ResearchontheinfluenceofgovernmentR&Dfundingonenterprisetechnologyinnovationdecision,behaviorandperformance

——Basedonthemicropaneldataofhigh-techenterprises

Abstract

Asthecorecontentoftechnologicalinnovationactivities,researchanddevelopment(R&D)isthecoredrivingfactorforcountriesandregionstoenhancetheirindependentinnovationabilityandpromoteeconomicandsocialdevelopment.Researchanddevelopmentactivitieshavethecharacteristicsof"quasi-publicgoods".ItisoneofthecorecontentsofscienceandtechnologypoliciesofvariouscountriestocorrectthepossiblemarketfailureofenterpriseR&DactivitiesthroughgovernmentR&Dsubsidies.WiththecontinuousincreaseofR&Dinvestmentinmanycountriesandregions,moreandmorescholarsathomeandabroadhavepaidattentiontotheinfluenceofgovernmentR&Dfundingonenterprisetechnologicalinnovation,whichalsohighlightsthetheoreticalandpracticalvalueofthisproblem.Duetothelackofdataatthemicrolevel,theempiricalresearchesontheeffectofChinesegovernmentR&Dsubsidiesaremostlyconcentratedattheindustrylevelorregionallevel,andmostoftherelevantresearchesfocusontheimpactofgovernmentR&DsubsidiesonenterpriseR&Dinvestmentorenterpriseinnovationperformance,andthereisalackofrelevantresearchesonthemulti-dimensionalimpactofgovernmentR&Dsubsidiesonenterprisetechnologicalinnovation.Atthesametime,afewempiricalresearchdataatthelevelofmicroenterprisesalsobasicallyadoptcross-sectiondata.DuetothelagofR&Dactivities,suchresearchwithcross-sectiondataassamplesisnotenoughtoobjectivelyreflecttheactualeffectofgovernmentR&Dsubsidies.Inviewofthis,thispaperfirstlylogicallydemonstratestheinfluencemechanismofgovernmentR&Dsubsidiesonenterpriseinnovationdecision-making,innovationbehaviorandinnovationperformance,andconstructsaconceptualmodelofgovernmentR&Dsubsidiesonthewholeprocessofenterprisetechnologyinnovation.Onthisbasis,thepaneldataof2,025high-techenterprisesinGuangdongProvincefrom2010to2014aretakenassamples.Byusingrandomeffectsmodel(PRM),fixedeffectsmodel(FEM)andsystemgeneralizedmomentestimation(SYS-GMM),wecananalyzefiveapproachesfromdifferentfundingmethods,differentgrowthstages,differententerprisesizes,differentindustrycharacteristicsandtimedelayeffect.ThispaperempiricallyexaminesthespecificeffectsofgovernmentR&Dsubsidiesoninnovationdecision-making,innovationbehaviorandinnovationperformance.Theresearchconclusionofthispapernotonlydeepenstheunderstandingofhowgovernmentpoliciesaffectenterprisetechnologicalinnovationininnovationsystemtheory,enterpriseresource-basedviewtheoryandinstitutionalembeddednesstheory,butalsoprovidestheoreticalsupportandempiricalevidenceforthegovernmenttodevelopmorescientific,rationalandtargetedscienceandtechnologypoliciestopromoteenterpriseR&Dactivities.

Keywords:governmentR&Dsubsidy;Enterpriseinnovationdecision;Enterpriseinnovationbehavior;Enterpriseinnovationperformance

Contents

TOC\o"1-3"\h\u

21276

Abstract

1Researchhypothesis

1.1TheMechanismofGovernmentR&DFundingImpactonCorporateTechnologicalInnovationChoices

Althoughenterprisesrecognizetheimportantroleoftechnologicalinnovationinmaintainingcompetitiveadvantage,However,duetothegreattechnologicalriskandmarketriskofinnovation,andthepositiveexternalityofcommontechnology,forthesakeofavoidingrisks,domesticenterprisesareunwillingtoinvesttoomuchresourcestocarryouttechnologicalinnovation,butprefertoobtainitfromotherways,suchastechnologyspilloverandsupplyfromothersocialsubjects.Therefore,reducingthetechnicalrisksandmarketrisksintechnologyresearchanddevelopmentcaneffectivelyguideenterprisestoinvestresourcesintechnologicalinnovation.GovernmentR&Dfundingcanaffecttheinternalresourceallocationofenterprises,effectivelyreducethecostofR&Dtechnology,improveR&Defficiency,andthenaffecttheinnovationdecisionofenterprises.Finally,governmentR&Dfundingwillhaveasignaleffecttoenhancetheexternalfinancingabilityofenterprises,andalleviatetheimpactoffinancingconstraintsoninnovationdecisionsofenterprises.Mostofthefundsforenterprisestocarryoutinnovationactivitiesareexternalfinancing,Inparticular,enterprisesneedalotoffinancialsupportforinnovativetechnologyresearchanddevelopment,Atthistime,enterprisesneedexternalfinancingtosolvetheirtechnologyresearchanddevelopmentproblems,However,thereisinformationasymmetrybetweenexternalinvestorsandenterprises,Thiswillcauseseriousfinancingconstraintstoenterprises,However,thegovernment'ssubsidytoenterprisesisnotarandombehavior,butanin-depthinvestigationandcanreflectthecompetitiveprospectsofenterprisestoacertainextent,soitcanbetransmittedtoexternalinvestorsasapositivesignal,increaseexternalfinancing,easethefinancingconstraintsofenterprisesandaffecttheirinnovationdecisions.

1.2TheProcessbywhichGovernmentR&DSubsidiesAffectBusinessTechnologicalInnovationOutcomes

Ontheonehand,thegovernmenthopestoincreasetheamountofpatentoutput,promotesocialprogressandeconomicgrowth,andimprovetheutilizationrateofinnovationresources.Ontheotherhand,governmentsubsidieshopetoincreasetheoutputvalueofenterprises'newproductsandpromotetheimprovementofgovernmentperformance.Intwoaspects,thegovernmentwillinevitablyexertinstitutionalpressureontheenterprisesreceivingfinancialsupport,forcingthemtocontinuetoinnovate.TherearethreemainwaysforgovernmentR&Dsubsidiestoinfluenceenterprises'innovationperformance.First,ithelpstoimproveenterprises'technologicalcapabilityandimproveR&Defficiency.GovernmentsubsidiesforenterpriseR&Denableenterprisestohavemoreresourcestoseekbetterexternaltechnicalforcesforcooperation,andtheinjectionofmorehigh-qualityinnovationresourcesimprovestheirowntechnicalcapabilities,whichmayimprovetheefficiencyofnewproductresearchanddevelopment.Second,createmarketdemandfornewproductsthroughgovernmentprocurement,reducethemarketriskofdevelopingnewproducts,andincreasethesalesrevenueofnewproducts.Third,theevaluationindexofgovernment-fundedprojectswillalsoencourageenterprisestoapplyformorepatents.

1.3ExaminationofKeyElementsInfluencingtheEfficacyofGovernmentR&DSubsidies

1.3.1Firmheterogeneity

(1)ThedifferencebetweenenterprisesizeandgovernmentR&Dsubsidies

Numerousstudieshavesuggestedthattheinnovativeactivitiesofsmesorstart-upscanallowthemtodestroylargerfirmsthathaveestablishedacompetitiveadvantageearlier.TheauthorbelievesthatthedifferenceoftheeffectofgovernmentR&Dsubsidiesonenterprisesofdifferentsizescannotbeconcludedsimply,anditmustbeconsideredcomprehensivelyfromtheaspectsofenterpriseinnovationdecision-making,innovationbehaviorandinnovationperformance.Theoreticallyspeaking,thesmallertheenterpriseis,themoredifficultitistoinvestinresearchanddevelopmentfunds(lessfundsandfinancing).GovernmentR&Dsubsidiesoftenhavetheeffectof"providinghelpinneed"forsmallenterprises,whichmaybringrelativelylargesubsidyeffect.Ontheotherhand,largeenterprisestendtohaveabundantfundsandinvestlargeamountsofR&Dfunds,whiletheproportionofgovernmentR&DfundsintheirR&Dfundsisrelativelysmall.Itislikea"finestoneintothesea"thatcannotcreatealargewave,andtheguidingeffectofgovernmentR&Dsubsidiesislimited.Therefore,thispaperputsforwardthefollowinghypothesis:

H4a:TheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationdecision-makingisnegativelycorrelatedwithenterprisescale,thatis,thesmallertheenterprisescale,thegreatertheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationdecision-making.

H4b:TheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationbehaviorisnegativelycorrelatedwithenterprisescale,thatis,thesmallertheenterprisescale,thegreatertheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationbehavior.

H4c:TheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationperformanceisnegativelycorrelatedwithenterprisescale,thatis,thesmallertheenterprisescale,thegreatertheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationperformance.

(2)Enterprisegrowthstageandgovernment

Ithasbeenwidelyrecognizedbytheacademiccirclethatenterprisesindifferentgrowthstageshavesignificantlydifferentdemandsforfunds.Thispaperarguesthatenterprisesatdifferentgrowthstageshavesignificantdifferencesincapitalaccumulation,technicalcapability,R&Dinstitutionsandotheraspects,andtheireffectsongovernmentR&Dsubsidiesmayalsohavesignificantdifferences.Theoreticallyspeaking,asenterprisesgraduallyenterthedeclineperiodfrominfancy,development,growthandmaturity,theirR&Dwillingnessandvitalityshowamarginaldecreasingeffect.Sincetheageofenterprisesisusedtomeasurethegrowthstageofenterprisesinthispaper,thefollowingassumptionsareproposed:

H5a:TheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationdecision-makingisnegativelycorrelatedwiththeageoftheenterprise,thatis,theyoungertheenterpriseis,thegreatertheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationdecision-makingis.

H5b:TheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationbehaviorisnegativelycorrelatedwithenterpriseage,thatis,theyoungertheenterpriseage,thegreatertheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationbehavior.

H5c:TheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationperformanceisnegativelycorrelatedwithenterpriseage,thatis,theyoungertheenterpriseage,thegreatertheinfluencecoefficientofgovernmentR&Dsubsidiesonenterpriseinnovationperformance.

（3）Industrysectorandgovernment

Theoreticallyspeaking,themoretechnology-intensiveindustries,themoreneedthesupportofR&Dactivities,becausetechnology-intensiveindustriesareoftencharacterizedbyhighrisk,highyieldandlongcycle,thepossibilityofmarketfailureisgreaterthanthatofresource-intensiveandlabor-intensiveindustries,sotheeffectofgovernmentR&Dsubsidiesshouldbemoreeffective.Thesampleofthispaperishigh-techenterprises,includingmanyspecificindustries,whichhavedifferentriskandperiodicityofR&D.Accordingly,thispaperproposesthefollowinghypothesis:

H6a:TherearesignificantdifferencesintheinfluenceofgovernmentR&Dsubsidiesoninnovationdecisionsofenterprisesindifferentindustries.

H6b:TheinfluenceofgovernmentR&Dsubsidiesontheinnovationbehaviorofenterprisesindifferentindustriesissignificantlydifferent.

H6c:TheinfluenceofgovernmentR&Dsubsidiesontheinnovationperformanceofenterprisesindifferentindustriesissignificantlydifferent.

1.3.2Policytime-lag

Theviewthatpolicyhastimelaghasbeenrecognizedbythevastmajorityofresearchers.Theoreticallyspeaking,theeffectofgovernmentR&Dfundingmusthavesometimelag.Forexample,Goldberg(1979)foundthatthefederalgovernment'sR&DinvestmentinthecurrentperiodhasasignificantnegativecorrelationwithenterpriseR&Dinvestment,butitshowsasignificantpositivecorrelationinthelagperiod.Levy&Terleckyj(1983)foundthattheimpactofotherformsoffundingotherthancontractonthetotalR&Dexpenditureofanenterpriseisnotsignificantinthecurrentperiod,withasignificantpositiveimpactafter3yearslag,andasignificantnegativeimpactontheproductioncapacityafter9yearslag.FrancescoBogliacino(2009)foundthatthelagperiodoftheimpactofgovernmentscienceandtechnologyfundsonenterpriseinnovationperformanceisabout3-4years,etc.Therefore,thispaperputsforwardthefollowinghypothesis:

H7a:TheeffectofgovernmentR&Dsubsidiesonenterpriseinnovationdecision-makinghasobvioustimelag,anditsinfluencecoefficientdecreasesgraduallywithlagperiod.

H7b:TheeffectofgovernmentR&Dsubsidiesonenterpriseinnovationbehaviorhasobvioustimelag,anditsinfluencecoefficientdecreasesgraduallywithlagperiod.

H7c:TheeffectofgovernmentR&Dsubsidiesonfirminnovationperformancehasobvioustimelag,anditsinfluencecoefficientdecreasesgraduallywithlagperiod.

1.3.3Differencesinfundingmethods

Theoretically,theeffectsofdirectgovernmentR&Dfundingandtaxreliefondifferentprocessesandlinksofenterpriseinnovationmaybequitedifferent:Asforenterpriseinnovationdecision-making,directfinancingusuallyhasanimpactonenterpriseinnovationdecision-makingbeforethestartofenterpriseresearchanddevelopmentactivitiesbecauseoftheformofR&Dcontract,whiletaxincentivesandexemptionsaremoreaimedatsubsidizingtheresultsaftertheoccurrenceofR&Dactivities.Therefore,theimpactofdirectfinancingistheoreticallybetterthanthatoftaxrelief.Asforenterprises'innovationinvestmentbehavior,taxincentivesandexemptionsarelinkedtoenterprises'R&Dinvestment,whichshouldhaveamoreeffectiveimpactonenterprises'R&Dinvestment.Althoughthegovernment'sdirectR&DfundingrequiresenterprisestoinvestacertainamountoftheirownR&Dfunds,itsimpactshouldbemorelimited.Therefore,inthislink,taxincentivesandexemptionsmaybemoreeffectivethandirectfunding.Inordertofacilitatetheempiricalresearch,thispaperputsforwardtheoriginalhypothesisthattheinfluencecoefficientofthegovernment'sdirectR&Dfundingislargerthanthatofthetaxincentivesandexemptions.Then,basedontheempiricalresults,theoriginalhypothesisofthethreelinksofenterprisetechnologicalinnovationistestedandfurtheranalyzedanddiscussed.

H8a:TheinfluencecoefficientofgovernmentR&Ddirectfundingonenterpriseinnovationdecision-makingislargerthanthatoftaxincentivesandexemptions.

H8b:TheinfluencecoefficientofgovernmentR&Ddirectfundingonenterpriseinnovationbehaviorislargerthanthatoftaxincentivesandexemptions.

H8c:TheinfluencecoefficientofgovernmentR&Ddirectfundingonenterpriseinnovationperformanceislargerthanthatoftaxincentivesandexemptions.

Accordingtotheconceptualmodelandresearchhypothesismentionedabove,thispapersummarizesandorganizes.Fordetails,seeTable2.1.Ingeneral,theresearchhypothesesinthispaperinclude8corehypotheses,whicharemainlydevelopedfromeightaspects:enterprisetechnologyinnovationdecision-making,behavior,performance,differententerprisesizes,differentgrowthstages,differentindustryareas,differentfundingmethods,andtimelag.Combinedwiththemeasurementindicatorsinvolvedinthispaper,the8corehypothesesareexpanded.The21sub-hypothesestobetestedempiricallyareformed,andtheempiricaltestsofthehypotheseswillbesummarizedandanalyzedintheresultsdiscussionsection.

Table2.1Generalsituationofresearchhypothesisinthispaper

2InvestigatingtheImpactofGovernmentR&DSubsidiesonBusinessTechnologicalInnovationActions

2.1Fullsampleempiricalanalysis

2.1.1ConfiguringtheModel

Thedata,variables,individualeffectsandtimeeffectshavebeentestedaccordingly,andtheabnormalsituationhasbeendealtwith.Nowtheempiricalanalysisofthewholesampleiscarriedout.Inordertoidentifythesettingformofthefixedeffectmodelmoreeffectivelyandscientifically,thestepwiseregressionmethodwillbeadoptedforempiricaltestofthefollowingequation.Ascanbeseenfromtheprevioustest,sincethedimensionsofvariablesamongindividualsarenotuniformandthenumericaldifferencebetweenindividualsislarge,inordertoensuretheeffectivenessofregression,itisnecessarytocarryoutregressionafterlogarithmofvariables(RDPP,theproportionofR&Dpersonnel,isalreadyapercentageindex,whichwillnottakelogarithm).Themodelequationstobetestedinthischapterarelistedbelow.

2.1.2RegressionOutcome

AfterLMtestandHusmantest,Models3-3through3-9wereadouble-effectmodelwithfixedindividualsandrandomtime,PanelEGLSregressionwasused,andWhitecrosssectionandWhiteCrosssectionwereusedforcoefficientstabilitycorrection.Accordingtotheregressionresults,thereisasignificantpositivecorrelationbetweengovernmentdirectR&Dfunding(GI)andenterpriseR&Dinvestment(EI).Ifothercontrolvariablesarenottakenintoaccount,thecorrelationcoefficientis0.032(significantat1%level),thatis,every1%increaseingovernmentdirectR&DfundingwilldriveenterpriseR&Dinvestmenttoincreaseby0.032%.AftergraduallyaddingcontrolvariablessuchastheproportionofR&Dpersonnel,netprofitandtotalassets,theresultingcoefficientdecreasesgradually.Model3-9isanequationwithallcontrolvariablesadded.TheinfluencecoefficientofgovernmentR&DdirectfundingonenterpriseR&Dinputis0.024(significantat1%level),andallcontrolvariablespassthesignificancetest.Amongthecontrolvariables,RDPPhasthelargestcausingcoefficient(2.035,whichissignificantat1%level).ItshouldbenotedthatlogarithmicratioofRDPPisnottakenhere.TheeconomicimplicationisthateverypercentagepointincreaseofRDPPwillincreaseR&Dinputby2.035%.TheindexofPROFITSandT_ASSETSalsoshowapositivecorrelationwithR&Dinvestmentofenterprises.

Table3.1RegressionresultsoftheinfluenceofgovernmentR&Ddirectsubsidy(GI)onenterpriseR&Dinvestment(EI)

Note:***,**,*respectivelymeans1%,5%,10%levelissignificant.

TheempiricalresultsinTable3.1showthattheinfluencecoefficientofgovernmentR&DdirectfundingonenterpriseR&Dinputis0.024(significantat1%level),whichindicatestosomeextentthatgovernmentR&Ddirectfundingispositivelycorrelatedwithenterprisetechnologicalinnovationinputwhenothervariablesremainunchanged.

2.2TimeLagEmpiricalInvestigation

2.2.1Modelsetting

TheSTATA"xtabond2"commandisusedtoperformGMMestimationonthesetdynamicmodel.Thedynamicpanelmodelequationmainlyverifiedisasfollows:

Where,lrepresentstheautoregressivelagorderofenterpriseR&Dinvestment(EI),anditsvaluesare1,2,3;mrepresentsthelagorderofgovernmentR&Ddirectfunding(GI),whichis0,1,2,3;nrepresentsthelagorderoftaxpreferencereduction(TC)enjoyedbyenterprises,andthevaluesare0,1,2,3.InModel2-11,thelagtermsofGIandTCareintroducedtoinvestigatethelongandshorttermeffectsofgovernmentR&Dfundingpolicies.Theidrepresentsthe"individualeffect",andthatistheerrorterm.

Sincemodels2-10and2-11containthelagtermofEI,theendogeneityofvariablesmayoccur.Therefore,Systemgeneralizedmomentestimationmethod(SystemGMM)mustbeadopted.TheSystemGMMmethodusestheorthogonalmomentofthehorizontalvariableandthefirstorderdifferencerandomterm,andtheorthogonalmomentofthefirstorderdifferencevariableandthehorizontalrandomterm.Thistreatmentcannotonlyeffectivelysolvetheendogeneityproblem,butalsoeliminatetheinfluenceofindividualeffect(LiPing&WangChunhui,2010).Inaddition,dynamicpanelGMMestimatesaresuitableforpaneldatawithsmallertimedimensionsandlargercross-sectiondimensions(wideandshortpanels).Thetimedimensionofthispaperis5andthesectiondimensionis2023,soitismoreappropriatetousetheGMMmethod.

2.2.2RegressionOutcomesandEvaluation

Bond(2002)believesthatifthesamplesaretoosmallortherepresentativenessoftheselectedinstrumentalvariablesisweak,theGMMestimatorofthedynamicpanelispronetoproducealargedeviation.HeproposedthattheOLSestimatorandthestaticfixedeffectestimatorcouldbeusedtotestthedeviation,andtheoreticallytheGMMestimatorshouldbebetweentheOLSestimatorandthefixedeffectestimator.Tothisend,thispaperfirstconductedgeneralOLSestimationandstaticfixedeffectestimationformodels2-10and2-11,inordertocomparethemwithdynamicGMMestimation.

Table2.2Staticpanelregressionresultsofmodels(2-10)

Note:Valuesinbracketsarestandarderror,*,**and***respectivelyindicatesignificantatthesignificancelevelof10%,5%and1%.

Table2.3Staticpanelregressionresultsofmodels(2-11)

Note:Valuesinbracketsarestandarderror,*,**and***respectivelyindicatesignificantatthesignificancelevelof10%,5%and1%.

ThekeytoGMMestimationistodetermineendogenousvariablesandinstrumentalvariables.SomescholarsbelievethattheindexofdirectgovernmentR&Dfunding(GI)isendogenous,thatis,governmentR&Dfundingisnotrandomlyallocatedtoenterprises,butisselective.Therefore,inthispaper,thelagtermandGIofEIareregardedasendogenousvariables,thefirstorderoflagisusedasinstrumentalvariables,corporatenetprofitandgovernmenttaxincentivesandexemptionsareregardedasexogenousvariables,andtheircurrentvalueisusedasinstrumentalvariables.Thispaperadoptsone-stepestimationmethod.UsetheSTATA"xtabond2"commandtogettheresultsoftheSystemGMMestimate(seeTable2.4)。

Table2.4LaganalysisoftheinfluenceofgovernmentR&DsubsidiesonenterpriseR&Dinvestment

Itcanbeseenfromtheresultsthatthecoefficientofthedelayedexplainedvariableis0.414(significantat1%level),indicatingthattheenterpriseR&DinputofthepreviousyearhasapositiveimpactontheenterpriseR&Dinputofthecurrentperiod.Fromthepracticalsenseofunderstanding,theenterprise'sR&Dinvestmenthasacertaincontinuityandstability,thegreatertheenterprise'sR&Dinvestmentinthepreviousyear,usuallythegreatertheR&Dinvestmentinthatyear.Accordingtotheregressionresults,whenthelagtermofEIisconsidered,thelagtermofgovernmentR&Ddirectsubsidy(lnGI)andtaxincentives(lnTC)cannotpassthesignificancetest,indicatingthatthelageffectofgovernmentR&Dsubsidypolicyisnotsignificantwhentheendogeneityofvariablesisconsidered.ThisisdifferentfromtheresultsofmixedOLSestimationandfixedeffectestimation.Thefirst,secondandthirdperiodsofgovernmentR&Dsubsidiesestimatedbythelattertwohavesignificantimpactsonenterprises'R&Dinvestment.TheauthorrevisedSYS-GMMestimatesformanytimes,buttheresultisstillthesame.Whennoobviousproblemswerefoundinmodelsetting,variableselectionanddataprocessinginthispaper,itindicatesthatthereisnotimelageffectontheinfluenceofgovernmentR&Dsubsidiesonenterprises'R&Dinputunderthepremiseofconsideringendogeneity.

2.3EmpiricalStudyofVaryingEnterpriseSizes

2.3.1Modelsetting

Inthispaper,thesamplegroupisdividedintofoursamplegroups:largeenterprises,mediumenterprises,smallenterprisesandmicroenterprises.Theregressionanalysisisconductedonthesamplegroupsrespectively,andthedifferenceoftheinfluenceofgovernmentR&Dsubsidiesonenterprisesofdifferentsizesisdiscussedaccordingtotheregressionresults.Inordertofocusonthedifferencesofdifferentscales,thegovernmentR&Ddirectsubsidies(GI)andtaxincentives(TC)areaddedintothemodelforempiricalstudy.ThecontrolvariablesaretheratioofR&Dpersonnel,corporatenetprofitandtotalassetsofenterprises.Themodelequationisasfollows：

2.3.2RegressionOutcomesandEvaluation

Fixed-effectmodelregressionwascarriedoutforthefoursamplegroupsrespectively,andtheregressionresultswereshowninTable2.5.Ascanbeseenfromthetable,thethreesamplegroupsoflargeenterprises,mediumenterprisesandsmallenterprisesallpassthesignificancetest,butthegovernmentpreferentialtaxrelief(TC)andenterprisetotalassets(T_ASSETS)ofmicroenterprisesfailtopassthesignificancetest.Forthesamplegroupoflargeenterprises,thecorrelationcoefficientofdirectgovernmentR&DsubsidiesonenterpriseR&Dinvestmentis0.012(significantat1%level),andthecorrelationcoefficientoftaxincentivesandexemptionsonenterpriseR&Dinvestmentis0.019(significantat1%level).Forthesamplegroupofmedium-sizedenterprises,thecorrelationcoefficientofdirectgovernmentR&DsubsidiesonenterpriseR&Dinvestmentis0.018(significantat1%level),andthecorrelationcoefficientoftaxincentivesonenterpriseR&Dinvestmentis0.032(significantat1%level).Forthesamplegroupofsmallenterprises,thecorrelationcoefficientofgovernmentR&DsubsidiesonenterpriseR&Dinvestmentis0.021(significantat1%level),andthecorrelationcoefficientoftaxincentivesonenterpriseR&Dinvestmentis0.042(significantat1%level).Forthesamplegroupofmicro-enterprises,thecorrelationcoefficientofdirectgovernmentR&DsubsidiesonenterpriseR&Dinvestmentis0.117(significantat1%level),whilethecorrelationcoefficientoftaxincentivesonenterpriseR&Dinvestmentfailstopassthesignificancetest.

ItcanbeseenthattheinducementcoefficientofgovernmentR&Ddirectsubsidy(GI)onenterpriseR&Dinvestmentbasicallyshowsadecreasingtrendwiththescale.Fromthestatisticalpointofview,forevery1%increaseofgovernmentR&Dsubsidiesforlargeenterprises,theincreaseofenterpriseR&Dinvestmentis0.012%.Formedium-sizedenterprises,theincreaseofR&Dinvestmentis0.018%forevery1%increaseofgovernmentR&Dsubsidies.WhenthegovernmentR&Dsubsidyforsmallenterprisesincreasesby1%,theenterpriseR&Dinvestmentincreasesby0.021%.Formicroenterprises,theincreaseofR&Dinvestmentis0.117%forevery1%increaseofgovernmentR&Dsubsidies.TheincentivecoefficientoftaxincentivesonR&Dinvestmentalsodecreaseswithscale,butitsincentivecoefficientislargerthanthatofdirectgovernmentsubsidies.

Table2.5ThedifferenceofinfluenceofgovernmentR&DsubsidiesonR&Dinputofenterprisesofdifferentsizes

Enterprisesize/

Explanatoryvariables

largeenterprise

mediumenterprise

s