2024年追逐夢(mèng)想：歐洲工業(yè)生產(chǎn)力的發(fā)展報(bào)告（英文版）-國(guó)際貨幣基金組織

ChasingtheDream:

Industry-Level

Productivity

DevelopmentsinEurope

SerhanCevik,SadhnaNaik,andKeyraPrimus

WP/24/258

NATn

2024

DEC

ARY

?2024InternationalMonetaryFundWP/24/258

IMFWorkingPaper

EuropeanDepartment

ChasingtheDream:Industry-LevelProductivityDevelopmentsinEuropePreparedbySerhanCevik,SadhnaNaikandKeyraPrimus1

AuthorizedfordistributionbyKazukoShironoDecember2024

IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.

Abstract

Europeancountriesarelaggingbehindinproductivitygrowth,withsignificantproductivitygaps

acrossindustries.Inthisstudy,weusecomparableindustry-leveldatatoexplorethepatternsandsourcesoftotalfactorproductivity(TFP)growthacross28countriesinEuropeovertheperiod

1995–2020.Ourempiricalresultshighlightfourmainpoints:(i)TFPgrowthisdrivenlargelybytheextenttowhichcountriesareinvolvedinscientificandtechnologicalinnovationastheleader

countryorbenefitingfromstrongerknowledgespillovers;(ii)thetechnologicalgapisassociated

withTFPgrowthascountriesmovetowardsthetechnologicalfrontierbyadoptingnewinnovationsandtechnologies;(iii)increasedinvestmentininformationandcommunicationstechnology(ICT)

capitalandresearchanddevelopment(R&D)contributessignificantlytohigherTFPgrowth;and(iv)theimpactofhumancapitaltendstobestrongerwhenacountryisclosertothetechnologicalfrontier.ThecorefindingsofthisstudycallforpolicymeasuresandstructuralreformstopromoteinnovationandfacilitatethediffusionofnewandexistingtechnologiesacrossEurope.

JELClassificationNumbers:

H32;H40;L50;L60;L80;O30;O40;O52

Keywords:

Totalfactorproductivity;technology;R&D;innovation;humancapital;Europe

Author’sE-MailAddress:

scevik@;

snaik@;

kprimus@

1TheauthorswouldliketothankSandraBaquie,HelgeBerger,LukasBoer,BorjaGracia,VincenzoGuzzo,FlorenceJaumotte,andYangYangforvaluablecommentsandsuggestions.

3

Productivityisn'teverything,but,inthelongrun,itisalmosteverything.

—PaulKrugman

I.INTRODUCTION

Followinganextendedperiodofeconomictranquilityandrapidincomeconvergence,Europe

hasexperiencedabarrageoflargeshocksinrecentyearsthatresultedindivergingtrendsin

productivitygrowth,whichiskeytoraisingmateriallivingstandards,expandingtheeconomy’sgrowthpotential,andstrengtheninginternationalcompetitiveness.Understandingthedriversoftotalfactorproductivity(TFP)growth—ameasureoftechnologicaladvancementsandthe

efficiencyinutilizingfactorsofproduction—isthereforenecessarytodeveloppoliciesthatcanhelpstrengthengrowthprospects.WeobservethataggregateTFPgrowthintheEuropean

Union(EU)declinedfromanaverageof0.7percentbetween1996and2007to0.1percentover

theperiod2009–2019and-2percentin2020duringtheCOVID-19pandemic(Figure1).WealsodetectsignificantvariationinaverageTFPgrowthratesacrosstheEUovertheperiod1996–2020,

withaminimumof-2percentinGreecetoamaximumof2percentinSlovakia.These

productivitydevelopmentsattheaggregatelevel,however,canreflectsignificantstructuraldifferencesinhumanandphysicalcapitalaccumulationandtechnologicalprogressatthe

industrylevel.Accordingly,toprovideagranularempiricalassessment,thispaperfocusesonindustry-levelproductivitydevelopmentsthatdeterminetheaggregate.

TheproductivitygapbetweentheEUandtheUShaswidenedaftertheglobalfinancialcrisis

(GFC)in2008,withEUcountrieslaggingbehindinproductivitygrowth(Cette,Fernald,and

Mojon,2016;FernaldandInklaar,2020).Furthermore,therearesignificantproductivitygaps

acrossEUcountriesandindustries,whichhavebecomemoreprominentaftertheGFC.Inthis

study,welookbeyondthebroadcontoursofproductivitygrowthandusecomparableindustry-leveldata—drawnfromtheEU-KLEMSdataset—toexplorethepatternsandsourcesofTFP

growthacross28Europeancountriesovertheperiod1995–2020.Wecontributetotheliteraturebyusingthelatestandmostcomprehensiveindustry-leveldatasetincludingthepandemic

periodanddevelopingagranularanalysisoftradableandnon-tradablesectorsofthe

Figure1.TotalFactorProductivity(TFP)Growth

AverageTFPGrowth

(Percent)

8

6

4

2

0

-2

-4

-6

-8

Inter-QuartileRangeEU:TFPGrowth

1996199820002002200420062008201020122014201620182020

AverageTFPGrowth,1996-20201/

(Percent)

2.5 21.5 10.5 0-0.5 -1-1.5 -2-2.5

EL

PT

LU

HR

ES

IT

CY

SE

IE

FR

BE

NL

UK

DK

HU

EE

SI

AT

DE

CZ

PL

FI

LT

BG

LV

RO

SK

1/Time-spancoveredforeachcountryvariesdependingondataavailability.

Sources:EUKLEMS;authors’calculations.

4

economy.ThisrichsectoraldatasetcoveringalargenumberofeconomiesoveranextendedperiodbeforeandaftertheGFCallowsustoanalyzetheheterogeneityoftradableandnon-tradablesectorswithinandacrosscountries,aswellaswithin-sectorandbetween-sector

developmentsthataresensitivetoaggregationbias(deVriesetal.,2012;üng?r,2017).

Ourempiricalresults,inlinewithpreviousstudies,highlightfourmainpoints.2First,TFPgrowthisdrivenlargelybytheextenttowhichcountriesareinvolvedinscientificandtechnological

innovationastheleadercountryorbenefitingfromstrongerknowledgespillovers.Second,thetechnologicalgap—measuredbyacountry’sTFPdistancetothefrontier—isassociatedwithTFPgrowthascountriesmovetowardsthetechnologicalfrontierbyadoptingnewinnovationsand

technologies,asshowninFigure2.Third,increasedinvestmentininformationand

communicationstechnology(ICT)capitalandresearchanddevelopment(R&D)contributes

significantlytohigherTFPgrowthacrossallEUcountries.Fourth,humancapitalasmeasuredbytheintensityofhigh-skilledlaborattheindustryleveldoesnotappeartohaveastatistically

significantimpactonTFPgrowth,butthereissomeevidencethatthiseffectisstrongerwhenacountryisclosertothetechnologicalfrontierandhumancapitalmattersmoreinnon-tradablesthantradablesectorsoftheeconomy.Foramoregranularassessment,wealsoexplorethe

interactionofindustry-levelfactorswiththetechnologicalgapandfindthatbothICTandnon-ICTcapitalexpenditurestendtomoderatethenegativeeffectofthetechnologicalgaponTFP

growth.Lastly,weestimatethemodelforsubsamplesandshowthatthetechnologicalgapisanimportantdriveroftheTFPslowdowninpost-GFCperiod.

TFPgrowthintheEUstagnatedaftertheGFCandturnednegativewiththeCOVID-19pandemic,withsizableproductivitygapsbetweendifferentindustriesandacrosscountries.Reversingthe

downwardtrendandboostingproductivitygrowtharekeytoraisinglivingstandardsamidadversedemographictransitionsandglobaleconomicrealignments.Asourindustry-level

Tradables

4-

2-

TechnologicalGap

Figure2.TFPGrowthandTechnologicalGap

Non-Tradables

TechnologicalGap

Note:Thesechartsshowabinnedscatterplotof9,151observations.

Sources:EUKLEMS;authors’calculations.

2Theindustry-levelanalysispresentedinthispaperisbroadlyconsistentwithfirm-levelestimationsforEurope,whichalsohighlighttheimportanceofR&Dinvestmenttoboostproductivitygrowth(IMF,2024a).

5

empiricalanalysisindicates,revampingtangibleandintangiblecapitalinvestmentinnew

technologiescangeneratehigherTFPgrowthdirectlyandindirectlybyclosingthetechnologicalgapvis-à-visthefrontier.Wealsofindsomeevidencethathumancapitalmattersmorewhenacountryisclosertothetechnologicalfrontier,especiallyinnon-tradablesectors.Basedonthesefindings,thepriorityshouldbetocreateaconduciveenvironmenttoraisebusinessinvestmentandimprovecapitalallocationbyprovidingincentivesforcapitalinvestmentandR&Dand

strengtheninghumancapitalaccumulationthrougheducationandhealthcare,whichcaninturnpromoteinnovationandfacilitatethediffusionofnewandexistingtechnologiestocountries

belowthetechnologyfrontier,withpositivespilloversacrossindustries(Akcigit,Baslandze,andStantcheva,2016;Akcigit,Hanley,andSerrano-Velarde,2021;IMF,2024b).

TheempiricalanalysispresentedinthispapersuggestsanimportantroleforpoliciesinreducingthetechnologicalgapamongcountriesinEurope.Narrowinginnovationandtechnologygaps

vis-à-visthefrontierandexpandingthefrontierarekeytoadvancingproductivitygrowthonasustainablebasis.Thisrequires(i)revampingtangibleandintangiblecapitalinvestmentinnewtechnologiesand(ii)strengtheninghumancapitalforrapidprogressinscienceandtechnology.ThepriorityshouldthereforebegiventocreatingaconduciveenvironmentforhigherbusinessinvestmentandbettercapitalallocationbyprovidingincentivesforcapitalinvestmentandR&Dandstrengtheninghumancapitalaccumulationthrougheducationandhealthcare,whichcanin

turnpromoteinnovationandfacilitatethediffusionofnewandexistingtechnologiestocountriesbelowthetechnologyfrontier,withpositivespilloversacrossindustries.

Theremainderofthisstudyisorganizedasfollows.SectionIIprovidesabriefoverviewofthe

relatedliterature.SectionIIIdescribesthedatausedintheanalysis.SectionIVintroducesthe

salientfeaturesofoureconometricstrategy.SectionVpresentstheempiricalresults,includingaseriesofrobustnesschecks.Finally,SectionVIsummarizesandprovidesconcludingremarkswithpolicyimplications.

II.LITERATUREREVIEW

Theconceptualframeworkfortheanalysispresentedinthispaperisbasedonthestandard

modelofconditionalconvergence.Thisimpliesthatcountriescancatchuptothetechnological

frontier.However,differencesinsteady-statelevelsofproductivitydependonstructuralfeaturesoftheeconomy,suchaslaborandproductmarketregulations,qualityofinstitutionaland

physicalinfrastructure,sociodemographicfactors,technologyinnovationandadoption,amongothers.Thereisarichliteratureaimingtoexplaincross-countrydifferentialsinproductivityandincomegrowthpatterns(Solow,1956,1957;Swan,1956;Bartelsman,Haltiwanger,andScarpetta,2013;CraftsandO’Rourke,2013;Cette,Fernald,andMojon,2016;égert,2016;Crafts,2018).

Forlaborproductivity,studiesfindthatphysicalandhumancapitalaccumulationarethemaindeterminantsoflaborproductivitygrowthandkeycontributorsofthedivergenceacross

countries(Lucas,1988;Wolff,1991;BenhabibandSpiegel,1994;Maudos,Pastor,andSerrano,2000;Barro,2001;KumarandRussell,2002;HendersonandRussell,2005;F?re,Grosskopf,andMargaritis,2006;EnfloandHjertstrand,2009;HanushekandWoessmann,2015;égert,dela

6

Maisonneuve,andTurner,2023).Thesefindingssuggestthatincreasingproductivityrequires

economicpoliciesdesignedtoreducebarrierstocapitaldeepeningandimprovetheeducationandhealthoftheworkforce.

Anotherimportantfactorincross-countrydifferencesinproductivitygrowthisscientificprogressandtechnologicalchange(NelsonandPhelps,1966;Romer,1987,1990;GrosmanandHelpman,1991;AghionandHowitt,1992;Greenwood,Hercowitz,andKrusell,1997;ArcelusandArozena,1999;Hulten,2001;Krüger,2003;GuellecandvanPottelsberghedelaPotterie,2004;Margaritis,F?re,andGrosskopf,2007;vanArk,O’Mahony,andTimmer,2008;Badunenko,Henderson,and

Zelenyuk,2008;Syverson,2011;Araujo,Vostroknutova,andWacker,2017).Griffith,Redding,andVanReenen(2004),InklaarandTimmer(2007),Jorgenson,Ho,andStiroh(2008)andSchiersch,Belitz,andGornig(2015)developamoregranularapproachtoanalyzeTFPgrowthandobtain

similarevidenceattheindustrylevel.ThesefindingsindicatethateconomicpolicyshouldalsoaimatfosteringR&Dandthediffusionofnewtechnologiestoboostproductivity.

PublicinfrastructureiscriticalforeffectiveandefficienteconomicactivityandtherebyTFP

growthbyenablingfirmstoinvestinmoreproductivemachinery,preventingdelaysin

production,andcontributingtoeducationandhealthcareoftheworkforce(Aschauer,1989;

Munnell,1992;Hulten1996;Straub,2011;Deng,2013;CalderonandServen,2014;égert,2016).Thequalityofinstitutionalinfrastructureisnolessimportantforpoliticalandsocioeconomic

stabilityandeconomicdevelopmentbysafeguardingcivilandpropertyrightsandprovidingasafelivingandworkingenvironment(North,1990;KnackandKeefer,1995;EasterlyandLevine,2003;Acemoglu,Johnson,andRobinson,2004;Rodrik,Subramanian,andTrebbi,2004;ChandaandDalgaard,2008).Physicalinfrastructureandinstitutionsalsoplayanimportantroleintradeopenness,financialdevelopmentandtheefficientallocationofresources,whichinturn

determineproductivitygrowthacrossfirmsandindustries(GrossmanandHelpman,1991;

Edwards,1998;Beck,Levine,andLoayza,2000;MillerandUpadhay,2000;Foster,Haltiwanger,andKrizan,2001;AlcaláandCiccone,2004;HisehandKlenow,2009;RestucciaandRogerson,2017;CevikandMiryugin,2018).

III.DATAOVERVIEW

Theempiricalanalysispresentedinthispaperisbasedonanunbalancedpanelofannual

observationson26industriesin28EUcountriesduringtheperiod1995–2020.3Theprimary

datasetforindustry-leveldataisobtainedfromtheEU-KLEMSdatabase,whichprovideshigh-

qualitymeasuresofeconomicgrowth,productivity,employmentcreation,capitalformationandtechnologicalchangeattheindustrylevelforallEUmemberstates.4Inparticular,itmakesdataavailableondifferenttypesofcapitalandskills-differentiatedcategoriesoflabor.Grossoutputisdecomposedintothecontributionsofintermediateinputs(i.e.,energy,materials,andservices)as

3IndustriesaregroupedaccordingtothestatisticalclassificationofeconomicactivitiesbasedontheNomenclaturedesActivitéséconomiquesdanslaCommunautéEuropéenne(NACE).

4Thelatestreleaseofthedatasetispubliclyavailableat

https://euklems-intanprod-llee.luiss.it/

(Bontadinietal.,2023).O’MahonyandTimmer(2009)provideadetaileddescriptionofthecontentsandconstructionoftheEU-KLEMSdatabase.

7

wellasvalue-added,whichinturnisdecomposedtothecontributionsfromdifferenttypesofcapitalandlabor.

Dataareavailableforsevendifferenttypesofcapital,whichareaggregatedbasedontheuser

costofcapitaltoproducecapitalserviceflowsthattakeintoaccountthedifferentmarginal

productivitiesofthedifferentcomponentsofacountry’scapitalaccumulation.Inadditionto

aggregatemeasures,theEU-KLEMSdatabasealsomakesavailablethebreakdownforICTand

non-ICTphysicalcapitalspendingasashareofgrossfixedinvestment.5Wealsoobtainindustry-leveldataonintangiblecapitalaccumulationasmeasuredbyR&Dexpenditureasashareof

grossfixedinvestment.

Laborinputsaredifferentiatedwithrespecttoskilllevelsasmeasuredbyeducationalattainment(primary,secondary,andtertiary),age,andgender.Inthispaper,weusetheshareoftotal

workinghoursprovidedbyworkerswithtertiaryeducationtomeasuretheshareofhigh-skilledlaborinagivenindustry.Comparablecross-countrydataattheindustrylevelareavailablefor

threedifferentskilllevels,withlaborinputsaggregatedbasedonmarginalproductivities.ThesegranulardataseriesallowtheassessmentofTFPdevelopmentsexcludingtheimpactofchangesinthecompositionandqualityofbothcapitalandlaborinputs.

Thedependentvariableinthisstudyisindustry-levelTFPgrowth,whichiscommonlymeasuredasaresidualafteraccountingforphysicalcapitaldeepeningandhumancapitalaccumulation.TheEU-KLEMSdatabaseprovidesadecompositionofGDPgrowthintoitsmaindeterminants

basedonaproductionfunctionwhichincludesproductivecapitalandemploymentlevels

adjustedforhoursworked,age,andforskillcomposition.Accordingly,TFPgrowthisdefinedasaresidualterm:

Δyijt=Δvijt?wvijt?wΔkijt?wΔLijt

whereΔyijtisTFPgrowthincountryiandindustryjattimetandV,K,andLdenotevalue-

added,capital,andlabor,respectively.Thecoefficientswandwaretheaverageshareof

capitalandlaborinputs,respectively.Tomitigatetheeffectsofextremeoutliers,wewinsorizeindustry-levelvariablesatthe5thand95thpercentiles.

ThemainexplanatoryvariablesofinterestattheindustrylevelaretheTFPgrowthfrontieras

measuredbythehighestlevelofTFPgrowthinagivenindustryandyearandthetechnologicalgapasmeasuredbythedistancetofrontierdefinedasthelevelofTFPofacountryinagiven

industryandyearrelativetothehighestlevelofTFPinthatindustryintheEU.Thisrelative

distancetothefrontierrepresentsthepotentialforincreasingTFPbyadoptingnewproductivity-enhancingknowledgeandtechnologies.Inaddition,weincludearangeofmacroeconomicandinstitutionalfactorsatthecountrylevel,suchasrealGDPpercapita,consumerpriceinflation,

tradeopenness,domesticcredittotheprivatesector(i.e.financialdevelopment),population,andbureaucraticqualityascontrolvariables,whicharedrawnfromtheIMF’sWorldEconomic

5ICTassetsincludecomputers,softwareandtelecommunicationequipment,whilenon-ICTassetsareproxiedbytransportationequipment.

8

Outlook(WEO),theWorldBank’sWorldDevelopmentIndicators(WDI),andtheInternationalCountryRiskGuide(ICRG)databases.

SummarystatisticsforthevariablesusedintheanalysisarepresentedinTable1.Weobserve

significantheterogeneityinaggregateandindustry-levelproductivitygrowthbetween28EU

countries,across26sectors,andovertheperiod1995–2020.AggregateTFPgrowthwas0.7

percentperyearonaverageintheEUbetween1996and2007butdeceleratedto0.1percent

aftertheGFC.Asaresult,overtheentiresampleperiodfrom1996to2020,averageTFPgrowthstoodat0.5percent,with11outof28EUcountriespresentinganegativeTFPgrowth.WealsoobservethatTFPgrowthinaccommodationandfoodservices—thelowestproductivitysector—was1.6percentagepointslowerthantheaverageTFPgrowthacrossallsectorsduringthe

sampleperiod,whileTFPgrowthinagriculture—thehighestproductivitysector—was1.4

percentagepointshigherthantheaverageTFPgrowth.Thereisalsosignificantheterogeneityinthetechnologicalgap.Withanaverageof-19.2percent,itvariesfrom-67.7percentto0percent.

Withregardstoindustry-levelfactors,weobservesimilarlylargevariationacrosssectors(Figure3).ICTcapitalspendingaveraged12.6percentofgrossfixedinvestment,withaminimumof0

percentandamaximumof44.6percent,whilenon-ICTcapitalspendingvariedfromaminimumof0.3percentofgrossfixedinvestmenttoamaximumof32.6percent,withanaverageof8.3

percent.Likewise,R&Dexpenditureasashareofgrossfixedinvestmentamountedtoanaverageof9.5percent,withaminimumof0percentandamaximumof47.8percent.

Table1.DescriptiveStatistics

Variable

Obs

Mean

SD

Min.

Max.

Industry-level

TFPgrowth

8,438

0.5

7.0

-50.0

100.0

Technologicalgap

9,151

-19.2

16.8

-67.7

0.0

ICTcapitalspending

11,879

0.1

0.1

0.0

0.4

Non-ICTcapitalspending

14,612

0.1

0.1

0.0

0.3

R&Dspending

12,315

0.1

0.1

0.0

0.5

Country-level

RealGDPpercapita

18,021

37873.8

18718.5

9544.1

122170.6

Inflation

17,969

5.3

41.3

-1.7

1061.2

Financialdevelopment

14,369

98.5

81.8

0.2

524.6

Tradeopenness

18,021

1.1

1.0

0.0

14.0

Bureaucraticquality

17,633

3.3

0.8

1.0

4.0

Population

18,021

18.1

22.8

0.4

83.2

Source:EU-KLEMS;IMF;WorldBank;ICRG;andauthors'calculations.

9

Figure3.Industry-LevelDevelopments

AverageTFPGrowth:Pre-GFC

AverageTFPGrowth:Post-GFC

(Percent)

Agriculture

FinanceandInsuranceManufacturing

WholesaleandretailtradeICT

Energy,waterandwasteTotal

Miningandquarrying

TransportationandstorageConstruction

Publicadmin,defenseandhealthAccomodationandfoodArtsandrecreation

ProfessionalandsupportserviceRealestate

(Percent)

ManufacturingAgriculture

WholesaleandretailtradeICT

Total

Professionalandsupportservice FinanceandInsurance TransportationandstoragePublicadmin,defenseandhealthRealestate

Construction Artsandrecreation MiningandquarryingEnergy,waterandwasteAccomodationandfood

-2-1.5-1-0.500.511.52

-2-1.5-1-0.500.511.52

MedianICTInvestmentbyIndustry

(Percent,measuredasashareoftotalinvestment)

MedianNon-ICTInvestmentbyIndustry

(Percent,measuredasashareoftotalinvestment)

ICT FinanceandinsuranceProfessionalandsupportservicesWholesaleandretailtrade

Artsandrecreation

Total

ManufacturingPublicadmin,defenseandhealth Accomodationandfood TransportationandstorageConstruction

Energy,waterandwasteMiningandquarryingAgriculture

Realestate

Transportationandstorage

ProfessionalandsupportserviceConstruction

WholesaleandretailtradeAgriculture

Total

AccomodationandfoodFinanceandInsuranceMiningandquarryingArtsandrecreation

Publicadmin,defenseandhealthManufacturing

Energy,waterandwasteICT

Realestate

010203040

AverageShareofHighSkillLaborbyIndustry

(Percent,measuredasshareoftotalemploymentineachindustry)

ICT

FinanceandInsurance

Realestate

TotalMiningandquarrying

ManufacturingWholesaleandretailtradeTransportationandstorage AccomodationandfoodConstruction

Agriculture

020406080

0204060

MedianR&DInvestmentbyIndustry

(Percent,measuredasashareoftotalinvestment)

ManufacturingProfessionalandsupportservicesPublicadmin,defenseandhealth

ICT

Total ArtsandrecreationWholesaleandretailtrade MiningandquarryingFinanceandInsurance

ConstructionEnergy,waterandwaste

AgricultureTransportationandstorageRealestate

Accomodationandfood

05101520

Sources:EUKLEMS;andauthors’calculations.

IV.ECONOMETRICMETHODOLOGY

Beyondglobalshocks,thereisamyriadoffactorscontributingtotheproductivityslowdownandcross-countryproductivitydifferentials.Inthispaper,wedevelopagranularanalysisbyusing

comparableindustry-leveldata—drawnfromtheEU-KLEMSdataset—andexplorethepatterns

andsourcesofTFPgrowthacrossthe28EUcountries.FollowingScarpettaandTressel(2002),

NicolettiandScarpetta(2003),Griffith,Redding,andVanReenen(2004),Acemoglu,Zilibotti,andAghion(2006),AghionandHowitt(2006),McMorrow,Werner,andTurrini(2010)andDabla-

10

Norrisetal.(2015),wemodelindustry-levelTFPgrowthusingthefollowingbaselinespecification:

+yj+μt+εijt

?Yijt=β0+β1?YLjt+β2(Yijt?1?YLjt?1)+βk∑kxt?1+βl∑kxjt?1(Yijt?1?YLjt?1)+ηi

inwhich?YijtisTFPgrowthincountryiandindustryjattimet.?YLjtdenotestheTFPgrowth

frontierintheEU,whichismeasuredbythehighestlevelofTFPgrowthinindustryjattimet.

TheTFPgrowthfrontiercapturestheextenttowhichcountriesareinvolvedincomparable

scientificandtechnologicalinnovationastheleadercountryorbenefitingfromknowledge

spillovers.(Yijt?1?YLjt?1)isthetechnologicalgapdefinedastheTFPdifferenceincountryiandindustryjattimetwithrespecttotheEUfrontier(highestlevelofTFP)inindustryjattimet.ThisrelativedistancetothefrontierrepresentsthepotentialforincreasingTFPbyadoptingnew

productivity-enhancingtechnologies.xt?1isavectorofindustry-levelandcountry-level

variables.Industry-levelvariablesincludeICTcapitalspending,non-ICTcapitalspending,R&Dspending,andtheshareofhigh-skilledlabor,whilecountry-levelvariablesincluderealGDPper

capita,consumerpriceinflation,tradeopenness,domesticcredittotheprivatesector,population,andbureaucraticquality.

Wealsoexplorehowindustry-levelfactors(ICTandnon-ICTcapitalexpenditures,R&Dspendingandtheshareofhigh-skilledlabor)interactwiththetechnologicalgap.Theseinteractionterms

aredesignatedbyxjt?1(Yijt?1?YLjt?1)incountryiandindustryjattimet.Thecoefficientsηi,

yjandμtdenotethetime-invariantindustry-specificeffectsandthetimeeffectscontrollingforcommonshocksthatmayaffectTFPgrowthacrossallindustriesattimet,respectively.6The

inclusionoffixedeffectsalsohelpsaddressendogeneityconcernsarisingfromomittedvariablebias.εijtistheidiosyncraticerrorterm.Robuststandarderrorsareclusteredattheindustrylevel.

V.EMPIRICALEVIDENCE

Ourbaselineresults,presentedinTable2,provideaconsistentassessmentofindustry-levelTFPgrowthin28EUcountriesovertheperiod1995–2020.Wedisplaythespecificationwithcountryfixedeffectsincolumn[1]forthewholesampleofindustries,incolumn[2]fortradablesectors,andincolumn[3]fornon-tradablesectors.7Wereplacecountryfixedeffectswitharangeof

country-levelcontrolvariablesandpresenttheseestimationsincolumn[4]forthewholesampleofindustries,incolumn[5]fortradablesectors,andincolumn[6]fornon-tradablesectors.

Becausedataontheshareofhigh-skilledlaborareavailableonlyfrom2008,wepresentthat

6Countryfixedeffectsarenotincludedwhenthemodelincorporatescountry-levelcontrolvariables.Theresultsarenotsensitivetoreplacingcountryfixedeffectswithcountry-levelvariables,whichprovideadditional

information.Wealsoobtainbroadlysimilarresultswiththeinclusionofcountry-yearandcountry-industryfixedeffects.

7Tradablesectorsincludeagriculture,forestry,fishing,mining,quarrying,andmanufacturing;whilenon-tradablesectorsincludeconstruction,wholesaleandretailtrade,transportation,storage,accommodation,foodservice,

ICT,finance,insurance,realestate,professionals