinginnovative financialengineeringapproachestofundinginnovationincancertherapy,thatcouldalsobeappliedto renewableenergytechnologies. 6. Conclusion Innovationinrenewableenergyhasgrowninrecentyears,inpartduetothesharpriseinventure capitalfinanceforrenewableenergystartupsintheearly2000s.however,theavailabilityofventure capitalfinanceforrenewableenergyhasfallendramaticallyinrecentyears.inthischapter,weask whetherweshouldworrythatthedeclineandshiftinvcwillslowtherateandalterthedirectionof innovationinrenewableenergy. Ourresultssuggestthatstartupsbackedbyventurecapitalfilepatentsthataremorelikelytohaveat leastonecitation,aremorelikelytobehighlycited,havefewerselfcitationsandaremorelikelytobe novelthanpatentsfiledbyincumbentfirms.althoughthelaginthepatentgrantsdonotallowusto directlyobservehowthefallinglevelsofvcfinancerelatetotheinnovationsbyvcbackedstartups,our resultssuggestvcfinancingisassociatedwithagreaterdegreeofeconomicexperimentationand therefore,weshouldworrythatvcshavemovedawayfromfundingsuchstartups. Ourchapterhasalsoaimedtoshedlightonsomeofthestructuralfactorsthathavemadesustained experimentationbyvcshard,withaparticularemphasisonthedifficultyofexitingtheirinvestmentsto incumbentfirmsthathavetheexpertiseandcapitaltofinancethescaleupofsuchtechnologies. Larger/longerFundsmaybeonealternative,sincesuchfundscouldgetpasttheuncertaintyandtoa stableplaceforexit),butmeasuresdesignedtoincentivizeincumbentsupthefinancingchaintobridge the valleyofdeath alsoseemlikepossiblesolutions. Weshouldnotethatouranalysisisnotmeanttosuggestthatprocessinnovationsareunimportant,or thatthefocusofvconotheraspectsofcleantechisnotvaluable.rather,ourobjectiveistohighlight thefactthattheshiftingfocusofventurecapitalislikelytohaveanimpactonboththerateandthe characteristicsofrenewableenergyinnovationinthecomingyears.totheextentthatthereisstilla

20 needforexperimentationwithnewtechnologiesadesiretocommercializeradicalinnovationsin renewableenergy,ourworkhighlightsthatinadditiontofallingnaturalgasprices,therearestructural factorsaboutenergythatmakesustainedexperimentationbyvcsdifficultinrenewableenergy. Althoughitisstillearlyinthelifecycleofthisindustry,ourdiscussionhasoutlinedsomespecificfactors thatmayfacilitatethedeploymentoflargeamountsofriskcapitalthatarenecessarytofinance renewableenergyinnovations. References Akcigit, Ufuk, and William Kerr, 2011, Growth through heterogeneous innovations. Harvard Business SchoolWorkingpaper. Balasubramanian,NatarajanandJeongsikLee,2008,Firmageandinnovation.IndustrialandCorporate Change,17: Bergemann, Dirk, and Ulrich Hege The Financing of Innovation: Learning and Stopping." RAND JournalofEconomics,36: Bergemann, Dirk, Ulrich Hege, and Liang Peng Venture Capital and Sequential Investments." WorkingPaper. Bloom,Nick.2009.TheImpactofUncertiantyShocks.Econometrica,77: BloombergNewEnergyFinance.2010.CrossingtheValleyofDeath;Solutionstothenextgeneration cleanenergyprojectfinancinggap.june2010whitepaper BP Statistical Review of World Energy. Review2012/statistical_review_of_world_energy_2012.pdf Conti, Annamaria, Jerry Thursby and Marie Thursby Are Patents Endogenous or Exogenous to StartupFinancing?WorkingPaper Fan,RongEnandKaiWeiChang,ChoJuiHsieh,XiangRuiWang,ChihJenLin:LIBLINEAR:ALibraryfor LargeLinearClassification.JournalofMachineLearningResearch9: (2008). Fernandez, JoseMaria, Roger M Stein and Andrew W Lo, 2012, Commercializing biomedical research throughsecuritizationtechniques.naturebiotechnology30, Gans,Joshua,DavidHsu,andScottStern,2002,Whendoesstartupinnovationspurthegaleofcreative destruction?randjournalofeconomics33, Goldfarb,B.,D.KirschandD.Miller,2007, WastheretooLittleEntryintheDotComEra? Journalof FinancialEconomics86(1): Gompers,Paul.1995.OptimalInvestment,Monitoring,andtheStagingofVentureCapital."Journalof Finance,50: Gompers,Paul,andJoshLerner,2002,TheVentureCapitalCycle,MITPress,Cambridge,MA. Guler,Isin,2007,ThrowingGoodMoneyAfterBad?AMultiLevelStudyofSequentialDecisionMaking intheventurecapitalindustry.administrativesciencequarterly,52: Hall,Robert,andSusanWoodward,2010,Theburdenofthenondiversifiableriskofentrepreneurship. AmericanEconomicReview100:3, Hellman,ThomasandManjuPuri.2002.VentureCapitalandtheProfessionalizationofStartupFirms: EmpiricalEvidence,JournalofFinance57(1): Henderson,RebeccaandKimClark.1990.ArchitecturalInnovationtheReconfigurationofExisting ProductTechnologiesandtheFailureofEstablishedFirms.AdministrativeScienceQuarterly 35(1):930.

21 Kaplan,Steve,BerkSensoy,andPerStrömberg,2009,Shouldinvestorsbetonthejockeyorthehorse? Evidencefromtheevolutionoffirmsfromearlybusinessplanstopubliccompanies.Journalof Finance64, Kerr,William,RamanaNandaandMatthewRhodesKropf.2013.EntrepreneurshipasExperimentation. WorkingPaper. Kortum, Samuel, and Josh Lerner, 2000, Assessing the contribution of venture capital to innovation. RANDJournalofEconomics31:4, March,James,1991,Explorationandexploitationinorganizationallearning.OrganizationalScience2, Nanda, Ramana, and Matthew RhodesKropf, 2012, Financing Risk and Innovation, Harvard Business SchoolWorkingPaper,No Nanda,Ramana,andMatthewRhodesKropf,2013,Investmentcyclesandstartupinnovation.Journalof FinancialEconomics,forthcoming. Roberts,Michael,JosephLassiterandRamanaNanda.2010.U.S.DepartmentofEnergy&RecoveryAct Funding:Bridgingthe"ValleyofDeath".HarvardBusinessSchoolCase Rosenberg, Nathan, 1994, Economic experiments. In Inside the Black Box, Cambridge: Cambridge UniversityPress. Rosenkopf, Lori and Atul Nerker, 2001, Beyond Local Search: BoundarySpanning, Exploration, and ImpactintheOpticalDiskIndustry,StrategicManagementJournal,22: Sahlman,W.,1990.Thestructureandgovernanceofventurecapitalorganizations.JournalofFinancial Economics27, Sahlman,W.,2010.Riskandrewardinventurecapital.HarvardBusinessSchoolNote Sahlman, William A., Ramana Nanda, Joseph B. Lassiter III, and James McQuade TerraPower: InnovationinNuclearEnergy.HarvardBusinessSchoolCase Samila, Sampsa, and Olav Sorenson, 2011, Venture capital, entrepreneurship and economic growth. ReviewofEconomicsandStatistics93, Singh, Jasjit and Lee Fleming, 2010, Lone Inventors as Sources of Breakthroughs: Myth or Reality? ManagementScience56:1,4156. Stern, Scott, 2005, Economic experiments: the role of entrepreneurship in economic prosperity. In: Understanding Entrepreneurship: A Research and Policy Report, Ewing Marion Kauffman Foundation. Sorensen, Jesper and Toby E. Stuart Aging, Obsolescence and Organizational Innovation. AdministrativeScienceQuarterly45: Tripsas, Mary and Giovanni Gavetti Capabilities, Cognition and Intertia: Evidence from Digital Imaging.StrategicManagementJournal,21: Tushman, Michael L., and Philip Anderson Technological Discontinuities and Organizational Environments.AdministrativeScienceQuarterly31(3): Ullman,JeffandAnandRajaraman2011.MiningofMassiveDatasets.CambridgeUniversityPress,NY, NY. Younge,Ken.2013(forthcoming).AMeasureofPatentNovelty,FungInstituteTechnicalReport.

22 Figure1:SourceofGlobalEnergyConsumedin2010(Total=132,000TWh) Figure2A:SeriesAFinancingforUSbasedstartupsinSolar,WindandBiofuels 350 SeriesAFinancing,bySector(US,only$MM) Biofuels Solar Wind

23 Figure2B:Industrial/EnergyShareoftotalVCinvestments(firstseriesfinancingsonly) 5.0% 4.5% 4.0% 3.5% 3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% Industrial/Energyshareofdeals Industrial/Energyshareoftotaldollars Figure2C:ShareofSeriesACleantechfinancingsbyVCsgoingtosolar,windandbiofuels startups 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0%

24 Number of patents Figure 3A: Renewable Energy Patents Share of all USPTO patents Figure 3B: Renewable Energy Share of All USPTO Patents

25 Number of Patents Figure 4: Renewable Energy Patents by US or Foreign US Patents Foreign Patents Figure 5: US Renewable Energy Patents by Organizational Type Stacked Shares of Total VC-backed Univ/Govt Incumbents

26 Figure 6: Concentration of Renewable Energy Patenting, US Inventors Share of Total Top 50 Firms Top 20 Firms Top 10 Firms Figure7:FundingGapsandthe ValleyofDeath Idea Vetting & Pre- Commercial Testing Establishing Commercial Viability Large-Scale Deployment Inability to bridge Valley of Death leads venture capital to dry up even in the precommercial stage Valley of Death Project / Asset Finance Established Companies Public Equity Markets

27 Table1 Patentingratesinrenewableenergy,bytechnologyandorganizationtype Thistablereportsthebreakdownof22,869renewableenergypatentsattheUSPTOthatweregrantedbetween1980 and2009.panelaprovidesabreakdownfortheentiresampleandpanelbprovidesabreakdownforusbased inventors.venturebackedstartupsrefertopatentswheretheassigneewasmatchedtoafirmthatreceivedventure Capitalfinance(identifiedusingdatafromCleantechi3andBloombergNewEnergyFinance).Patentsgrantedto academicinstitutionsorgovernmentlabswereidentifiedusingatextmatchingalgorithmfollowedbymanualreview. IncumbentfirmsrefertotheresidualcategoryofassigneeswhowerenotclassifiedaseitherVCbackedorfrom academia/government.unassignedpatentsarethosenotaffiliatedwithanyorganizationandaretypicallyseenas independentinventors. PANELA:ALLRENEWABLEENERGYPATENTSATUSPTO( ) Venturebacked Startups Incumbent Firms Academiaand Government Unassigned Total Percent Solar 473 5, ,502 9,644 42% Wind 169 1, ,129 3,047 13% Biofuels 177 4, ,834 30% Hydroelectric 78 1, ,058 2,375 10% Geothermal % Total ,340 1,847 5,733 22, % PANELB:USBASEDINVENTORSONLY Venturebacked Startups Incumbent Firms Academiaand Government Unassigned Total Percent Solar 402 2, ,884 5,565 41% Wind ,492 11% Biofuels 143 2, ,302 32% Hydroelectric ,509 11% Geothermal % Total 686 7,547 1,290 4,066 13, %

28 Table2 AssigneeswithfiveormorepatentstoUSbasedinventors Rank Solar Patent Count Bio Patent Count Wind Patent Count Hydro Patent Count Geothermal 1 AppliedMaterials,Inc 56 StineSeedCompany 212 GE 204 GE 16 KelixHeatTransferSystems 7 2 SunPowerCorporation 35 DuPont 78 GenedicsCleanEnergy,LLC 9 OceanPowerTechnologies 12 EarthToAirSystems,LLC 5 3 Solopower 33 MerschmanSeeds 76 ClipperWindpowerTechnology 8 AlticorCorporateEnterprises 11 GE 5 4 GE 31 NovoGroup 30 NorthernPowerSystems,Inc. 6 LockheedMartin 6 5 Boeing 25 UOP 30 FloDesignWindTurbineCorp. 5 VerdantPower 6 6 KonarkaTechnologies 24 Chevron 29 FrontierWind,LLC 5 7 StionCorporation 24 Monsanto 29 RenScienceIPHoldingsInc. 5 8 IBM 20 SyngentaAG 25 9 DuPont 18 M.S.TechnologiesLLC EmcoreSolarPower,Inc. 17 Dow Nanosolar,Inc. 17 MichiganStateUniversity GuardianIndustriesCorp. 16 ADM Xerox 13 UniversityofIllinios SolyndraLLC 12 UniversityofWisconsin TEConnectivity 12 Amyris TwinCreeksTechnologies 12 RoyalDSM LockheedMartin 11 NewMarketCorporation Miasole 11 U.S.DepartmentofAgriculture SolariaCorporation 9 GE UTC 9 UniversityofCalifornia EnergyInnovations,Inc. 8 BattelleMemorialInstitute 9 22 GM 8 Ceres,Inc Solexel,Inc. 8 UniversityofSouthernCaliforn 7 24 Solfocus,Inc. 8 BASF 6 25 UniversityofCalifornia 8 ConocoPhillipsCompany 6 26 WorldFactory,Inc. 8 ExxonMobil 6 27 Foxconn 7 ZeaChemInc IostarCorporation 7 Agritope/Aventis 5 29 NorthCarolinaStateUniver 7 Battelle 5 30 UniversityofCentralFlorida 7 CatalyticDistillationTechnolog 5 31 XantrexTechnologyInc. 7 Coskata,Inc Chevron 6 LUCATechnologies,Inc GenedicsCleanEnergy,LLC 6 LenLoChem,Inc QualcommInc. 6 PennsylvaniaStateUniversity 5 35 NASA 6 RiceUniversity 5 36 UnitedSolarOvonicLLC 6 SyntroleumCorporation 5 37 VarianSemiconductor 6 Xyleco,Inc AMPT,LLC 5 39 AdvancedEnergyIndustries 5 40 AlcatelLucent 5 41 AppleInc ArchitecturalGlassandAlum 5 43 Batelle/MRIGlobal 5 44 Plextronics,Inc PrimestarSolarInc. 5 Patent Count

29 Table3 Citationstopatents ThistablereportstheresultsfromNegativeBinomialregressionswherethedependentvariableisthecountofcumulativecitationsreceivedbyeachpatent fiveyearsfromthepatent'sapplication.column(1)reportsresultsforthefullsample.columns(2)and(3)excludepatentswithnoassignees.column(3) looksonlyatthelatterhalfofthethirtyyearperiod,from columns(4),(5)and(6)aretheequivalenttocolumns(1),(2)and(3)respectively butforthesubsampleofusbasedinventorsonly.allregressionsincludefixedeffectsforthepatent'sgrantyearaswellasatechnologyfixedeffect(for Solar,Wind,Biofuels,HydroandGeothermal).Parenthesesreportrobuststandarderrors,clusteredbyassignee.***p<0.01,**p<0.05,*p<0.1 FullSample USbasedInventorsonly (1) (2) (3) (4) (5) (6) (a)venturecapitalbackedstartup 0.606*** 0.639*** 0.672*** 0.601*** 0.641*** 0.670*** (0.165) (0.158) (0.177) (0.181) (0.173) (0.199) (b)incumbentfirms * 0.164* (0.073) (0.075) (0.108) (0.087) (0.089) (0.132) (c)unassigned 1.365*** 1.472*** (0.176) (0.175) PvalueonChi2testfordifferencebetween(a)and(b) 0.002*** <0.001*** <0.001*** 0.011** 0.003*** 0.002*** Patentapplicationyearfixedeffects Y Y Y Y Y Y Technologyfixedeffects Y Y Y Y Y Y Observations 22,869 17,136 11,611 13,589 9,523 6,155

30 Table4 Shareofpatentswithatleastonecitation ThistablereportstheresultsfromOLSregressionswherethedependentvariabletakesavalueofoneifthepatentreceivedatleastonecitationandzero otherwise.resultsarerobusttorunninglogitregressions.column(1)reportsresultsforthefullsample.columns(2)and(3)excludepatentswithno assignees.column(3)looksonlyatthelatterhalfofthethirtyyearperiod,from columns(4),(5)and(6)aretheequivalenttocolumns(1),(2) and(3)respectivelybutforthesubsampleofusbasedinventorsonly.allregressionsincludefixedeffectsforthepatent'sgrantyearaswellasatechnology fixedeffect(forsolar,wind,biofuels,hydroandgeothermal).parenthesesreportrobuststandarderrors,clusteredbyassignee.***p<0.01,**p<0.05,* p<0.1 FullSample USbasedInventorsonly (1) (2) (3) (4) (5) (6) (a)venturecapitalbackedstartup 0.105*** 0.130*** 0.136*** 0.077* 0.114*** 0.123*** (0.034) (0.027) (0.029) (0.040) (0.030) (0.032) (b)incumbentfirms 0.055*** 0.053*** 0.058*** 0.061*** 0.063*** 0.070*** (0.016) (0.014) (0.017) (0.019) (0.018) (0.022) (c)unassigned 0.353*** 0.381*** (0.018) (0.020) PvalueonWaldtestfordifferencebetween(a)and(b) *** 0.003*** * 0.058* Patentapplicationyearfixedeffects Y Y Y Y Y Y Technologyfixedeffects Y Y Y Y Y Y Observations 22,869 17,136 11,611 13,589 9,523 6,155

31 Table5 Shareofpatentsthatarehighlycited ThistablereportstheresultsfromOLSregressionswherethedependentvariabletakesthevalueofoneifthepatentwasabovethe90thpercentileinterms ofcitationsreceived,andzerootherwise.resultsarerobusttorunninglogitregressions.percentilesarecalculatedrelativetocitationsreceivedbyother patentsinthesametechnologyandapplicationyearandarebasedoncumulativecitationsreceivedbyeachpatentfiveyearsfromthepatent'sgrant. Column(1)reportsresultsforthefullsample.Columns(2)and(3)excludepatentswithnoassignees.Column(3)looksonlyatthelatterhalfofthethirty yearperiod,from columns(4),(5)and(6)aretheequivalenttocolumns(1),(2)and(3)respectivelybutforthesubsampleofusbased inventorsonly.sincepercentilesarecalculatedwithatechnologyyearcell,allregressionsimplicitlyincludefixedeffectsforthepatent'sapplicationyearas wellasatechnologyfixedeffect(forsolar,wind,biofuels,hydroandgeothermal).parenthesesreportrobuststandarderrors,clusteredbyassignee.*** p<0.01,**p<0.05,*p<0.1 FullSample USbasedInventorsonly (1) (2) (3) (4) (5) (6) (a)venturecapitalbackedstartup 0.075*** 0.075*** 0.083*** 0.086*** 0.086*** 0.101*** (0.026) (0.026) (0.027) (0.028) (0.028) (0.028) (b)incumbentfirms ** 0.037** 0.031* (0.013) (0.013) (0.014) (0.016) (0.016) (0.016) (c)unassigned 0.125*** 0.114*** (0.015) (0.015) PvalueonWaldtestfordifferencebetween(a)and(b) 0.018** 0.018** 0.008*** 0.054* 0.054* 0.009*** Patentapplicationyearfixedeffects Y Y Y Y Y Y Technologyfixedeffects Y Y Y Y Y Y Observations 22,869 17,136 11,611 13,589 9,523 6,155

32 Table6 Degreeofselfcitation ThistablereportstheresultsfromNegativebinomialregressionswherethedependentvariableisthenumberofnumberofbackwardcitationsthatareself citations.allregressionscontrolforthetotalnumberofbackwardcitations,sothecoefficientsreflecttheshareofpriorartbeingcitedthatisselfcitation. Theycanthereforebeinterpretedasthedegreetowhichtheassigneeisengagedinincrementalorexploitativeinnovation.Column(1)reportsresultsfor thefullsample.columns(2)and(3)excludepatentswithnoassignees.column(3)looksonlyatthelatterhalfofthethirtyyearperiod,from Columns(4),(5)and(6)aretheequivalenttoColumns(1),(2)and(3)respectivelybutforthesubsampleofUSbasedinventorsonly.Allregressionsinclude fixedeffectsforthepatent'sapplicationyearaswellasatechnologyfixedeffect(forsolar,wind,biofuels,hydroandgeothermal).parenthesesreport robuststandarderrors,clusteredbyassignee.***p<0.01,**p<0.05,*p<0.1 FullSample USbasedInventorsonly (1) (2) (3) (4) (5) (6) (a)venturecapitalbackedstartup (0.266) (0.266) (0.281) (0.296) (0.296) (0.308) (b)incumbentfirms 0.405** 0.405** 0.466** 0.379** 0.380** (0.188) (0.188) (0.232) (0.181) (0.181) (0.219) (c)unassigned 5.259*** 5.283*** (1.045) (1.082) PvalueonChi2testfordifferencebetween(a)and(b) 0.007*** 0.007*** 0.013** 0.056* 0.057* 0.087* Patentapplicationyearfixedeffects Y Y Y Y Y Y Technologyfixedeffects Y Y Y Y Y Y Observations 22,869 17,136 11,611 13,589 9,523 6,155

33 Table7 Characteristicsofinnovation:patentnovelty ThistablereportstheresultsfromOLSregressionswherethedependentvariableisthenoveltyofthepatentclaims,calculatedasdescribedinthe appendix.column(1)reportsresultsforthefullsample.columns(2)and(3)excludepatentswithnoassignees.column(3)looksonlyatthelatterhalfof thethirtyyearperiod,from columns(4),(5)and(6)aretheequivalenttocolumns(1),(2)and(3)respectivelybutforthesubsampleofus basedinventorsonly.sincethenoveltymeasureiscalculatedforeachfocalpatentatagivenpointintime,relativetopatentsfromthethreepreviousyears inthesametechnologyarea,allregressionsimplicitlyincludefixedeffectsfortimeandtechnology.parenthesesreportrobuststandarderrors,clusteredby assignee.***p<0.01,**p<0.05,*p<0.1 FullSample USbasedInventorsonly (1) (2) (3) (4) (5) (6) (a)venturecapitalbackedstartup * (0.014) (0.015) (0.012) (0.018) (0.022) (0.017) (b)incumbentfirms 0.039*** 0.039*** 0.052*** 0.050*** 0.049*** 0.068** (0.012) (0.012) (0.017) (0.019) (0.018) (0.029) (c)unassigned 0.009** (0.004) (0.005) PvalueonWaldtestfordifferencebetween(a)and(b) * Patentapplicationyearfixedeffects Y Y Y Y Y Y Technologyfixedeffects Y Y Y Y Y Y Observations 22,869 17,136 11,611 13,589 9,523 6,155

34 Appendix Assigneeswiththemostpatentsbetween ,includingUSandForeigninventors Rank Incumbents Patent Count VCbackedfirms Patent Count AcademiaandGovernment 1 GE 541 KonarkaTechnologies,Inc. 52 IndustrialTechnologyResearchInstitute 38 2 DuPont 235 Solopower,Inc. 33 UniversityofCalifornia 36 3 StineSeedCompany 213 LMGlasfiberA/S 30 UniversityofWisconsin 27 4 Vestas 136 RepowerSystemsAG 30 MichiganStateUniversity 22 5 Canon 131 Nanosolar,Inc. 29 U.S.Navy 21 6 Siemens 92 StionCorporation 24 U.S.DepartmentofAgriculture 19 7 Boeing 88 ClipperWindpowerTechnology,Inc. 18 NorthCarolinaStateUniversity 18 8 MitsubishiGroup 88 PowerLightCorporation 14 FraunhoferSociety 17 9 NovoGroup 83 Miasole 13 UniversityofIllinios MerschmanSeeds 81 SolyndraLLC 12 InstitutFrancaisduPetrole Sanyo 79 TwinCreeksTechnologies,Inc. 12 NASA BASF 76 XantrexTechnologyInc. 11 UniversityofCentralFlorida Sharp 68 EnergyInnovations,Inc. 10 UniversityofFlorida ExxonMobil 65 HansenTransmissionsInternationalNV 9 PrincetonUniversity DKBGroup 64 OceanPowerTechnologies,Inc 9 MassachusettsInstituteofTechnology(MIT) AppliedMaterials,Inc 61 SolariaCorporation 9 NationalInstituteofAdvancedIndustrialScienceandTechnology Chevron 60 Ceres,Inc. 8 BattelleMemorialInstitute Monsanto 56 Nanosys,Inc 8 U.S.Army Samsung 52 Solexel,Inc. 8 IowaStateUniversity 9 20 NordexEnergyGmbH 48 Solfocus,Inc. 8 CaliforniaInstituteofTechnology 8 21 SunPowerCorporation 45 ZeaChemInc. 8 KoreaAdvancedInstituteofScienceandTechnology 8 22 KanekaCorporation 44 MetabolixInc. 7 PennsylvaniaStateUniversity 8 23 LockheedMartin 42 OryxeEnergyInternational,Inc. 7 UniversityofSouthernCalifornia 8 24 Dow 39 SunPowerCorporation,Systems 7 NationalResearchCouncilofCanada 7 25 Sumitomo 39 ConverteamLtd 6 NationalTaiwanUniversity 7 26 Gamesa 38 EnlinkGeoenergyServices,Inc. 6 U.S.DepartmentofEnergy 7 27 RAGFoundation 37 TigoEnergy,Inc. 6 UniversityofArizona 7 28 SchottAG 37 VerdantPower 6 ClemsonUniversity 6 29 CSIR 34 Coskata,Inc. 5 Queen'sUniversityatKingston 6 30 BP 33 FloDesignWindTurbineCorp. 5 RiceUniversity 6 31 RoyalDSM 32 KiorInc. 5 SwissFederalInstituteofTechnology(EPFL) 6 32 UTC 32 LUCATechnologies,Inc. 5 UniversityofColorado 6 33 IBM 31 MarineCurrentTurbinesLimited 5 UniversityofMichigan 6 34 RoyalDutchShell 31 Plextronics,Inc. 5 UniversityofToledo 6 35 SonyCorporation 31 PrimestarSolarInc. 5 AtomicEnergyCouncil 5 Patent Count

35 Appendix2:NewMeasureofNovelty WehavedevelopedanewmeasureofNoveltythatisnotbasedoncitationmeasures.Instead,wedraw onatextualanalysisofpatentapplicationstolookatthesimilarityofpatentclaimsanddescriptionsfor patentsinagiventechnologyarea.intuitively,ourdefinitionissuchthatpatentswithgreatertextual similaritytoneighboringpatentsareconsideredtobelessnovel.ourmeasureofnoveltyshouldbe particularlyusefulinthecontextofsciencebasedpatenting,wheretechnicaltermsaremoreunique andthereforemorelikelytosignaldifferencesinthecharacteristicsofinnovation,andformorerecent timeperiods,whereinitialforwardcitationsmaybeanoisypredictorofultimateoutcomes. WhileamorecompleteexpositionandvalidationofthemeasureofNoveltyisbeingdeveloped elsewhere(seeyounge,2013readythisfall),thegeneraloutlineforthecalculationofthemeasureis asfollows: First,thecalculationalgorithmreviewseverypatentclaimanddescriptioninthesampletobuildalistof alltermsused;thelistoftermsconstitutesahighdimensionalpositivespacewhereineachterm representsadimensionintothatspace.second,thealgorithmpositionseachpatentinthevectorspace byassigningitasetofcoordinateswherethemagnitudeofeachdimensioniscalculatedasthe"term frequencyinversedocumentfrequency"(tfidf)ofeachterminthepatent.intuitively,tfidfgivesa greaterweighttoadimensionwhenatermoccursmorefrequentlyinthepatent,andgivesalesser weighttoadimensionifthewordisfrequentlyobservedinotherpatentsaswell.third,thealgorithm calculatesthe similarity betweeneverypossiblecombinationoftwopatents,bycalculatingthecosine oftheangleformedbetweentheirvectors.themeasurementofsimilarityisbounded[0,1],witha measurementof1representingaperfectsimilaritybetweentwopatents. Havingthusarrivedatapairwiselistofsimilaritycomparisons,betweeneverypossiblecombinationof patentsinthesample,thealgorithmthencalculatesameasurementofnoveltyforeachfocalpatentby examiningthedistributionofsimilaritiesrelativetoacomparisonsetofpatents.thecomparisonsetis drawfromthepriorthreeyearsandfromthesametechnologyareaasthefocalpatent(e.g., Solar ). Toassessthe novelty ofapatent aconceptconnotingfewneighborsinthetechnologylandscape wetakethe5thpercentileoftherankordereddistributionofsimilaritiestiedtothecomparisonset.for easeofinterpretation,wereversethenoveltymeasurebysubtractingitfrom1,arrivingata measurementfornoveltythatisbounded[0,1],where1representsapatentthatisentirelydissimilar fromallotherpatents.whenneeded,weaveragepatentlevelmeasuresofnoveltyuptothefirmor categorylevel.