牛津經(jīng)濟研究院-技術支出：在經(jīng)濟衰退中推動增長和韌性（英）

OXFORD

EC0N0MICS

TECHNOLOGY

SPENDING:

DRIVINGGROWTHANDRESILIENCETHROUGHECONOMICDOWNTURNS

inassociationwith

SEPTEMBER2024

OXFORD

ECON0MICS

Tableofcontents

CONTENTS

EXECUTIVESUMMARY 4

Survivingandthrivinginthenextdownturn:

Therighttechnologyinvestmentspayoff 4

Keyfindings 5

SECTION1:Introduction 7

Box1:Auniquetechnologyspendingdataset 9

SECTION2:Evolutionoftechspend 10

TechspendhasoutpacedGDPandotherinvestments 11

Innovativedisruptionhasfavoredcertaintypesoftech

spendmorethanothers 12

SECTION3:Techspendasadriverofbusinessoutcomes 14

Morethanjustastrongcorrelationbetweentechspend

andrevenuegrowthinnormaltimesandduringdownturns 15

Techspendhasbeenimportantforresilienceandrecovery 15

Box2:Reviewofthedynamicpanelmodelusedinthisstudy 16

Datadeep-diverevealsfourkeytakeawaysfor

decisionmakers 18

SECTION4:Conclusion 22

SECTION5:Areasforfurtherinvestigation 24

Appendix1:Techspendduringrecessions 28

Appendix2:Industry-specificanalysis 31

Appendix3:Aboutoureconometricanalysis 37

ABOUTOXFORDECONOMICS 38

3

4

OXFORD

ECON0MICS

EXECUTIVESUMMARY

SURVIVINGANDTHRIVINGINTHENEXTDOWNTURN:

THERIGHTTECHNOLOGYINVESTMENTSPAYOFF

Howorganizationsinvestintechnologyduringeconomicdownturnscanbeastrongindicatoroftheirperformanceduringhardtimesandtherecoveriesthatfollow.

Conventionalwisdomholdsthatcostcuttingisanessentialwaytosurvivearecession,andwhilethatmayholdtrueinsomeareas,newresearchfromEYandOxfordEconomicsshowsthat

strategicinvestmentsintechnologyduringa

downturnmaybetheleadingstrategyfor

survivingarecession—andthrivinginarecovery.

Thatdoesnotmeanorganizationsshouldblindlythrowmoneyattechnology—asmanycompanieslearnedduringtheCovid-19era,itispossibletooverspendevenwhenrealopportunitiesexist.

Muchofthefrenziedbuyingduringtheearly

2020sresultedinaglutoftechnologyassets

thatbroughtcostlycomplexitytocomputing

environmentsandcreatedatechhangover

thathashadanegativeeffectonatleastshort-termspending.

Nordoesspendingthrougharecessionmeanthatthereisasinglecorrectwaytoapproach

technologyinvestmentswhentheeconomy

goessouth.Onekeytoasuccessfulstrategyisunderstandingthenatureofthesituation—notalldownturnsarecreatedequal.Overthelastthreedecades,eacheconomiccrisishashadunique

causesandeffects,anddisparateimpacts,onparticularindustries.Thereisnoone-size-fits-allmodelfortechnologyspending.

WhenEYandOxfordEconomicsteamedupforthisresearch,wefoundlittleempiricalevidenceoftechnologyinvestmentstrategiesduring

downturnsthatcouldenablerobustguidancetocorporatedecisionmakers.Muchoftheexistinganalysisfocusesinsteadonindividualbusinesscasesorspecificproducts.

Againstthisbackground,EYandOxford

Economicscompiledasignificantnewdatasetoftechnologyspendingfrompubliclyavailableindustrydatatoextractkeylessonsfor

investmentstrategiesacrossabroadspectrum

ofindustriesandproductcategoriesinadvancedeconomies,usingeconometricanalysiscombinedwithinsightsfromacademicliteratureand

businessleaders.

Onekeytoasuccessfulstrategyis

understandingthenatureofthesituation

—notalldownturnsarecreatedequal.

Impactonrevenuegrowth

(percentagepoints,2yravg)

KEYFINDINGS

Ourresearchshowsthat,inbroadterms,industries

thatspendwiselyontechnologythrougheconomic

downturnsemergestrongerandmorecompetitive

whileavoidingtheworstofthetechnologyhypecycle.

Amongthecoremessages:

Recentanalysisofmacrotrendsconfirmsthe

increasingsignificanceoftechnologyspendinginalleconomicsectors.Thisisevidencedby

businessescommittinganincreasingshareof

revenuestotechspendovertime.Thesurgeis

morepronouncedinservicesindustries,where

increasedtechspendhasbecomeacrucialdriverofeconomicactivity.Comparedtoproducing

industries,servicesectorsarewitnessingamore

substantialproductivityboostfromtechnologicaladvancements.Importantly,newerstudiestendtofindahigherimportanceofsoftwareandcloud-

relatedproducts(preferredbyservicesindustries),whileolderstudiestendtofindabiggerimpact

ofhardwareandcommunicationstechnologies.

Techspendingcansupportrevenueresilience

duringadownturn.Onaverage,revenuesshrankby3.5percentagepointslessinindustriesenteringthe2008globalfinancialcrisis(GFC)inthetop

quartileintermsofrelativetechspend.AndduringCovid-19,industriesinthemiddleofthepack

averagedrevenuelosses8pptslargerthanthoseatthetopoftechspending.Notably,regressionresultssuggestthatshiftsinthecompositionoftechspendplayanimportantroleindownturn

resilience.Inthenearterm,reducingthelevelofoperationalexpenditures(opex)relativetorevenueshasbeenconsiderablyharmfulto

revenuegrowthduringdownturns,suggestingthatthiscouldbeasub-optimalstrategyfor

protectingprofitabilityintimesofcrisis.By

contrast,changesintechcapitalexpenditures

(capex)hadlimitedimpactsongrowth,flaggingthatmanyindustriesmayhaveoverspentduringtheGFCandCovid-19crisis.

Chart1:Two-yearavg.annualgrowthgapinindustrieswith1pptlowertechopexspendasashareofrevenues

0

-0.1

-0.2

-0.3

-0.4

COVID

downturns

Baseimpact

Additionalrecessionspeci?cimpactSource:OxfordEconomics’estimates

Thebarsinthechartillustratethe

estimatedgapsinrevenuegrowth

betweenindustrieswithdifferentlevelsoftechopexacrosstheentiresampleandduringdownturns.

Thegraybarsshowthegapovertheentiresample,andtheredbarsshowhowmuchbiggerthegapbecomesduringadownturn.Theseregressioncoefficientsessentiallyquantify

averageoutcomedifferencesacrossindustriesandcountries.Technologyexpendituresarenormalizedand

analyzedasashareofrevenues.

Toaccountforsector-specific

performance,thesefigureswere

estimatedusingOxfordEconomics’dynamicpanelmodeleconometricframeworkdiscussedinBox2andinthetechnicalappendix.

5

Last3

6

OXFORD

ECON0MICS

Techspendingfuelsrecovery.Investment

intechbeforeandduringadownturncan

domorethanjustkeepbusinessesafloat;it

cancarrythemthroughandhastenrecovery.Excludingheavymanufacturersandpublic

services,averagepost-downturnrevenue

growthwithinaparentsectorwasbetween

1.2and3.8pptshigherinindustriesthat

maintainedarelativelyhighleveloftech

spendduringtheGFCrecession.1Onaverage,industriesthatgointodownturnswithlarger

sharesoftechspendalsoappeartobebetterplacedtoseizegrowthopportunitiesaftera

recession.Ourregressionresultsalsopoint

toaconsistentleveloftechopex(pre-and

duringadownturn)askeyfactorsforpost-

downturncompetitiveness.Bycontrast,capitalexpenditures(capex)intechwerenotfound

tohaveanimmediateimpactonresilience,butinsomeindustriestherewereindicationstheycouldbeimportantdriversofgrowth24-36

monthsahead,especiallyafterrecessions.

Knowyourdownturn,andtargettechnologybasedonspecificneeds.Differentlessons

applyindifferenttypesofdownturns,

whetheratechstockmarketcrash,aglobalfinancialcrisis,orpandemic-drivenshutdownoftheworldeconomy.Althoughspending

throughadownturnmaybecomethenew

normal,thatdoesnottranslateintothrowingmoneyattechnology.Understandingthe

macroeconomiccontext,includingpoliticalandsocio-culturaltrendsaswellasthe

sourceoftheshock,iscriticaltomaking

therightinvestmentdecisions.Afinancialcrisisthatreducescreditavailabilityimpliesadifferentsetoftechnologicalinvestmentsthanapandemicthatcausespeopleto

stayhomeandsociallydistance.Further,thosesameshocksimplyadifferentsetofinvestmentsforcompaniesoperatingindifferentsectors.

Afinancialcrisisthatreduces creditavailabilityimpliesadifferentsetoftechnologicalinvestmentsthanapandemic thatcausespeopletostayhomeandsociallydistance.

1Forcomparisonpurposes,thegranularindustriescoveredinthisstudyaregroupedintosixparentsectors:privateservices,publicservices,retailandwholesaletrade,transportationandstorage,heavymanufacturing,andlightmanufacturing.

SECTION1:

INTRODUCTION

7

OXFORD

ECON0MICS

Section1:Introduction

INTRODUCTION

Althoughthereisawealthofresearchdemonstratingthepositiveandsignificanteffectsoftechnologyinvestmentsonproductivityandrevenuegrowth,2,3,4a

relativelycommonstrategyencouragescompaniestoreducetheirspendingontechnologytoprotectprofitabilityduringeconomicdownturns.

AsshowninChart2,thisisthemainreasonbusinessspendingontechhasexhibitedacyclicalpatternoverthepast25years,withsignificantfluctuationsingrowth.Thereisalsoevidenceofscarring,orlong-term

damage,totechnologyspendingfollowingeconomicdownturns.

Theanalyticalworkpresentedinthispaper

revealsthatthisapproachhasbeenshortsightedandhasultimatelyharmedlong-termbusiness

success.Industriesacrossdifferentcountries

haveexhibitedvaryingresponsestoeconomic

downturns,andourempiricalstudydemonstratesthesignificantimpactofthesedifferenceson

revenuegrowth.Inparticular,thegranularity

ofouruniquetechnologyspendingdataset

(describedinBox1)allowedustouncoveruniqueinsightsabouttechspendanditscomposition

onresilience,post-downturncompetitiveness,andinnovation.

Forexample,ourdatasetrevealsthat,onaverage,companiesingoods-producingindustrieshave

cuttheirtechnologyspendingthreetimes

morethanthoseinserviceindustriesduring

challengingeconomictimes.Aftercontrolling

forsector-specificperformanceconstraints,oureconometricresultsshowthatthishastypically

contributedtolowerrevenuegrowthduringthelastthreeglobalrecessionsandinsomeinstanceshasalsoconstrainedpost-recoveryperformance.

Chart2:Globaltechspendgrowth

%growth,constantprices

12

6

0

-6

-12

199820002002200420062008201020122014201620182020

Source:OxfordEconomics’estimates

2Stiroh,K(2005).ReassessingtheImpactofITintheProductionFunction:AMeta-AnalysisandSensitivityTests.Annalesd’économieetdeStatistiqueNo.79/80,pp.529-561.Availableat

/10.2307/20777587

3Spiezia,V(2012).ICTinvestmentsandproductivity:MeasuringthecontributionofICTStogrowth.OECDJournalofEconomicStudies,Vol.2012/1.Availableat

/10.1787/eco

_studies-2012-5k8xdhj4tv0t

8

4OECD(2015).TheImpactofR&DInvestmentonEconomicPerformance:AReviewoftheEconometricEvidence.OECDDirectorateforScience,TechnologyandInnovation.Availableat

/document/DSTI/EAS/STP/NESTI

(2015)8/en/pdf

Section1:Introduction

Howtechnologyinvestments

enablegrowthandcompetitiveness

Twokeyfactorshelpexplainthestrong

relationshipbetweentechnologyspendingandgrowth.

Technologyisadriverofinnovationandcompetitivenessinthemodernglobaleconomy:

Companiesthatinvestintechnologyare

morecapableofadaptingtochanging

marketconditions,enhancingtheirproductsandservices,securingmarketaccess,andstayingaheadoftheircompetition.

Reducingtechnologyspendingduringa

downturncanimpedeacompany’sability

toinnovateandcompete,ultimatelyresultingindecreasedprofitability.

Technologyhelpsmaintainandimproveoperationalefficiency:

Byinvestinginnewtechnologies,companiescanoptimizetheirprocesses,lowercosts,

andincreaseproductivity.

Reducingtechnologyspendingduringa

downturncanconsequentlyleadtodecreasedoperationalefficiency,resultinginhigher

costsandlowerprofitabilityovertime.

Ofcourse,itispossibletooverspendon

technology.Companiesmuststrikea

balancebetweeninvestingenoughtoremaincompetitiveandinnovative,whileavoiding

certaintypesoftechnologythatmaynot

provideasufficientreturnoninvestment.Itiscrucialforbusinessleaderstocarefullyassessthecompositionoftheirtechnologyspendingagainstrelevantindustrybenchmarks.Overall,lessonsfromourstudyareintendedtoserve

asaguidelineforexecutivestofindtheoptimalcourseofactiongivencompanycharacteristicsandwheretheyareinthebusinesscycle.

BOX1:AUNIQUETECHNOLOGYSPENDINGDATASET

ThisstudyisbasedonanoriginaldatasetoftechnologyspendingputtogetherbyOxfordEconomicsin2023toprovidevaluableinsightsintotheimpactofdifferenttypesoftechnologyinvestmentsoncompanyrevenuesand

resilienceduringeconomicslowdownsanddownturns.

Agloballyconsistent,industry-leveldatasetfora

sufficientlylargenumberofcountriesisessentialfor

hypothesistestingtodeductrobustclaimsregardingtheassociationoftechnologyspendingwithperformance.

Thedatasetincludesinformationfrom11countriesandupto35industriespercountry,spanningfrom1995to2021.

Fivecoreenterprisetechnologyitemsareconsideredalongwithtwoadditionaldriversofenterprise

value,whichweconsiderpartofthebroader

“techspend”umbrella.Thesearethefollowing:

Devices:Electronics,computers,communicationequipment

Enterprisesoftware:“Pre-packaged”software

ITservices:Customand“own-account”softwareandrelatedservices

Internetandcloudservices:Dataprocessing,internetportals,searchengines

Communicationservices:Wired,wireless,satellite

ResearchandDevelopment(R&D):Innovativeactivitiestogeneratenewproducts

Brandmanagement:Advertising,media,andpublicrelationsystems,artisticoriginals

Foreach,twoadditionalbreakoutsareincluded:

Capex(capitalexpenditure):Long-termexpenditurerelatedtoplansforfuturegrowthandefficiency,

thatarepurchasedforalifespanlongerthanoneyear(e.g.,purchasesofITequipment,enterprisesoftwarerequiredtorunproductionfacilities)

Opex(operatingexpenses):Day-to-dayexpenditurerelatedtoeachindustry’sproductionrequirementsthatarepurchasedforuseinmakingtheindustry’s

product(e.g.,annualsoftwaresubscription,semiconductorsinstalledincars)

Thedataiscompiledfrompubliclyavailablesources,primarilyfromorganizationssuchastheOECD,theUSBureauofEconomicAnalysis,andtheAustralianBureauofStatistics.

9

SECTION2:

EVOLUTIONOFTECHSPEND

10

11

Section2:Evolutionoftechspend

OXFORD

ECON0MICS

Abroadreviewofexistingliteratureandabrieflookatourtechnology

spendingdatasetshowshowtechspendhasevolvedoverthepast25years,withparticularfocusoneconomicdownturns.

KEYFINDINGSANDTAKEAWAYS

Datareflectsthattechspendhasbecomeevermoreimportantacrossallsectorsoftheeconomy,outpacingrevenuegrowthovertime—especiallyforcapex.

Therisehasbeenparticularlynotableforserviceindustries,whichmayseealargerproductivitybenefitfromtechcomparedtoproducingindustries.

Techspendhasrespondeddifferentlyduringdifferentdownturns.Afteraconsiderabletechhangoverintheearly2000s,thebeliefthattechboostsproductivityandprofitabilityhas

drivenstrategyevenduringdownturns,withaspendingpeakduringtheCovid-19pandemic.

TECHSPENDHASOUTPACEDGDPANDOTHERINVESTMENTS

Thedigitalrevolutionhassupporteda

considerableriseoftechnology-related

expenditureoverthepast25years,outpacing

bothtotalinvestmentandGDP(Chart3)—hinting

Chart3:WorldGDPandinvestment

attherisingimportancebusinessesassigntotechspendingasacatalystforgrowth.Therisehas

beenparticularlynotableforserviceindustries,whichmayseealargerproductivitybenefit

Index(1997=100),in?ationadjusted

300

200

100

GDP

TotalInvestmentTechspend

0

199720002003200620092012201520182021

Source:OxfordEconomics’estimates

12

OXFORD

ECON0MICS

Section2:Evolutionoftechspend

fromtechcomparedtoproducingindustries

(Chart4)—therelativeshareoftechspendasa

percentageofrevenueswentfrom6%before

2000toover9%afterthepandemic.

Chart4:Techspend(percentofrevenues)

%%%

%%%

HeavymanufacturingLightmanufacturing

4

3

6

4

2

1

0

2

0

20002005201020152020

20002005201020152020

PrivateservicesPublicservices

12

4

8

2

4

0

0

2000200520102015202020002005201020152020

TradesTransportationandstorage

4

3

2

2

1

0

0

2000200520102015202020002005201020152020

Source:OxfordEconomics’estimates

INNOVATIVEDISRUPTIONHASFAVOREDCERTAINTYPESOFTECHSPENDMORETHANOTHERS

Theshareofrevenuesthatindustriesspendon

enterprisesoftwarehasgrownsix-foldsince1997,reachingaglobalaverageof0.6%in2021.This

reflectstheincreasingrelianceofbusinessesonsoftwaresolutionstoenhancetheirproductivity,innovation,andcompetitiveness.However,

softwareisnottheonlycategorythathasseen

asignificantincreaseinspending.Internetand

cloudservices,whichincludedatastorage,

processing,andanalytics,havealsogrown

rapidly,especiallyinthelastfewyears.Infact,theshareofrevenuesspentoninternetandcloud

serviceshasalmosttripledsincetheGFC,andisnowclosetomatchingsoftwarespending(seeChart5).Bycontrast,communicationserviceshaveexperiencedasteadydecline.Theshift

towardmobiledevicesandtheincreaseduseof

cloud-basedservices(e.g.,VoIP)havelikelyhurtcommunicationservices;however,asweexploreinthisstudy,communicationservicesarestillanimportantdriverofrevenuegrowth.

Whatexplainsthisshiftintechnologyspending?Oneofthekeyfactorsistherisingimportance

ofintangibleassets,suchasdataandcomputerservers,inbusinessoperations.Theseassets

enablebusinessestoleveragethepowerof

cloudcomputing,artificialintelligence,and

otheremergingtechnologiestocreatevalueandgaincompetitiveadvantage.Oureconometric

analysis,whichisavailableintheappendixof

ourreport,providesempiricalevidenceforthisrelationship.However,thedataweuseinour

studydoesnotfullycapturethecomplexityand

13

Section2:Evolutionoftechspend

OXFORD

ECON0MICS

dynamismofthetechnologylandscapebecausestandardindustryclassificationsarebasedon

historicaldefinitionsusedbystatisticalagenciesandmaynotreflectthecurrentandfuture

realitiesoftechnologyusage.Forexample,cloudservicesarearelativelynewphenomenonthat

spanacrossmultiplecategories,suchassoftware,

ITservices,andinternetservices.Assuch,they

maynotbeadequatelyrepresentedbyanysinglecategoryinourdata.Therefore,werecommendthatmoregranularandtimelydatasourcesbe

developedandusedtobetterunderstandthestrategicimplicationsofshiftingfromonetypeoftechnologytoanother.

Chart5:Globaltechspendevolutionbytype(percentofrevenues)

SoftwareDevicesR&DCommunicationservices

6.01.50.75

%

4.0

2.0

0.0

%

1.0

0.5

1.0

0.50

0.25

0.00

%

%

0.5

0.0

0.0

200020102020200020102020200020102020200020102020

ITservicesInternet&cloudAdvertising

Source:OxfordEconomics’estimates

%

1.5

1.0

0.5

%

0.4

%

0.80.60.4

0.2

0.0

200020102020

0.30.2

0.1

0.0200020102020

0.0200020102020

Howrecessionsshapedtechspend

Asmightbeexpected,thenatureofthedownturnaffectedtechspend—bothinthevolumeofspendandthetypesoftechnologytargeted.

Duringthedotcomcrisis,ITwasviewed

withskepticismbyexecutiveswhocut

theirtechnologybudgetstosurvivethe

downturn.Techspendingenteredaperiodofausterity.However,by2008,executivesbegantorealizethattechnologywasno

longera“nicetohave”butanecessity.

Thetypesoftechspendcutshavealso

variedsignificantlyaftereachofthepast

tworecessions.Forexample,globaldevices

spendtookahitduringtheGFCbut

wereboostedduringtheCovidpandemic,whichalsosawariseininvestments

indigitalinfrastructure.

Ingeneral,industrieshavemovedawayfromoperationalexpenditure(opex)software

spendduringeverydownturn.Instead,theyhavetakenadvantageofthesituationto

investinmorelong-termsolutions,suchassoftwarecapitalexpenditure(capex).

Formoredetailsonspendingduringvariouseconomicdownturns,seeAppendix1.

SECTION3:

TECHSPENDASADRIVEROFBUSINESSOUTCOMES

14

Section3:Techspendasadriverofbusinessoutcomes

OXFORD

ECON0MICS

KEYFINDINGSANDTAKEAWAYS

Fromcorrelationanalysis,wefindthatsectorsthatspendheavilyontechappearto

experiencegreatergrowth,especiallyinthelongterm.Keepingtechspendaconsistentlyhighshareofthetotalthusseemscrucialforsuccessaswellasresilience.

DuringCovid-19,industriesinthemiddleofthepackaveragedmuchlargerrevenuelossescomparedtothoseatthetop.

Weverifytheseobservationsviaeconometricanalysisusingadynamicpanelmodelandcontrollingfortheinfluenceofothersourcesofrevenue.

Wefindthatmaintainingorevenraisingopexisparticularlyimportantforsupporting

revenuegrowthindownturns.Thelargestandmostsignificantimpactisfoundforthe

Covid-19downturn.

OurestimatessuggestthatmanyindustriesinmanufacturingandthepublicandprivateservicessectorsunderspentintechopexduringtheGFCdownturnandoverspentin

techcapexduringtheCovid-19pandemic.

MORETHANJUSTASTRONGCORRELATIONBETWEENTECHSPEND

ANDREVENUEGROWTHINNORMALTIMESANDDURINGDOWNTURNS

Ouranalysisrevealsanuanced,multifaceted

connectionbetweentechnologyadoption

andrevenuegrowthacrossdifferentindustries.Whileapositivecorrelationisevident(asshowninChart6),oureconometricanalysisuncoversadeeperdynamic(seemainresultssection

fordetails).Weobserveavirtuouscyclewhere

acceleratedrevenuegrowthfuelstechnologyinvestment,whichinturnamplifiesbusinessperformance.Thiseffectisparticularlystrongwhencomparingcompanieswithinthesameindustrygroup,especiallythosededicating

alargershareofcoststotechnology.

Chart6:Techspendasadriverofrevenuegrowth5

Avgrevenuegrowth(1997-2021),

in?ationadjusted(%)

5

4

3

2

1

0

Techmaturity

HighLow

Heavy

manufacturing

Light

manufacturing

Privateservices

Publicservices

TradesTransportationandstorage

Source:OxfordEconomics’estimates

5Hightechmaturityincludesindustry-countryobservationswithanabovemedianlevelofteachspendrelativetorevenues.15

BOX2:AREVIEWOFTHEDYNAMIC

PANELMODELUSEDINTHISSTUDY

Adynamicpanelmodel(DPM)isatypeofregressionmodelthathasbecomethestandardtooleconomistsusetounderstandtheeffectsofbothtimeandindividualcharacteristicsonadepen-

dentvariable.Inthiscase,aDPMisidealtodirectlyestimatehowrevenuegrowthhasbeeninfluencedbychangesintechnology

expenditures,itsownpastvalues,aswellasbyotherexplanatoryvariables,suchasthesizeoftheindustryandexposureto

large-scaleeconomicdownturns(i.e.,globalrecessions).

However,estimatingadynamicpanelmodelaccuratelyisnot

straightforwardbecauseitinvolvessometechnicalchallenges.Oneofthemistheproblemofendogeneity,whichmeansthatsomeoftheexplanatoryvariablesmaybecorrelatedwithwhatisleft“unexplained”bytheregressionequation.Thiscanbias

theestimatesofthecoefficientsandleadtowrongconclusions.Severalstepsweretakentoaddressthisproblem:

First,thevariablesincludedintheregressionsweretransformedtoremovethecommoninfluenceoftimetrends.Inaddition,

thedependentvariableandselectregressorswereexpressedinrelativetermstoensurecomparabilityacrossperiodsand

countries(e.g.,techinvestmentasashareofindustryrevenues).

Additionally,laggedvariablevalueswereincludedasinstrumentsbecausethesewerefoundtobecorrelatedwiththeexplanatoryvariables,butnotwiththe“errorterm.”Theseinstrumentswereassumedtobepredeterminedandexogenousandwereused

toprojecttheendogenousvariablesintheregressionequationtoprovideconsistentestimatesofthecoefficients.

Onechallengethatariseswhenusinglaggedvaluesasinstrumentsisthatlaggedvaluesmaynotbesufficientlycorrelatedwiththe

currentvaluesoftheexplanatoryvariables,especiallyifth