unfairwp在线

WP/08/4DoesTechnologicalDiffusionExplainAustralia'sProductivityPerformanceThierryTressel2008InternationalMonetaryFundWP/08/4IMFWorkingPaperResearchDepartmentDoesTechnologicalDiffusionExplainAustralia'sProductivityPerformancePreparedbyThierryTressel1AuthorizedfordistributionbyMasahikoTakedaandAlessandroPratiJanuary2008AbstractThisWorkingPapershouldnotbereportedasrepresentingtheviewsoftheIMF.
TheviewsexpressedinthisWorkingPaperarethoseoftheauthor(s)anddonotnecessarilyrepresentthoseoftheIMForIMFpolicy.
WorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtofurtherdebate.
Thispaperanalyzestheimpactofproductandlabormarketpoliciesontechnologicaldiffusionandmulti-factorproductivity(MFP)inapanelofindustriesin15OECDcountriesovertheperiod1980to2003,withaspecialfocusonAustralia.
Weuseasimpleconvergenceempiricalframeworktoshowthat,onaverage,convergenceofMFPwithinindustriesacrosscountrieshasslowed-downinthe1990s.
Incontrast,Australianindustrieshavesignificantlycaught-upwithindustryproductivitybestpracticesoverthepast16years,andhavebenefitedfromthediffusionofInformationandCommunicationTechnologies(ICTs).
Weshowthatreformsofboththelaborandproductmarketssincetheearly1990scanexplainAustralia'sproductivityperformanceandadoptionofICTs.
JELClassificationNumbers:D24,J08,J24,K2,L6,L8,L9,O3.
Keywords:Australia,Productivity,TechnologyDiffusion,LaborMarket,ProductMarketAuthor'sE-MailAddress:ttressel@imf.
org1TheauthorwouldliketothankDavidCoe,MasahikoTakeda,HaliEdisonandparticipantsataseminarattheAustralianTreasuryduringtheArticleIVconsultationoftheIMF(June2007)forusefulremarksandsuggestions.
ManzoorGillandJoseRomeroprovidedexcellentresearchassistance.
2ContentsPageI.
Introduction4II.
ProductivityPerformanceandReformsinAustralia7A.
AustralianProductivityPerformancesince1990:KeyFacts7B.
TheReformProcessinAustralia9III.
EmpiricalModelandDataDescription10A.
Theory.
10B.
EmpiricalSpecification.
11C.
TheLong-RunImpactofCovariatesXonMFPandCapital-LaborRatioLevels.
.
.
12D.
DataSourcesandMethodology.
13E.
GrowthAccounting14F.
LevelAccounting.
15G.
ExchangeRatesforInternationalComparisons.
16IV.
EmpiricalAnalysis.
16A.
AFirstLookattheData.
16B.
RegressionResults18C.
RobustnessTests20D.
DoReformsExplainAustralia'sProductivityPerformance20E.
OtherDeterminantsofMFPConvergence:theRoleofHumanCapitalandR&D.
.
21V.
Conclusion23References.
24Tables1.
SectoralDecomposition.
132.
SummaryStatistics(1980–2003)283.
Correlations.
294.
Multi-FactorProductivityLeaders.
305.
ConvergenceofAustralianIndustries'TechnologyLevelinaPanelofOECDCountries.
326.
ImpactofProductMarketRegulationsofMFPGrowth347.
ImpactofLaborMarketInstitutionsonMFPGrowth.
358.
DisentanglingtheEffectsofProductandLaborMarketInstitutionsonMFPGrowth369.
ImpactofLaborandProductMarketInstitutionsonICTCapitalDeepening3710.
PredictedImpactofProductandLaborMarketReforms.
3811.
ControllingforOtherDeterminantsofMFPGrowth393Figures1.
Australia'sProductivityPerformance72.
InvestmentsinInformationandCommunicationTechnologies.
83.
SectoralContributionstoRealGDPGrowth.
84.
EmploymentProtectionLegislationsinOECDCountries95.
ProductMarketReformsinAustralia.
106.
AustralianIndustriesMFPLevels(relativetoU.
S.
31AppendixesI.
TheImpactofProductandLaborMarketInstitutionsonLaborProductivity40II.
DroppingCountriesOnebyOne41III.
Regressionswith3-yearAverages.
424I.
INTRODUCTION"LessonsfromAppleCo.
:notinventedhere…andverywelcome"TheEconomist,June2007Sincetheendofthe1991recession,Australiahasexperiencedalonglastingeconomicexpansionduringwhichlaborproductivityhasincreasedbycloseto30percent,almostthesamefeatastheonerealizedbytheU.
S.
economyoverthesameperiod.
2Moreover,theInformationandCommunicationTechnology(ICT)revolution,whichhasbeenbehindtheproductivityrevivalintheU.
S.
,seemstohavediffusedaroundtheglobetoAustraliaaswell.
3Overthepast16years,Australiahasalsoundertakenmanyreforms,inparticularofthelaborandproductmarkets,andhasfollowedsoundandprudentmonetaryandfiscalpolicies,whichhaveenhancedtheflexibilityoftheeconomyanditsresiliencetoshocks.
WhatexplainsthissuccessstoryTheobjectiveofthispaperistolookatAustralia'sproductivityperformancefromtheperspectiveofproductivitybestpracticesdiffusionacrossOECDcountries.
Welookatthediffusionoftechnologyfromtwoangles:(i)multi-factorproductivityand(ii)thediffusionofICTcapital,andassesstheroleofpoliciesinfosteringdiffusion.
Morespecifically,weaskwhetherAustralia'sreformsexplainthesustainedproductivityaccelerationobservedsince1991.
Moregenerally,thispaperprovidesnewevidenceontheinteractionbetweentechnologydiffusionandpoliciesinOECDcountries.
Wefindthatproductmarketreformshaveasignificantandpositiveimpactonmulti-factorproductivity(MFP),inparticularinindustriesthatuseICTcapitalgoodsmoreintensively.
4However,themagnitudeoftheestimatedimpactisnotsufficienttoexplaintheproductivityaccelerationobservedintheseindustriesinAustralia.
Next,wefindthatlabormarketflexibilityseemstobeassociatedwithfasterproductivitygainsinindustriesthataremorehumancapitalintensive.
Themagnitudeoftheestimatedeffectislarge.
WealsofindthatcountrieswithmoreflexiblelabormarketshaveexperiencedafasterICTcapitaldeepening.
Theseresultssuggestthat,inthecaseofAustralia,acombinationoflaborandproductmarketreformscouldhaveinducedproductivitygainsoftheorderofmagnitudeofthoseobservedinthe1990s.
Finally,wefindthattechnologicaldiffusioncruciallydependsondomesticR&Dintensityandhumancapital,evenaftercontrollingforproductmarketandlabormarketpolicies,suggestingthatawiderangeoffactorsaffectthediffusionoftechnologybestpracticesacrossOECDcountries.
52DavisandRahman(2006)andDolmanetal.
(2007)provideanoverviewofAustralia'sproductivityperformanceandofitsdeterminants.
3SeeOECD(2003),Parhametal.
(2001),ProductivityCommission(2004),andCardarelli(2001a).
4Inthepaper,wewillusealternativelymulti-factorproductivity(MFP)ortotalfactorproductivity(TFP)tonameHicks-neutralproductivityeffectsthatarenotbiasedinfavorofanyfactorofproduction.
5WeusetheEU-KLEMSandIGAdatabaseoftheGGDC,whicharethebestexistingdatasetsprovidinginformationonproductivityacrossOECDcountriesandatadisaggregatedlevelwithabreak-downbetween(continued…)5AgrowingliteraturehastriedtoidentifythefactorsdrivingthediffusionoftechnologywithinindustriesacrossOECDcountries.
TheroleofdomesticR&DininnovationandinspeedinguptheadoptionofexistingtechnologieshasbeenemphasizedbyGriffithetal.
(2004),andCameronetal.
(2005).
AcharyaandKeller(2007)andCameronetal.
(2005)showthattechnologicaltransfers(includingR&Dspillovers)takeplacethroughinternationaltrade.
6Keller(2002)alsoshowsthatinternationalR&Dspilloversdependsnegativelyonthedistancebetweencountries.
7Vandenbussche,AghionandMeghir(2006)findthattertiaryeducationhasahighergrowth-enhancingeffect,attheaggregatelevel,theclosertothetechnologyfrontieracountryis.
Whilewefindlimitedsupportingevidenceinfavorofthetradechanneloftechnologicaldiffusioninoursample,weconfirmtheroleofR&Dandofhumancapitalattheindustrylevel.
TheroleofproductmarketpoliciesinspeedingupthediffusionofproductivitybestpracticesacrossOECDcountrieshasbeenemphasizedbyScarpettaandTressel(2002),NicolettiandScarpetta(2003),andConwayetal.
(2006).
8However,ScarpettaandTressel(2002)andNicolettiandScarpetta(2003)donotcoverthelate1990s,whichpreventsanalyzingtheimpactofreformsonthediffusionofICTgoodsacrossOECDcountries.
Conwayetal.
(2006)focusonlaborproductivity,anddonotexploretheroleoflabormarketinstitutions,humancapitalorR&Dinfosteringtechnologyadoption.
TheyalsodonotanalyzetheprocessofICTcapitaldeepeningattheindustrylevel.
Finally,theroleoflabormarketregulationswasemphasizedbyScarpettaandTressel(2002,2004),andfromamacroeconomicperspectivebyGustandMarquez(2004)andKentandSimon(2007).
9ThepaperisalsorelatedtotheliteratureonthesourceoftheproductivitysurgeintheUSinthesecondhalfofthe1990s(JorgensonandStiroh,2000,OlinerandSichel,2000).
Stiroh(2002)showsthatthemostICT-intensiveindustriesexperiencedsignificantlylargerproductivitygainsthanotherindustries.
HealsofindsastronglinkbetweenICTcapitalsharesandtherelativeaccelerationoflaborproductivity,e.
g.
theproductivitysurgeoccurredinindustriesthatproduceanduseinformationtechnologiesmoreintensively.
10Jorgenson,Hodifferenttypesofcapitalandlabor(seesectionIII.
D).
However,problemsremainwithproductivitydataandnationalstatisticalagenciesareintheprocessofgainingexperienceinthecompilationsofthesedata.
Thissuggeststhatestimatesmaybesubjecttoerrors(seeforinstanceAustralianBureauofStatistics,2007).
6CoeandHelpman(1995)firstemphasizedtheroleofinternationaltradeininternationalR&Dtransfers.
7Keller(2004)providesasurveyofthemicroandmacroliteratureonR&Dspillovers.
8GriffithandHarrison(2006)showedthatEUSingleMarketProgrammereformshaveincreasedcompetition,andsubsequentlyraisedinnovationintensityandproductivitygrowthformanufacturingsectors.
9Salgado(2002)alsoexplorestheimpactofstructuralreforms(trade,productmarketandlabormarket)onaggregateproductivitygrowth.
Hefindsnosignificanteffectintheshort-run,andsomesignificantpositiveeffectofproductmarketandtradereformsonproductivitygrowthinthelong-run.
10Stiroh(2006)findsthatthesecondproductivitysurgeintheU.
S.
(2000-04)ismorebroad-based,andthatnon-ICTindustriesalsoexperiencedproductivitygains.
6andStiroh(2005)findthat,bothattheindustryandaggregatelevel,economicgrowthintheU.
S.
over1977–2000wasexplainedbyinvestmentsininformationtechnologyandhighereducation.
Turningtointernationalcomparisons,Inklaaretal.
(2005)performedgrowthaccountingdecompositionsattheindustrylevelandfoundthatICTcapitaldeepeningtookplaceinservicesectorsbothintheU.
S.
andinfourEuropeancountries,butthatthecontributiontoaggregategrowthislowerinEurope.
Moreover,thesesectorsdidnotexperienceaTFPaccelerationinEuropeancountries.
Inklaaretal.
(2006)showthatdifferencesinproductivitygrowthandlevelsaretoalargeextentdrivenbymarketservices.
TheysuggestthatproductivitygrowthinservicesindustriesinCanadaandAustraliamayberelatedtoaprocessofconvergence(lowinitiallevelsofproductivity).
Inthispaper,wegoonestepfurtherbyexploringwhichfactorscouldexplainwhytheprocessofconvergenceacceleratedinAustraliapreciselyafter1990,andnotbefore.
Moregenerally,werelatethediffusionofthe"ICTrevolution"acrosscountriesandwithinindustriestoexistingpoliciesinthelaborandproductmarkets,andtoR&Dandhumancapital.
Whatarethechannelsthroughwhichlessstringentproductmarketpoliciesmayenhanceproductivitygrowth11Beyondthetraditionaltextbookargumentthatcompetitionbringsaboutallocativeefficiency,recenttheoriesandevidencehaveemphasizedtheimpactofcompetitiononinnovationinadynamicSchumpeterianenvironment.
InaSchumpeterianworld,firms'abilitytoinnovate,adoptnewtechnologiesandreorganizetheirproductionstructuredependsoncompetitivepressuresreflectedintheprocessofentryandexit.
Acemogluetal.
(2006)arguethatbenefitsfromcompetitionpoliciesarelargerinmoreadvancedeconomieswheninnovation-basedgrowthstrategiesbecomemoreimportant.
Aghionetal.
(2005)developamodelinwhichthereisaninvertedU-shaperelationshipbetweencompetitionandinnovation,andfindsupportforthetheoryfromUKfirmleveldata.
Aghionetal.
(2006)findapositiveimpactofentryonproductivityinmoretechnologicallyadvanced,lessmatureindustries.
Ourresultsonthelabormarketareconsistentwithanumberoftheories.
Forinstance,SaintPaul(2002)showsthatcountrieswitharigidlabormarketspecializeinlowrisk"secondaryinnovations"whichimproveexistingproducts,whilecountrieswithmoreflexiblelabormarketstendtoproducenewproducts.
Histheoryimpliesthatlabormarketrigidityismorelikelytoharminnovationandproductivityinmorehigh-skill,innovativeindustries.
Similarly,HopenhaynandRogerson(1993)andBertola(1994)showhowproductivitycanbereducedasemploymentprotectiondistortsemploymentawayfromanefficientallocation,therebyreducingcapitalaccumulationandgrowth.
12Allinall,existingtheoriessuggestthatlabormarketrigiditiesaremorelikelytohaveadverseeffectsonproductivityinindustriesthatdependmoreoninnovationandthecreationofnewgoods.
11SeealsoScarpettaandTressel(2002)foradiscussion.
12Therealsoexistsalineoftheoreticalliteratureshowingthatemploymentprotectionmayhavepositiveeffectsonproductivitybyencouragingfirmspecifichumancapitalaccumulationthroughlongerjobtenure(seeSaintPaul(1997)).
7Thepaperisorganizedasfollows.
SectionIIdescribesAustralia'sproductivityperformanceandthereformsundertakeninthe1990s.
SectionIIIpresentstheempiricalmodelandthedata.
SectionIVdiscussesourempiricalresults.
SectionVconcludes.
II.
PRODUCTIVITYPERFORMANCEANDREFORMSINAUSTRALIAA.
AustralianProductivityPerformancesince1990:KeyFactsSincetheendofthe1991recession,Australiahasexperiencedastrongproductivityperformance.
Aggregatelaborproductivityhasincreasedbyabout30percentbetween1990and2006,drivenbystrongproductivitygrowththroughoutmostofthe1990s,butfollowedbyaslow-downinrecentyears.
ThisoverallperformanceissimilartothatoftheUnitedStatesoverthesameperiodandbetterthanOECDaverage.
13Productivitygainswereaccompaniedbystronggrossfixedcapitalformationandincreasedlaborparticipationsothat,by2007,theunemploymentratehadfallentoa32yearslow.
Duringthisperiod,AustraliahasalsostronglyincreasedinvestmentsinInformationandCommunicationTechnologies(ICT).
Figure1.
Australia'sProductivityPerformanceLaborProductivity(1991=100)AverageAverage(excl.
Nordiccountries)UnitedStatesAustralia10010511011512012513019911993199519971999200120032005LaborProductivitydefinedasRealGDP/EmploymentSource:WorldEconomicOutlook200713TheaverageinFigure1includesallofthe15countriesusedintheregressionanalysis.
Nordiccountries(Denmark,FinlandandSweden)experiencedthestrongestproductivitygains,followedbysimilarperformancesfortheU.
S.
,theU.
K.
,andAustralia.
8Figure2.
InvestmentsinInformationandCommunicationTechnologiesICTinvestment(inpercentoftotalinvestment)AllindustriesMarketservices1/0%5%10%15%20%25%19601963196619691972197519781981198419871990199319961999200220051/:wholesaletrade,retailtrade,businessservicesandtransportSource:AustralianBureauofStatisticsLaborproductivitygainswereunevenlydistributedacrosssectors,andweremainlylocatedinmarketservices,includingwholesaletradeandretailtrade,aswellasinbusinessservicesandtransportsinmorerecentyears.
14Hence,fromasupply-sidepointofview,marketserviceshavebeenthemaindriversofrealGDPgrowthoverthepast16years.
Othersectorsexperiencedmixedperformance:manufacturing'scontributionhasbeendecliningsteadily,whileagricultureandmining'soutputgrowthandproductivityperformancefollowedcyclicalpatternswithnocleartrend.
Figure3.
SectoralContributionstoRealGDPGrowthAgricultureMiningManufacturingMarketservices1/RealGDPgrowthOtherservices2/-2%-1%1%2%3%4%5%6%199119931995199719992001200320051/Wholesaletrade,retailtrade,communication&businessservices,transport&storage,finance&insurance2/Includesnon-marketservicesSource:Author'scalculations,AustralianBureauofStatistics14Author'scalculations.
SeealsoInklaaretal.
(2006)forgrowthaccounting.
9B.
TheReformProcessinAustraliaSincethelate1980s,Australiahasundertakenbroadandfarreachingreformsintheareasoftheproductandlabormarketswhichhaveenhancedthecapacityoftheeconomytorespondflexiblytoshocks.
Australiahasalsofollowedsoundmonetaryandfiscalpolicies.
Intheareaofthelabormarket,therehasbeenalong-termtrendofindustrialrelationsreforms.
Bytheendofthe1980s,thecostsassociatedwiththecentralizedsystemofwagebargainingandsetting,wherebywageincreaseswerepassedonacrosstheboardtoallindustries,hadbecameapparent.
Labourmarketreformsundertakenduringtheperiodofthisanalysisincludedthefollowingsteps.
In1991,undertheKeatinggovernment,bargainingagreementsonemploymentconditionsweredecentralizedattheenterpriselevelsoastobetteralignwagestoproductivitygains.
The1993reformsexpandedandacceleratedtheuseofenterprisebargaining,whilein1996,theCoalitionGovernmentfurtheredthedecentralizationprocessbyintroducingthepossibilityofindividualcontractsthroughtheAustralianWorkplaceAgreements(thepowersoftheAustralianIndustrialRelationsCommissionwerealsoreduced).
Thesesteps,togetherwithotherreformsandongoingstructuralchangesofindustrializedeconomiesaroundtheworld,contributedintheseculardeclineintheunionizationrate.
15Asaresultoftheseconsecutivereforms,AustraliastandsoutamongOECDcountriesasacountrywithaflexiblelabormarketcharacterizedbyalowlevelofcorporatism,adecentralizedwagebargainingsystem,andaflexibleemploymentprotectionlegislation.
Figure4.
EmploymentProtectionLegislationsinOECDCountriesStrictnessofEmploymentProtectionLegislation00.
511.
522.
533.
544.
5UnitedStatesUnitedKingdomCanadaIrelandAustraliaDenmarkFinlandAustriaJapanBelgiumKoreaNetherlandsFranceNorwaySwedenGermanySpainGreecePortugalItalySource:Nicolettietal.
(2001)Index15Whileoutsidetheperiodofthisanalysis,the2006"Workchoices"reformsaimedatfurthersimplifyingworkplaceagreementproceduresbyformingasinglenationalindustrialsystem,reducingthenumberofminimumemploymentconditions,andliberalizingunfairdismissallaws.
10Australiahasalsosignificantlyderegulateditsproductmarketssincetheendofthe1980s.
16Significanttradeliberalizationwasinitiatedinthe1980s.
Startingin1988,phasedreductionsintariffswereimplementedacrossallindustries,sothatbytheendofthe1990svirtuallyalltariffshavebecomenegligible.
Infrastructurereformswerealsoinitiatedinthelate1980s,coveringderegulationandrestructuringofairandcoastaltransport,andtelecommunication.
Publicenterpriseswerealsoprogressivelycommercialized,corporatizedandprivatizedonalargescale.
Between1995and2000,theNationalCompetitionPolicy(NCP)furtherreducedanti-competitiveregulations,andreformedgovernmentbusinessesandthetransportandutilitiessectors.
Goingforward,theNationalReformAgendawillcontinue(i)reducinganti-competitivebarriers(energy,transport,infrastructure);(ii)simplifyingregulations;and(iii)enhancinghumancapital.
Figure5.
ProductMarketReformsinAustraliaIndexofProductMarketRegulation(OECDRegulatoryImpactIndicator)AustraliaUnitedStatesAverage0.
080.
10.
120.
140.
160.
180.
20.
22198019821984198619881990199219941996199820002002Source:ConwayandNicoletti(2006)III.
EMPIRICALMODELANDDATADESCRIPTIONAcloselookatthedatashowsthatproductivityperformanceusuallyvariessignificantlyacrossindustriesintheOECD.
Moreover,reallocationofresourcesacrossindustrieshastypicallyplayedalimitedroleinexplainingcross-countrydifferencesinaggregateproductivitygrowthinOECDcountries.
Forthesetworeasons,werestrictouranalysistoproductivityperformanceattheindustrylevel.
A.
TheoryConsideraconstantreturntoscaleproductionfunctionforindustryjincountryi:16SeeBanks(2005)andZiegelschmidtetal.
(2005)foramoredetaileddescription.
11()ijtNICTijtICTijtijijtijtLKKFAY,,=whereICTKisthestockofICTcapitalservices,NICTKthestockofothercapitalservices,andLameasureofthestockoflaborservicesthataccountsfortheskillcompositionoflabor(seethegrowthaccountingmethodologydescribedbelow).
TheparameterAistheHicksneutraltechnologyparameter(alsocalledmultifactorproductivityortotalfactorproductivity):itaccountsforalltechnologies(definedlooselysoastoalsoincludeorganizationalandmanagerialfactors)thatarenotbiasedtowardsanyfactorofproduction.
Inthisframework,innovationsthatarespecifictooneoranotherfactorofproductionwillbeembodiedinthedefinitionofcapitalorlaborinputs.
Forexample,thedirecteffectoftheICTrevolutionappearsasachangeinthecompositionofcapitalservices(fromnon-ICTtoICTcapitalservices).
Similarly,changesinhumancapitalareembodiedinthestockoflaborservices.
Hence,anyeffectsofICTcapitalorhumancapitalonMFPcanbeseenasanexternalitycontributingtoincreasingreturnsattheindustrylevel.
Thisproductionfunctionassumesthat,ineachcountry,firmsandindustriescanchoosedifferentcombinationsoffactorsofproduction.
Asinthestandardneoclassicalmodel,theHicks-neutralworldtechnologyfrontierineachindustryjisdefinedby:()ijtFjtAMaxA=Inthisworld,Hicks-neutraltechnologiesdiffuseperfectlyacrosscountriesifandonlyif,foreachcountryi:FjijAA=.
Inaddition,inthestandardneoclassicalmodel,thesteady-stateaggregatecapital-laborratioisafunctionoftheHicks-neutraltechnologyparameter:()ijijijAhLK=,whereKisameasureoftheaggregatestockofcapitalservices.
17Hence,inasteady-state,theICTcapitallaborratiowilldependonthecompositionofcapitalbetweenICTandtraditionalcapitalgoods(non-ICT)andtheproductivityparameter:()ijijICTijijICTijAhKKLK=B.
EmpiricalSpecificationWeconsideralinearizedprocessofdiffusionofHicks-neutralmultifactorproductivity(MFP)acrosscountrieswhereby,withineachindustry,theMFPlevelincountryiisinfluencedbytechnologicaltransfersfromthetechnology-frontiercountry.
WeassumethatMFPforagivenindustryjincountryifollowsanauto-regressivedistributedlagADL(1,1)processinwhichthelevelofMFPisco-integratedwiththelevelofMFPofthetechnologicalfrontiercountryF.
18Notethat,inthisframework,theleadercountryFisnotnecessarilythesameacrossallindustries,andcanchangeovertimewithinanyindustryj.
Specifically,thedynamicsoftechnologydiffusionismodeledas:17Therelationshipislog-linearforaCobb-Douglasproductionfunction.
18Usingindustrydata,asimilarspecificationwasusedbyGriffithetal.
(2004),ScarpettaandTressel(2002),NicolettiandScarpetta(2003),Cameronetal.
(2005),andConwayetal.
(2006).
Vandenbusscheetal.
(2006),KentandSimon(2007)andGustandMarquez(2004)followedthesameapproachatthemacroeconomiclevel.
12ijtijtFjtFjtijtijtXAAAAελβββ++++=113211lnlnlnlnAssumingthat,ineachindustry,MFPgapsareconstantinthelongrun(areasonablesteady-stateassumptionthatdoesnotrequireconvergencetothesameleveloftechnologyacrosscountries),thespecificationbecomes:()ijtijtFjtijtFjtijtXAAAAελββ++Δ=Δ11112ln1lnlnThediffusionofICTtechnologiesacrosscountriescanbeanalyzedbylookingatthedynamicsoftheICTcapitaltolaborratio.
Inasimilarfashion,theshort-rundynamicsoftheICTcapitallaborratioismodeledas:()'1113211lnlnlnlnlnijtijtICTijtICTFjtICTFjtICTijtICTijtKKXKLKLKLKLεγμααα+++++=Again,withconstantICTcapitallaborratiogapsinthelongrun,theshort-rundynamicsbecomes:()'11112lnln1lnlnijtICTijtjtICTFICTiICTFjtICTijtKKXKLKLKLKLεγμαα+++Δ=ΔWeincludeafullsetofcountriesfixedeffectstocontrolforunobservedcountryfactorsthatcouldaffectproductivity;theyalsopermittocontrolforpossiblesystematicdifferencesacrosscountriesinthemeasurementofvariablesofinterest.
Finally,weincludeafullsetofindustry-yearfixedeffectstocontrolforgeneraltrendsandco-movementsofindustryMFPacrosscountries.
19Theempiricalanalysisisperformedatanannualfrequency,ratherthanoverlongerperiods.
Therearetworeasonsforthischoice:(1)takingaveragesoverseveralyearsmaysmoothoutandthereforehideproductivityaccelerations,ordecelerations,dependingonwhichperiodofthecycletheaverageisperformed;20(2)theimpactofreformsonproductivitygrowthmaymaterializerelativelyquicklyandfadeoutafterseveralyears,soaveragesmaynotpickupsuchdynamicsverywell.
C.
TheLong-RunImpactofCovariatesXonMFPandCapital-LaborRatioLevelsThelong-runimpactofavectorofcovariatesXontheMFPgapacrosscountriesinanygivenindustryjcanbederivedfromthepreviousspecification.
Inthelongrun,anypotentialgapsinthelevelsofMFPwillbeconstant,implyingthatMFPgrowthwillbethesameforallcountriesineachindustry.
ThemarginalimpactofXonthesteady-stateMFPgapisthengivenby:11lnβλ=XAAFjij19Industry-yearfixedeffectsalsocontrolfortheMFPgrowthoftheleaderineachindustry.
20Cyclesareunlikelytobesynchronizedacrosscountriesandindustries,so,inacross-countrysettingitwouldbeextremelydifficulttocomputeseveralyearaveragesallstartingatthesamestageofcyclesindifferentcountriesandindustries.
13ThereforethecovariatesXcanbeseenasaffectingMFPgrowthintheshortrun,andaffectingconvergenceofMFPlevelsacrosscountriesinthelongrun.
D.
DataSourcesandMethodologyThemainsourceofinformationforEuropeancountries,theU.
S.
andJapanistheEUKLEMSdatabasethatprovidesgrowthaccountsatadisaggregatedsectorallevel.
21Asexplainedbelow,andincontrasttopreviousindustrydatasets,thisdatasetprovidesinformationonthecompositionofinputsthatiscomparableacrosscountries.
TheEUKLEMSdatabaseisaugmentedwiththeIndustryGrowthAccounting(IGA)databaseforAustraliaandCanada,whichisconstructedfollowingasimilarmethodology.
22WealsousedtheOECDSTANdatasetforadditionalgrowthaccountinginformation,theOECDANBERD2&3datasetsforR&D,andtheOECDbilateraltradedatabase.
CountriesincludedareAustralia,Austria,Belgium,Canada,Denmark,Finland,France,Germany,Italy,Japan,Netherlands,Spain,Sweden,theU.
K.
,andtheU.
S.
ThesectoraldecompositionretainedfortheanalysisisgiveninTable1.
Sectorsrelatedtothepublicsectorweredroppedfromtheanalysis.
23Whensectoralaggregationsvariedacrossdataset,standardgrowthaccountingmethodswereusedtoobtaincomparablelevelsofaggregationsforeachvariableusedintheanalysis.
TheindexofemploymentprotectionlegislationisfromNicolettietal.
(2001)andtheindexofproductmarketregulationsisfromConwayetal.
(2006b).
Table1.
SectoralDecomposition21Thedatabaseispubliclyavailableathttp://www.
euklems.
net/andisdescribedinTimmeretal.
(2007a),andinTimmeretal.
(2007b).
22TheIndustryGrowthAccounting(IGA)databaseispubliclyavailableathttp://www.
ggdc.
net/dseries/iga.
htmlandisdescribedinInklaaretal.
(2005).
23SectorsexcludedareCommunitySocialandPersonalServices(ISICsectors75to99)ISICRev.
3CodeIndustryNameICTClassification01-05Agriculture,forestryandfishingNon-ICT10-14MiningandquarryingNon-ICT15-16FoodproductsNon-ICT17-19Textile,clothing&leatherNon-ICT20WoodproductsNon-ICT21-22Paper,printing&publishingICTusing23PetroleumandcoalproductsNon-ICT24ChemicalproductsNon-ICT25Rubber&PlasticsNon-ICT26Non-metallicmineralproductsNon-ICT27-28MetalproductsNon-ICT29Machineryandequipment,N.
E.
C.
ICTusing30-33ElectricalandopticalequipmentICTproducing34-35TransportequipmentNon-ICT36-37ManufacturingN.
E.
C.
;recyclingICTusing40-41Electricity,gasandwatersupplyNon-ICT45ConstructionNon-ICT50-52Wholesaleandretailtrade;repairsICTusing55HotelsandrestaurantsNon-ICT60-63Transport&storageNon-ICT64PostandtelecommunicationsICTproducing65-67FinancialintermediationICTusing71-74BusinessservicesICTusingIndustryList(MarketEconomy)14E.
GrowthAccountingThegrowthaccountingmethodologyusedinthedatabaseisdescribedinJorgensonetal.
(2005).
Thestockofcapitalservicesisbasedontheaggregationofvariousrealcapitalassets.
Foreachrealassettypes,thecapitalstockisconstructedusingtheperpetualinventorymethod(industrysubscriptjisomittedforsimplicity):()tStStSIKK,1,,1+=δGrowthintheaggregatecapitalstockismeasuredbythesumofcapitalserviceflowsforeachassettypeweightedbytheshareofeachcomponentinthevalueofcapitalcompensationaveragedovertwoyears([]1,,,21+=tStSKtSvνν):24tSKtSStKK,,lnlnΔΣ=ΔνInacompetitivemarket,therateofreturnoneachcapitalgoodisequaltotherentalpriceforeachassettypes:tStSSttSppRr,,,+=δwheretherentalpriceoftheassetisthesumofanominalrateofreturnR,adepreciationrateSδspecifictotheassets,minusthegrowthratetSp,oftheassetprice.
DetailsontheconstructionoftherentalpricescanbefoundinTimmeretal.
(2007b)fortheEUKLEMSdatabaseandinInklaaretal.
(2005)fortheGGDCIndustryGrowthAccountingDatabase.
Usingthismethodology,twoseriesofcapitalstockareconstructed:oneforInformationandCommunicationTechnologycapitalgoods(ICTK)andoneforothertypesofcapitalgoods(NICTK).
ICTcapitalgoodsincludecomputingequipment,communicationequipmentandsoftwares.
25Ineachindustryj,undertheassumptionofacompetitivefactormarketthatequalizesthepriceofeachfactorofproductiontoitsmarginalproduct,fullinpututilizationandconstantreturnstoscale,thelogchangeinvalueaddedVAjcanbedecomposedasthesumoflogchangesinthetwotypesofcapital,laborandHicksneutralmulti-factorproductivityA,whichisthereforederivedasaresidual:jNICTjNICTICTjICTjLjAKKvLVAlnlnlnlnlnΔ+Δ+Δ+Δ=ΔννAsforthecapitalstock,laborinputisalsomeasuredaslaborservices(numberofhoursworked)aggregatedoverdifferenttypesoflabor,includingdifferentskilllevels(industrysubscriptjomittedforsimplicity):2624Thismethodologyallowstoaccountforthefactthatsomeassetsaremoreproductivethanothers.
25Fornon-ICTcapitalgoods,theIGAdatabaseincludestransportequipment,non-ICTmachineryandequipmentandnon-residentialstructures.
TheEUKLEMSdatabaseincludesabroadersetofcapitalgoods.
26SeeTimmeretal.
(2007)foradescriptionofthelaborservicevariableintheEUKLEMSdatabase.
ForAustralia,laborserviceincludingthedistinctionbetweenskilledandunskilledlaborinputsarefromScarpettaandTressel(2002).
15∑Δ=ΔkkLkLLlnlnνwhereLkvisthe2yearaverageshareoflabortypekintotallaborcompensation.
Forallaggregationpurposes,wefollowthe"aggregationoverindustries"methodfirstdevelopedbyJorgenson,GollopandFraumeni(1987)anddescribedinJorgensonetal.
(2005).
Thismethodologyallowstoperformagrowthaccountingdecompositionattheindustrylevelwhileavoidingtheassumptionofperfectmobilityofinputsacrossindustries.
Forinstance,aggregaterealaggregatevalueaddedisdefinedasaTornqvistindexoverindustryvalue-added:∑Δ=ΔjjjVAwVAlnlnwherejwisthetwoperiodaverageshareofthenominalvalue-addedofindustryjintotalnominalvalueadded.
Byidentification,onecanderiveaggregategrowthincapitalservicesandlaborservicesas:Δ=ΔΔ=ΔΔ=Δ∑∑∑jjLjLjjICTjICTjICTjICTjNICTjNICTjNICTjNICTLvwLKvwvKKvwvKlnlnlnlnlnlnνWhere:∑=jNICTjjNICTvwν,∑=jICTjjICTvwνand∑=jLjjLwνν.
Aggregationofothervariablesaredonebyusingthevalue-addedsharesofeachindustryjwasweights.
F.
LevelAccountingIneachindustry,anindexofMFPisdefinedbycomparingeachcountrytoahypothetical"average"countryforeachyear:27=NICTjtNICTijtNICTICTjtICTijtICTjtijtLjtijtijtKKKKLLVAVAAlogloglogloglogνννwhereanupperbarstandsforanaverageacrosscountriesofvalueadded,laborandcapitalstocks.
ThefactorsharesLν,ICTνandNICTνaresimplearithmeticaveragesofthefactor27SeeScarpettaandTressel(2002)andInklaaretal.
(2006)foramoredetaileddescription.
ThemethodwasoriginallyproposedbyCaves,ChristensenandDiewert(1982)andisthelevelaccountinganalogytotheTornqvistlevelapproachusedintheindustrygrowthaccountingdecomposition.
16sharesinthecountryconsideredandfactorshareofthe"average"country.
Thefactorsharesofthe"average"countryaregeometricaveragesoverallcountries'factorshares.
Next,the(log)levelofMFPrelativetotheleadingcountryisdefinedasthelogdifferencebetweentheindexofMFPofagivencountryandtheindexofMFPofthecountrywiththehighestindexinaparticularindustryandyear:FjtijtAAlogwhere(){}ijtiFjtAMaxAloglog=.
G.
ExchangeRatesforInternationalComparisonsTobecomparableacrosscountries,allvariablesusedinthegrowthaccountingmustbeexpressedinacommoncurrency.
UntilrecentlyonlyGDPPPPexchangerateswereavailable.
Theproblemisthat,inpresenceofsystematicdifferencesinpriceevolutionsacrossindustries,theuseofGDPPPPsislikelytoresultinsystematicbiaseswhencomparingproductivitiesacrosscountries(seeforinstanceSorensen,2001).
Suchsystematicdifferencesinpriceevolutionsacrossindustriesareverylikelytobesignificant,asillustratedforexamplebythedrasticpricereductionsofICTgoodsoverthepast15yearsrelativetothepriceofothergoods.
Inthisstudy,weusethegrossoutputPPPsconstructedbyTimmeretal.
(2007c).
28Theyfindconsiderablevariationsofoutputpricesacrosscountrieswithinindustries,suggestingthatPPPdoesnotholdwithinindustries.
Moreover,pricesrelativetotheUSpricelevelalsovaryimportantlyacrossindustrieswithincountries,suggestingthatindustrylevelpriceadjustmentsareimportant.
Forinstance,inAustralia,foodproducts'relativepriceis1(e.
g.
relativetotheUSpricelevel),butitis1.
70forRadio,TelevisionandCommunicationEquipment,1.
53forTextilesbutonly0.
85forCokeandPretroleumProducts.
29IV.
EMPIRICALANALYSISA.
AFirstLookattheDataTable2presentsdescriptivestatisticsofthekeyvariables.
Overtheperiodstudied(1980–2003),industrylaborproductivitygrewat2.
8percentonaverageinoursample.
Growthofmulti-factorproductivityreached1.
3percentonaverage,andthedistancetotheMFPleaderwas39percentonaverage.
30AustralianindustriesexperiencedsomewhathigherlaborproductivityandMFPgrowth,andwerefartherawayfromtheMFPleaderonaverage,28Thedataarepubliclyavailableathttp://www.
euklems.
net/29SeeTimmeretal.
(2007c),Table5.
30DistancetothefrontieriscomputedasFjtijtAA1.
Otherstudieshavefounddistancestothefrontierattheindustrylevelvaryingfrom20percentto50percent.
17butnotsignificantlyso(distancetotheMFPleaderwas43percentonaverage).
AccumulationofICTcapitalserviceswasfasterthanthatofmoretraditionalcapitalgoods:theICTcapitaltolaborratiogrewat13.
9percentonaverage,whilethenonICTcapitaltolaborratiogrewatonly2.
6percent.
AustraliastandsoutashavingexperiencedmuchfasteraccumulationofICTcapitalthanothercountriesinthesample,butnotforothertypesofcapital.
Table3showsthatlaborproductivitygrowthisstronglycorrelatedwithMFPgrowth,suggestingthatdifferencesoflaborproductivityacrossindustriesandcountriesaremainlyexplainedbydifferencesintheefficiencyoftheuseoffactors,ratherthanbydifferencesincapital-laborratiosacrossindustriesandcountries.
Hence,toexplaindifferencesinlaborproductivity,itiskeytounderstandwhatdeterminestheefficiencyoftheuseoffactorsacrossindustriesandcountries.
Table4showsMFPleadersinselectedindustries.
Australiahasconsistentlyrankedwellinminingandintransportandstorages.
TheU.
S.
hasbeentheleaderinmanufacturingindustriestakenaswholeamongOECDcountries.
Withinmanufacturing,FinlandandtheU.
S.
havebeenthetwoleadersintheproductionofICTgoodsinrecentyears.
ContinentalEuropeancountrieshavebeentheMFPleadersinservicesindustries(suchaswholesaleandretailtrade,businessservices).
31RelativeMFPlevelsalsochangesignificantlyovertime,asillustratedbyAustralia'sexperiencebetween1980and2003(Figure6).
TheFigurealsoshowsthattheevolutionofMFPlevelsovertimehasastrongindustry-specificcomponent.
ThesepatternsaregenerallyconsistentwithindustrylaborproductivitylevelspresentedinDolmanetal.
(2007),andsuggestthatindustry-levelstudiesareimportanttounderstandtheevolutionofcountries'productivitylevels.
Turningtopolicyvariables,therearesubstantialvariationsinlaborandproductmarketregulationsacrossOECDcountries.
Theindexofthestrictnessofemploymentprotectionlegislation(EPL),whichincludesbothregularandtemporarycontracts,isfromNicolettietal.
(2001).
32AsshownonFigure4,Anglo-Saxoneconomies(theU.
S.
,theU.
K,Canada,AustraliaandIreland)haveveryflexiblelabormarketsasmeasuredbyEPL,comparedwithotherOECDcountries.
Attheotherextreme,SouthernEuropeancountries(Spain,Portugal,Italy,Greece),aswellasGermanyhavestrictemploymentlegislation.
WeconsidertheEPLindexasabroadindicatorofflexibilityofthelabormarket.
Overthepastdecades,OECDcountrieshavealsoderegulatedtheirproductmarketssignificantly,asshownonChart5(seeNicolettietal.
(2001),andConwayetal.
2005).
33By31Leadersbasedonlaborproductivityareavailableuponrequest.
32Regulationsforregularcontractsinclude(i)proceduralinconveniencesfacedbyemployerswhendismissingaworker,(ii)advancenoticeofdismissalandseverancepayments,and(iii)standardsandpenaltiesforunfairdismissals.
Regulationsfortemporarycontractsinclude(i)theobjectivereasonsforwhichtheycanbeoffered;(ii)themaximumnumberofsuccessiverenewals;and(iii)themaximumcumulateddurationofthecontract.
33SeeConwayandNicoletti(2006)fordescriptionontheconstructionoftheindustry-specificproductmarketregulatoryimpactvariable.
18theendoftheperiod,AustraliahasderegulateditsproductmarketsignificantlymorethanotherOECDcountries,reachingalevelofproductmarketregulationslightlybelowtheU.
S.
level.
Theindustryspecifictime-varyingindicatorofproductmarketregulationsisfromConwayandNicoletti(2006b),andmeasurestheextentofanti-competitiveproductmarketregulationsinnon-manufacturingsectorsaswellastheir'knock-on'effectsonmanufacturingsectors.
34B.
RegressionResultsBasicconvergenceregressionsforMFPgrowthandtheICTcapital-laborratioarereportedinTable5.
ThetoppanelpresentstheresultsforMFPgrowth.
BasedontheestimatesshowninColumn(1)ofpanelA,simplecalculationsimplythat,foratypicalindustrywitha40percentgaptotheleaderinthatindustry,thediffusionoftechnologyacrossOECDcountriesimpliesanMFPgrowthof0.
21percentagepointsannuallyonaverage.
35GiventheactualannualMFPgrowthobservedinthesample(1.
3percent),thisimpliesthatdiffusionoftechnologiesonaverageexplainsonlyabout15percentofobservedMFPgrowth.
Columns(1)and(2)ofPanelAalsoshowthatconvergenceofMFPlevelshassignificantlysloweddownoverthepasttenyears.
36Interestingly,whenallowingforaconvergencefactorspecifictoAustralia,theaverageeffectdropsinsizeandsignificance,whiletheAustralianeffectisstronglysignificant(columns(3)and(4)).
ThissuggeststhatAustraliaisoneofthecountrieswhereconvergenceofMFPlevelshasplayedasignificantrole,andthat,onaverage,convergencehasnottakenplaceinotherOECDcountries.
Basedontheestimatedcoefficientsofcolumn(3),andtheaveragegaptotheindustryleaderinAustralia,simpleback-of-the-envelopecalculationsimplythatconvergenceledtoanannualMFPgrowthofabout0.
4percentinAustralia.
Moreover,column(4)suggeststhatthisfasterconvergenceofMFPinAustraliatookplacemainlyafter1990.
Finally,incolumns(5)and(6)weshowthattheICTrevolutiondoesnotseemtohaveledtosignificantlydifferentconvergenceexperiencesbetweenICTusing,ICTproducingandotherindustriesonaverage(columns(5)and(6)).
Finally,wealsofindthat,inAustralia,MFPinICTusingsectorshasgrown0.
1percentagepointsfasterthanAustralia'sownaverage(basedontheestimatedcoefficientoftheAustraliadummyincolumn(3)).
34Non-manufacturingsectorsincludeelectricity,gasandwatersupply,airandroadtransport,roadfreight,postandtelecommunications(coveredover1975-2003),andretaildistributionandprofessionalbusinessservicesfor1996-98and2003.
Theindexisbasedonseveralsub-indicescoveringbarrierstoentry,publicownership,marketstructure,verticalintegrationandpricecontrols.
Italsoindirectlymeasuretheindirectimpactoftheseregulationsformanufacturingsectors.
35Thisiscomputedasfollows:-0.
0095*Log(0.
6).
36Conwayetal.
(2006)findasimilarresultwhenlookingatconvergenceoflaborproductivityacrossOECDcountries.
19PanelBofTable5exploresconvergenceintheuseofICTinputsbylookingattheconvergenceofICTcapital–laborratios.
Again,Australiaappearsasanoutlier,and,asimpliedbythecoefficientontheAustraliadummyincolumn(3),hasexperiencedanICTcapitaldeepening3.
5percentagepointsabovethesampleaverageof13.
9percent,anestimateconsistentwiththesummarystatisticsdiscussedearlier.
AswithMFP,Australia'sconvergenceofitsICTcapital-laborratioseemstoexplainmostoftheaverageconvergenceeffect(columns(3)and(4)).
Table6explorestheroleofproductmarketregulations(PMR)inexplainingdifferencesintheconvergenceofMFPbestpracticesattheindustrylevelacrossOECDcountries.
TheaverageeffectofPMRisnotrobustlysignificant,whenenteredeitherasalineareffectoranon-lineareffectthatdependsonthegaptotheMFPleader.
However,wefindarobustlysignificantpositiveeffectofproductmarketderegulationinICTusingindustries,butnoeffectinotherindustries.
Theseeffectsarerobustwhencontrollingforpotentialexternalities(orincreasingreturnstoscale)causedbyhumancapitalorICTcapital.
37Interestingly,wealsofindaverysignificantandpositiveeffectofhumancapitalonMFPgrowth(butonlyfortheperiod1990–2003),eventhoughourmeasureofMFPhasalreadynettedoutthecompositioneffectsoflaborinput.
Thisresultisconsistentwithrecentstudiesemphasizingtheroleofhumancapitalinspeedingconvergenceattheaggregatelevel(seeVandenbusscheetal.
,2006).
InTable7,weexploretheimpactoflabormarketflexibilityonMFPgrowth.
Column(1)showsastronglynegativecorrelationbetweentheindexofEPLandindustrylevelMFPgrowthacrosscountries.
Incolumn(2)wesplitindustriesaccordingtotheshareofhighskillworkersintotalhoursworkedineachcountry.
Industriesthatareabovethecountrymedianarelabeled"highskillindustries",andthosebelow"low-skillindustries".
WeshowthatthenegativeassociationbetweenMFPgrowthandtheEPLindexisstronglysignificantinhigh-skillindustries,butnotinlowskillindustries.
Thisresultisrobusttotheinclusionofcountryfixedeffects,andremainssignificantaftercontrollingfortheshareofhighskillworkersinlaborcompensationandtheshareofICTcapitalintotalcapitalcompensation.
Hence,labormarketflexibilityseemsparticularlyimportantforMFPgrowthinindustriesthataremoreintensiveinhumancapital.
Table8showsthatbothlabormarketflexibilityforhighskillindustriesandproductmarketderegulationforICTusingindustrieshavecontributedinboostingMFPgrowthinOECDcountries.
Moreover,increasedflexibilityofthelabormarketandproductmarketreformsseemtoexplainthe"Australianeffect"identifiedinTable5(seecolumn(7)).
WecontinuetofindthatahighershareofskillworkersinlaborcompensationremainssignificantlyassociatedwithfasterMFPgrowthafter1990.
38However,wedonotfindanyclearsignsofexternalitiescausedbyICTcapitalwithinindustriesonaverage.
37Conwayetal.
(2006)obtainasimilarresultforlaborproductivity.
38Theeffectofhumancapitalisinsignificantwhencountryfixedeffectsareomitted.
Thismightreflectproblemsofcomparabilityofmeasuresofhumancapitallevelsacrosscountries.
However,deviationfromcountryandindustrymeansarelikelytobebettermeasured.
Thefactthatwefindastrongeffectwithin(continued…)20Finally,Table9showsthatcountrieswithmoreflexiblelabormarketsseemtohaveexperiencedafasterICTcapitaldeepeningthanothercountries.
Thisresultisnotdrivenbydifferentcompositionsofthestockofcapital(wecontrolforthepreviousyear'sshareofICTcapitalintotalcapital)orbyalowerinitialstockofICTcapital(wecontrolfortheinitialICTcapital-laborratio).
Resultsontheroleofproductmarketregulationsaremoreambiguous:theeffectsofPMRbecomeinsignificantwhenthedegreeoflabormarketflexibilityisaccountedfor,orwhenindustry-yearfixedeffectsareincluded,suggestingthatanynegativeassociationbetweenthePMRindexandMFPgrowthreflectedindustries'specificities.
C.
RobustnessTestsAnumberofrobustnesstestsarereportedinAppendixTables1–3.
First,weshowthatweobtainverysimilarresultswhenlookingattheeffectsoflabormarketflexibilityandproductmarketregulationsonlaborproductivitygrowthinsteadofMFPgrowth(Appendix1).
Next,weshowthatthemainresultsarenotdrivenbyanyspecificcountryinthesample.
InAppendix2,wereportthecoefficientandt-statisticsforthevariableofinterestwhencountriesaredroppedone-by-one,fortheregressionwiththehighestandlowestp-valuerespectively.
Thisconfirmsthatourresultsarenotdrivenbyanyoutliercountryinthesample.
Finally,wereportregressionsusing3yearaveragesinsteadofregressionsatanannualfrequency.
Using3yearaveragemayhelpsmoothoutnoiseinthedata,butmayalsoreducetheprecisionofestimatedeffects,forthereasonsdiscussedinsectionIII.
D.
DoReformsExplainAustralia'sProductivityPerformanceTable10exploreswhetherproductandlabormarketreformscanexplainAustralia'slongperiodofhighproductivitygrowth.
ItusesestimatesofvariousregressionstocompareactualandpredictedMFPperformances.
InPanelA,weaskwhetherproductmarketreformsundertakensincetheearly1990scanexplaintheorderofmagnitudeofaverageproductivitygrowthaccelerationinICTusingindustriesafter1991relativetothe1980s.
MFPgrowthreached1.
2percentonaverageeachyearbetween1991and2003,comparedwith0.
4percentbetween1980and1990.
InICT-usingindustries,thePMRindexwentdownfrom0.
22to0.
18inAustraliabetweentheearly1990sand2003.
Givenanaveragegaptotheleaderofabout40percent,wecanestimatetheMFPgrowthaccelerationcausedbyPMRreformsusingregressionestimates:weobtainapredictedaccelerationofMFPgrowthof0.
1percentonaverageeachyear(usingthechangeinendofperiodPMRindicestoestimatetheMFPacceleration).
Moreover,giventhatthegaptotheleaderhadsomewhatwidenedduringthe1980sfrom42percentto47percent,thecatch-upeffectalsoimpliesanaccelerationofMFPgrowthof0.
07percent.
Overall,thisimpliesthatlessthan20percentofAustralia'sproductivityaccelerationisexplainedbyproductmarketreformsandconvergence,soproductmarketderegulationsalonecannotexplainAustralia'sproductivityacceleration.
countriesafter1990,however,isstronglysuggestiveofhumancapitalexternalitiesattheindustrylevel,withincountries.
21Theimpactofproductmarketreformswasnotnegligible,butotherfactorsmusthavealsocontributedtotheproductivityperformance.
39Incontrast,theeffectoflabormarketflexibilityasimpliedbyourregressionresultsseemstobeoftheorderofmagnitudeofobservedMFPgrowthdifferentialsacrosscountries,andthereforemaybeabettercandidateinexplainingMFPcatch-upwithinindustriesinOECDcountriesingeneral,aswellasAustralia'sspecificexperienceinthe1990safterthelabormarketwasderegulated.
PanelBtestswhetherdifferencesofEPLacrosscountriesaregoodpredictorsofdifferencesinMFPgrowthovertheperiod1990–2003.
Forthisexercisewecomparethecountrieswiththemostflexiblelabormarketswiththecountrieswiththeleastflexiblelabormarkets.
WefindthatEPLexplainsMFPgrowthdifferentialsquitewell:thefourcountrieswiththemostflexiblelabormarketsin2001(theUS,theUK,AustraliaandCanada)experiencedanMFPgrowthof1.
13percentinhighskillindustriesduringthatperiod.
Thecountrieswiththelessflexiblelabormarkets(Germany,SpainandItaly)experiencedanannualMFPgrowthof0.
68percentinhighskillindustries.
FromtheestimatedcofficientontheEPLvariable,ourregressionresultsimplythatdifferencesinEPLacrossthetwogroupspredictanMFPgrowthdifferentialof0.
58percentagepointsforthecountrieswithflexiblelabormarketsrelativetothecountrieswithinflexiblelabormarkets.
Hence,theMFPgrowthdifferentialobservedbetweenthetwogroupsofcountries(0.
44percent)isoftheorderofmagnitudeofthedifferentialpredictedbytheregressionresults(0.
58percent).
E.
OtherDeterminantsofMFPConvergence:TheRoleofHumanCapitalandR&DTable11showsthatourkeyresultsonlabormarketflexibilityandproductmarketderegulationremainsignificantwhencontrollingforvariousothervariables,includingR&Dintensity,andtraderelatedvariables.
WealsouncoverastrongandsignificantassociationbetweenR&DintensityandMFPgrowth,asinotherstudies(seeinparticularGriffithetal.
(2004)).
Specifically,wefindastrongindirectbenefitofR&D,butnodirectsignificantassociationbetweenR&DintensityandMFPgrowthaftercontrollingfortheindirecteffect.
40TheindirecteffectofR&DisproxiedbyaninteractiontermbetweenR&DintensityandtherelativelevelofMFP.
ThisindirecteffectofR&DsuggeststhatdomesticR&Dmayenhancetechnologicaltransfer,byincreasingdomesticabsorptivecapacity,thusallowingcountriestoimportandadapttoforeigninnovationsfaster.
ThesizeoftheeffectofR&DintensityonMFPgrowthmaybebiasedbecauseofendogeneity,andthereforeshouldbeinterpretedcautiously.
R&Dintensity,whichmeasuresanoutcome,couldreflectanindustry'sexpectedfutureproductivityperformance:industriesmaydomoreR&Dwhentheyexpecttobeabletoinnovatemoreeasily.
Moreover,R&Dspendingsmayonlyreflectacountry'sindustrystructure,andtheoverallbusiness39Themagnitudesoftheeffectsobtainedinourregressionsareconsistentwiththosefoundinearlierstudies(seeScarpettaandTressel,2002,NicolettiandScarpetta,2003,andConwayetal.
,2006).
40Previousstudies(Griffithetal,2004,Cameronetal.
,2005)hadfoundadirectpositiveeffectofR&DonMFPgrowth.
SeealsoCardarelli(2001b).
22environment,asshowninDavisandTunny(2005).
41However,controllingforindustry-specificco-movementsshouldpartially,butnotentirelyaddresssuchconcerns,bynettingoutanyindustryspecificpatterns.
Asimple,albeitroughwaytoevaluatetheimportanceofendogeneityconcernsistoseehowthesizeandsignificanceoftheR&Dintensityvariablechangewhenitislaggedbymorethanoneyear.
WefindthatthedirecteffectofR&Dintensitybecomesinsignificantwhenitislagged3yearsormore.
42Therearetwopossible,non-mutuallyexclusiveexplanationsforthisresult:(a)theinitialeffectwasdrivenbyendogeneity,(b)theactualdirecteffectofR&Donproductivitymaterializesduringthefirsttwoyears.
Incontrast,wefindthattheindirecteffectofR&Dviatheadoptionofforeigntechnologiesremainstronglysignificantevenafter5years,suggestingthattheindirecteffectmaybelesslikelytobedrivenbyendogeneity.
43TheestimatesofTable11suggestthat,ifAustralianfirmsincreasedtheirR&Dintensityfrom1.
8percenttothesampleaverageof3.
6percent,MFPgrowthwouldincreaseby0.
04percentannually.
44WealsoshowinTable11thattheeffectofhumancapitalbecomesstronglysignificantwhencontrollingfortheeffectsofR&D,givingsupporttotheargumentthathumancapitalmaycreateexternalitiesevenattheindustrylevel,andmayhaveindirecteffectsonlaborproductivity,inadditiontodirecteffectsonlaborproductivitycausedbyinputcompositioneffects.
Theeffectsareeconomicallysignificant:anincreaseintheshareofhigh-skillworkersintotallaborcompensationbyonestandarddeviation(13percent)wouldleadtoanaccelerationofindustryMFPgrowthof0.
5percentannually.
Dolmanetal.
(2007)andDavisandRahman(2006)showthatAustraliahasnarrowedthegapineducationalattainmentinrecentyearsasyoungercohortsofworkersaremoreeducatedthatoldercohorts.
Finally,wealsocontrolfortraderelatedfactors,butnonearesignificantlyandrobustlycorrelatedwithMFPgrowth.
Wecontrolfortheexportorientationofeachsector,asmeasuredbytheshareofexportsinindustryvalue-added.
WealsocontrolfortechnologicaldiffusionthroughtrademeasuredbytheimportweightedR&DintensityofU.
S.
industries.
Indeed,anumberofpapers(includingCoeandHelpman(1995)andAcharyaandKeller(2007))havefoundastrongandsignificanteffectoftechnologydiffusionthroughtrade.
45Howeverwedonotfindanysignificantpositiveeffectsoftradeinducedtechnologicaldiffusioninoursampleofcountries.
41TheyalsofindaweakcorrelationbetweenR&Dandmoredirectmeasuresofinnovation,suggestingthatlowerlevelsofR&Ddonotnecessarilyimplythatacountryislessinnovative.
SeeProductivityCommission(2007)forathoroughanalysisofAustralia'spublicsupportofscienceandinnovation.
42Theseresultsarenotreportedandareavailableuponrequest.
43Theseresultsarenotreportedandareavailableuponrequest.
44AnincreaseofR&Dintensitybyonestandarddeviationto8.
9percentwouldincreaseMFPgrowthbyabout0.
1percentannually.
45Keller(2002)showsthatgeographyanddistancetoinnovatorsaffectinternationaltechnologicaldiffusion.
SeeKeller(2004)forasurvey.
23V.
CONCLUSIONThispaperhaslookedatAustralia'spastproductivityperformanceinthebroadercontextoftechnologydiffusionacrossOECDcountries.
Inmanyrespects,Australiahasbeenanoutliersincetheearly1990s:Australiahasenjoyedalongeconomicexpansionandexperiencedstrongproductivitygains;Australiahasalsoundertakenmanyreforms,inparticularinthelaborandproductmarketareas.
Weshowthatthesereforms,whentakentogether,maylargelyexplaintheproductivityperformanceoftheAustralianeconomy.
Morebroadly,wefindthat,amongOECDcountries,reformsoftheproductmarketsignificantlyaffectMFPgrowthinindustriesthatuseICTintensively.
Moreover,labormarketflexibilityseemstohaveastrongandpositiveimpactonMFPinsectorsthataremoreintensiveinhumancapital,andtendstofostertheaccumulationofICTcapital.
Finally,weconfirmtheimportanceofR&Dforthespeedoftechnologicaldiffusion,andfindevidenceofhumancapitalexternalitiesattheindustrylevel.
However,thisdoesnotnecessarilyimplythatmoreR&D—oneinputofinnovationamongmanyothers—wouldalwaysresultinhigherproductivity.
24REFERENCESAcemoglu,D.
,Aghion,P.
,andZilibotti,F.
,2006,"DistancetoFrontier,SelectionandEconomicGrowth,"JournaloftheEuropeanEconomicAssociation,Vol.
4,No.
1pp.
37–74.
Acharya,R.
,andW.
Keller,2007,"TechnologyTransferThroughImports,"CEPRDiscussionPaperNo.
6296.
Aghion,P.
,Bloom,N.
,Griffith,R.
,Blundell,R.
,andHowitt,P.
,2005,"CompetitionandInnovation:AnInvertedURelationship,"QuarterlyJournalofEconomics,Vol.
120No.
2,pp.
701–28.
———,Blundell,R.
,Griffith,R.
,Howitt,P.
,andS.
Prantl,2006,"TheEffectsofEntryonIncumbentInnovationandProductivity,"(unpublished;Cambridge,Massachusetts:HarvardUniversity).
AustralianBureauofStatistics,2007,"ExperimentalEstimatesofIndustryMultifactorProductivity,"InformationPaper(5260.
0.
55.
001).
Banks,G.
,2005,"StructuralReformAustralian-Style:LessonsforOthers"PresentationtotheIMF,WorldBankandOECD,May2005,ProductivityCommission,Canberra.
Bertola,G.
,1994,"Flexibility,InvestmentandGrowth,"JournalofMonetaryEconomics,Vol.
34,No.
2,pp.
215–38.
Cameron,G.
,ProudmanJ.
,S.
Redding,2005,"TechnologicalConvergence,R&D,TradeandProductivityGrowth,"EuropeanEconomicReview,Vol.
49,pp.
775–807.
Caves,D.
,Christensen,L.
,andE.
Diewert,1982,"MultilateralComparisonsofOutput,Input,andProductivityUsingSuperlativeIndexNumbers,"EconomicJournal,Vol.
92.
Coe,D.
,andE.
Helpman,1995,"InternationalR&DSpillovers,"EuropeanEconomicReview,Vol.
39,pp.
859–87.
Conway,P.
,deRosa,D.
,Nicoletti,G.
,andF.
Steiner,2006a,"Regulation,CompetitionandProductivityConvergence,"OECDEconomicsDepartmentWorkingPapersNo.
509(Paris:OECDPublishing).
Conway,P.
,andG.
Nicoletti,2006b,"ProductMarketRegulationintheNon-ManufacturingSectorsofOECDCountries:MeasurementsandHighlights,"OECDEconomicsDepartmentWorkingPapersNo.
530(Paris:OECDPublishing).
Conway,P.
,Janod,V.
,andG.
Nicoletti,2005,"ProductMarketRegulationinOECDCountries:1998to2003"OECDEconomicsDepartmentsWorkingPapersNo.
419(Paris:OECDPublishing).
25Davis,G.
,andJ.
Rahman,2006,"PerspectivesonAustralia'sProductivityProspects,"AustralianTreasuryWorkingPaper2006–04(Canberra:Australia).
Davis,G.
,andG.
Tunny,2005,"InternationalComparisonsofResearchandDevelopment,"EconomicRoundup,Spring,AustralianTreasury(Canberra:Australia).
Dolman,B.
,Parham,D.
,andS.
Zheng,2007,"CanAustraliaMatchU.
S.
ProductivityPerformance"ProductivityCommissionStaffWorkingPaper(Canberra:Australia).
Griffith,R.
,Redding,S.
,andJ.
VanReenen,2004,"MappingtheTwoFacesofR&D:ProductivityGrowthinaPanelofOECDIndustries",ReviewofEconomicsandStatistics,Vol.
86,No.
4,pp.
883–95(firstversion2000).
Griffith,R.
,andR.
Harrison,2006,"ProductMarketReformandInnovationintheEU,"CEPRDiscussionPaperNo.
5849.
Gust,C.
,andJ.
Marquez,2004,"InternationalComparisonsofProductivityGrowth:TheRoleofInformationTechnologyandRegulatoryPractises,"LabourEconomics,Vol.
11No.
1,pp.
33–58.
Hopenhayn,H.
,andR.
Rogerson,1993,"JobTurnoverandPolicyEvaluation:AGeneralEquilibriumAnalysis,"JournalofPoliticalEconomy,Vol.
101,No.
5,pp.
915–38.
Inklaar,R.
,O'Mahony,M.
,andM.
P.
Timmer,2005,"ICTandEurope'sProductivityPerformance;Industry-LevelGrowthAccountComparisonswiththeUnitedStates,"ReviewofIncomeandWealth,Vol.
51,No.
4,pp.
505–36.
———,Timmer,M.
P.
,andB.
vanArk,2006,"MindtheGap!
InternationalComparisonsofProductivityinServicesandGoodsProduction,"GGDCResearchMemorandum,GD–89.
Cardarelli,R.
,2001a,"IsAustraliaa"NewEconomy"SelectedIssuesPaper(Washington:InternationalMonetaryFund).
———,2001b,"TechnologyTransferandR&D:ACross-CountryRegression,"SelectedIssuesPaper(Washington:InternationalMonetaryFund).
Jorgenson,D.
,Ho,M.
,andK.
Stiroh,2005,"GrowthofU.
S.
IndustriesandInvestmentsinInformationTechnologyandHigherEducation,"inMeasuringCapitalintheNewEconomy,ed.
byCarolCorrado,JohnHaltiwangerandDanielSichel(Chicago:UniversityofChicagoPress).
Jorgenson,D.
,andK.
Stiroh,2000,"RaisingtheSpeedLimit:U.
S.
EconomicGrowthintheInformationAge,"BrookingsPapersonEconomicActivity:1,pp.
125–211,BrookingsInstitution.
26Keller,W.
,2004,"InternationalTechnologyDiffusion,"JournalofEconomicLiterature,Vol.
XLII,752–82.
———,2002,"GeographicLocalizationandInternationalTechnologyDiffusion,"AmericanEconomicReview,Vol.
92,pp.
120–42.
Kent,C.
,andJ.
Simon,2007,"ProductivityGrowth:TheEffectofMarketRegulations,"ReserveBankofAustralia,RDP2007–04(Sydney:Australia).
Nicoletti,G.
,andS.
Scarpetta,2003,"Regulation,ProductivityandGrowth,"EconomicPolicy,April,pp.
9–72.
Nicoletti,G.
,Bassanini,A.
,Ernst,E.
,Jean,S.
,Santiago,P.
,andP.
Swaim,2001,"ProductandLaborMarketsInteractionsinOECDCountries,"OECDEconomicsDepartmentWorkingPaperNo.
312(Paris:OECDPublishing).
OECD,2003,ICT&EconomicGrowth:EvidencefromOECDCountries,IndustriesandFirms(Paris:OECDPublishing).
Oliner,S.
,andD.
Sichel,2000,"TheResurgenceofGrowthintheLate1990s:IsInformationTechnologytheStory"JournalofEconomicPerspectives,Vol.
14,No.
4,pp.
3–22.
Parham,D.
,Roberts,P.
,andSun,H.
,2001,InformationTechnologyandAustralia'sProductivitySurge,ProductivityCommissionStaffResearchPaper,AusInfo,(Canberra:Australia).
ProductivityCommission,2004,"ICTUseandProductivity:ASynthesisfromStudiesofAustralianFirms,"CommissionResearchPaper(Canberra:Australia).
———,2007,"PublicSupportforScienceandInnovation,"ProductivityCommissionResearchReport"(Canberra:Australia).
Salgado,R.
,2002,"ImpactofStructuralReformsonProductivityGrowthinIndustrialCountries,"IMFWorkingPaper02/10(Washington:InternationalMonetaryFund).
SaintPaul,G.
,1997,DualLaborMarkets–AMacroeconomicPerspective(Cambridge,Massachusetts:MITPress).
SaintPaul,G.
,2002,"EmploymentProtection,InternationalSpecialization,andInnovation,"EuropeanEconomicReview,Vol.
46,pp.
375–95.
Stiroh,Kevin,2002,"InformationTechnologyandtheU.
S.
ProductivityRevival:WhatDotheIndustryDataSay"AmericanEconomicReview.
27———,2004,"ReassessingtheImpactofITintheProductionFunction:AMeta-AnalysisandSensitivityTests,"FederalReserveBankofNewYork.
———,2006,"TheIndustryOriginsoftheSecondSurgeofU.
S.
ProductivityGrowth,"(unpublished;NewYork:FederalReserveBankofNewYork).
Scarpetta,S.
,andT.
Tressel,2002,"ProductivityandConvergenceinaPanelofOECDIndustries:DoRegulationsandInstitutionsMatter"OECDEconomicsDepartmentWorkingPaperNo.
342(Paris:OECDPublishing).
———,2004,"BoostingProductivityviaInnovationandAdoptionofNewTechnologies:AnyRoleforLaborMarketInstitutions"WorldBankPolicyResearchPaper3273(Washington:WorldBank).
Sorensen,A.
,2001,"ComparingApplesandOranges:ProductivityConvergenceandMeasurementAcrossIndustriesandCountries:Comment,"AmericanEconomicReview,Vol.
91,No.
4,pp.
1160–67.
Timmer,M.
,Ypma,G.
andB.
vanArk,2007c,"PPPsforIndustryOutput:ANewDatasetforInternationalComparisons,"GroningenGrowthandDevelopmentCentre,ResearchMemorandumGD–82(Holland:UniversityofGroningen).
Timmer,M.
P.
,O'Mahony,M.
,andB.
vanArk,2007,TheEUKLEMSGrowthandProductivityAccounts:AnOverview(unpublished;Holland:UniversityofGroningen).
Timmer,M.
,vanMoergastelT.
,andE.
Stuivenwold,2007b,EUKLEMSGrowthandProductivityAccounts,Version1.
0,PartIMethodology(unpublished;Holland:UniversityofGroningen).
Ziegelschmidt,H.
,Koutsogeorgopoulou,V.
,Bojornerud,S.
,andM.
Wise,2005,"ProductMarketCompetitionandEconomicPerformanceinAustralia,"OECDEconomicsDepartmentWorkingPapersNo.
451(Paris:OECDPublishing).
28Table2.
SummaryStatistics(1980–2003)1/VariableObsMeanStd.
Dev.
MFPgrowth44811.
3%3.
8%MFPgrowth-Australia3251.
5%3.
8%MFPgrowth-UnitedStates3831.
1%3.
9%DistancetoMFPfrontier448139.
3%19.
0%Distancetofrontier-Australia32543.
4%17.
8%Distancetofrontier-UnitedStates38336.
7%17.
9%Laborproductivitygrowth44812.
8%4.
0%Laborproductivitygrowth-Australia3253.
1%4.
5%Laborproductivitygrowth-UnitedStates3832.
8%4.
0%ICTcapitaldeepening448113.
9%7.
5%ICTcapitaldeepening-Australia32517.
7%6.
4%ICTcapitaldeepening-UnitedStates38315.
7%6.
5%Non-ICTcapitaldeepening44812.
6%3.
1%Non-ICTcapitaldeepening-Australia3252.
4%3.
6%Non-ICTcapitaldeepening-UnitedStates3832.
2%2.
9%R&Dintensity29443.
6%8.
2%R&Dintensity-Australia2361.
8%2.
6%R&Dintensity-UnitedStates2543.
8%6.
0%391215.
9%13.
5%392010.
9%11.
1%38954.
4%8.
7%448112.
6%12.
6%Employmentprotectionlegislationindex44812.
241.
18Productmarketregulation(PMR)42140.
220.
18AnnualpercentchangeinPMR4214-1.
7%3.
2%ICTusingindustries448129.
4%ICTproducingindustries44818.
7%Otherindustries448162.
0%1/Summarystatisticsarecomputedonthesampleofregression(1)Table5ICTcapitalcompensation(inpercentoftotalcompensation)Highskilllaborcompensation(inpercentoftotallaborcompensation)Hoursworkedbyhighskillworkers(inpercentoftotalhoursworked)Annualpercentchangeinshareofhoursworkedbyhighskillworkers29Table3.
CorrelationsMFPgrowthMFPgrowth1ICTcapitaldeepening0.
001810.
86-0.
00080.
283610.
940.
00Laborproductivitygrowth0.
90210.
10340.
189810.
000.
000.
00-0.
02550.
040.
01870.
00410.
010.
000.
090.
66-0.
0243-0.
0389-0.
0626-0.
0037-0.
123410.
010.
000.
000.
680.
00-0.
0454-0.
07540.
0102-0.
03550.
00210.
206410.
000.
000.
320.
000.
820.
00-0.
0113-0.
0491-0.
008700.
0315-0.
09420.
23110.
310.
000.
461.
000.
010.
000.
00-0.
0116-0.
20440.
1024-0.
02630.
0018-0.
3187-0.
01540.
147710.
220.
000.
000.
000.
860.
000.
100.
00Note:p-valuesareinitalicsEmploymentProtectionLegislationProductMarketRegulationProductMarketRegulationEmploymentProtectionLegislationshareofhighskillworkersintotallaborcomp.
shareofICTcapitalintotalcapitalcomp.
ICTcapitaldeepeningNon-ICTcapitaldeepeningNon-ICTcapitaldeepeningGrowthinshareofhoursbyhighskillworkersshareofICTcapitalintotalcapitalcomp.
productivitygrowthGrowthinshareofhoursbyhighskillworkersshareofhighskillworkersintotallaborcomp.
30Table4.
Multi-FactorProductivityLeadersIndustryNo.
1No.
2No.
3Agriculture1990NetherlandsUnitedStatesBelgium2003GermanyDenmarkBelgiumMining1990DenmarkBelgiumAustralia2003DenmarkUnitedKingdomAustraliaTotalManufacturing1990UnitedStatesBelgiumGermany2003UnitedStatesSwedenFinlandElectricalandOpticalEquipment1990NetherlandsAustriaSpain2003FinlandUnitedStatesFranceWholesale,retailtrade1990FranceNetherlandsGermany2003FranceNetherlandsFinlandTransport&storage1990NetherlandsAustraliaUnitedStates2003NetherlandsUnitedStatesAustraliaBusinessServices1990BelgiumNetherlandsDenmark2003BelgiumNetherlandsSweden31Figure6.
AustralianIndustriesMFPlevels(relativetoU.
S.
)Source:author'scalculationsGaptoU.
S.
TFPLevel(inpercent)AGRICULTURE-50%-45%-40%-35%-30%-25%-20%-15%-10%-5%0%19801983198619891992199519982001GaptoU.
S.
TFPLevel(inpercent)MINING-4%-2%0%2%4%6%8%10%12%14%19801983198619891992199519982001GaptoU.
S.
TFPLevel(inpercent)TOTALMANUFACTURING-60%-50%-40%-30%-20%-10%0%19811984198719901993199619992002GAPtoU.
S.
TFPlevel(inpercent)EGW-20%-10%0%10%20%30%40%50%19801983198619891992199519982001Source:author'scalculationsGaptoU.
S.
TFPLevel(inpercent)CONSTRUCTION-30%-25%-20%-15%-10%-5%0%5%10%15%19801983198619891992199519982001GaptoU.
S.
TFPLevel(inpercent)WHOLESALE&RETAILTRADE-35%-30%-25%-20%-15%-10%-5%0%19811984198719901993199619992002GaptoU.
S.
TFPLevel(inpercent)Transport&Storage-20%-10%0%10%20%30%40%50%60%70%19801983198619891992199519982001GaptoU.
S.
TFPLevel(inpercent)BUSINESSSERVICES-15%-10%-5%0%5%10%15%20%1980198319861989199219951998200132Table5.
ConvergenceofAustralianIndustries'TechnologyLevelinaPanelofOECDCountriesDependentvariable:annualMFPgrowth(1)(2)(3)(4)(5)(6)1980-19951980-1990TFPlevel,relativetoleader(t-1)-0.
0095-0.
020.
00003-0.
01[3.
11]***[5.
74]***[0.
02][2.
82]***TFPlevel,relativetoleader(t-1)-0.
008-0.
008(ICTusingindustries)[2.
30]**[2.
38]**TFPlevel,relativetoleader(t-1)-0.
013-0.
013(ICTproducers)[2.
28]**[2.
18]**TFPlevel,relativetoleader(t-1)-0.
010-0.
009(NonHighTechindustries)[3.
08]***[2.
92]***TFPlevel,relativetoleader(t-1)-0.
019-0.
017Australiaspecific(interactionwithdummyvariable)[2.
06]**[1.
53]Australia(relativetoaverage)0.
0110.
009(dummyvariableforAustralia)[2.
35]**[1.
25]ICTusinginAustralia0.
01[2.
64]***Observations448127584472162644814481R-squared1/0.
020.
020.
0010.
010.
020.
02CountryfixedeffectsYESYESNONOYESYESIndustry-timefixedeffectsYESYESYESYESYESYESRobusttstatisticsinbrackets,observationsareclusteredbycountry-year.
*significantat10%;**significantat5%;***significantat1%FtestofequalityofconvergenceparametersforICTusing,ICTproducingandNon-ICTsectors(Regression5):pvalue=0.
531/R-squaredexcludeindustry-yearfixedeffectsNote:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicatedPanelA.
Multi-FactorProductivity33Table5.
continuedDependentvariable:annualgrowthinICTcapitallaborratio(1)(2)(3)(4)(5)(6)1980-1990ICTcapital-laborratio,relativetoleader(t-1)-0.
006-0.
0030.
0030.
002-0.
003[4.
27]***[2.
15]**[1.
66][0.
57][2.
20]**ICTcapital-laborratio,relativetoleader(t-1)-0.
016-0.
016Australiaspecific[-4.
21]***[-3.
12]***ICTcapital-laborratio,relativetoleader(t-1)-0.
002(Non-ICTindustries)[-1.
49]ICTcapital-laborratio,relativetoleader(t-1)-0.
004(ICTusingindustries)[-2.
07]**ICTcapital-laborratio,relativetoleader(t-1)-0.
008(ICTproducingindustries)[-2.
04]**ShareofICTcapitalintotalcapital(t-1)-0.
002[0.
92]ShareofICTcapitalintotalcapital(t-2)-0.
0070.
006-0.
008-0.
006[3.
56]***[2.
83]***[-2.
34]***[-3.
47]***ShareofICTcapitalintotalcapital(t-2)-0.
004(Non-ICTindustries)[1.
96]*ShareofICTcapitalintotalcapital(t-2)-0.
01(ICTusingindustries)[3.
30]***ShareofICTcapitalintotalcapital(t-2)-0.
02(ICTproducingindustries)[5.
72]***Australiarelativetoaverage0.
0350.
020(dummyvariableforAustralia)[3.
61]***[1.
23]Observations451843784378176043784378R-squared1/0.
270.
270.
030.
030.
270.
27CountryfixedeffectsYESYESNONOYESYESIndustry-timefixedeffectsYESYESYESYESYESYESRobusttstatisticsinbrackets,observationsareclusteredbycountry-year.
*significantat10%;**significantat5%;***significantat1%FtestofequalityofconvergenceparametersforICTusing,ICTproducingandNon-ICTsectors:pvalue=0.
0001(Reg.
5),pvalue=0.
3522(Reg.
6)1/R-squaredexcludeindustry-yearfixedeffectsNote:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicatedPanelB.
ICTcapitaldeepening34Table6.
ImpactofProductMarketRegulationsonMFPGrowthDependentvariable:annualMFPgrowth(1)(2)(3)(4)(5)(6)(7)(8)1990-20031990-20031990-2003TFPlevel,relativetoleader(t-1)-0.
010-0.
013-0.
007-0.
014-0.
008-0.
009-0.
011-0.
005[3.
25]***[-3.
75]***[-1.
47][-3.
82]***[-1.
68]*[2.
76]***[-3.
17]***[-1.
08]TFPlevel,relativetoleader(t-1)0.
0080.
0100.
010InteractedwithPMR[1.
81]*[1.
91]*[1.
41]TFPlevel,relativetoleader(t-1)0.
0180.
021InteractedwithPMRinICTusingindustries[3.
04]***[2.
79]***TFPlevel,relativetoleader(t-1)0.
0060.
015InteractedwithPMRinICTproducingindustries[0.
65][0.
73]TFPlevel,relativetoleader(t-1)0.
0050.
004InteractedwithPMRinotherindustries[0.
65][0.
45]SectoralPMR(t-1)0.
008-0.
002-0.
0130.
003-0.
008[0.
80][-0.
16][-0.
87][0.
23][-0.
52]Highskilllaborshare(t-1)0.
00020.
00060.
00030.
00060.
00020.
0006[1.
43][2.
86]***[1.
52][2.
96]***[1.
09][2.
6]***ICTcapitalshare(t-1)0.
01200.
0153[1.
85]*[1.
82]*SectoralPMR(t-1)-0.
028-0.
025-0.
043inICTusing[2.
21]**[-1.
83]*[-2.
6]***SectoralPMR(t-1)0.
031-0.
068-0.
0828416inICTproducing[1.
08][-1.
87]*[-1.
59]SectoralPMR(t-1)0.
0160.
0200.
0136761innon-ICT[1.
38][1.
36][0.
72]Observations42143539223535392235421435392235R-squared0.
020.
030.
03490.
02550.
03770.
020.
0260.
037CountryfixedeffectsYESYESYESYESYESYESYESYESIndustry-timefixedeffectsYESYESYESYESYESYESYESYESRobusttstatisticsinbrackets,observationsareclusteredbycountry-year.
*significantat10%;**significantat5%;***significantat1%Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicated35Table7.
ImpactofLaborMarketInstitutionsonMFPGrowthDependentvariable:annualMFPgrowth(1)(2)(3)(4)(5)(6)1990-2003TFPlevel,relativetoleader(t-1)-0.
0005-0.
00067-0.
01002-0.
01092-0.
01093-0.
00251-0.
39[0.
34][3.
18]***[3.
12]***[3.
12]***[0.
65]EPL-0.
002[-2.
02]**EPL-0.
0023-0.
0015-0.
0015-0.
0015-0.
0019HighSkillindustries[2.
36]**[3.
04]***[3.
15]***[3.
16]***[3.
68]***EPL-0.
00095LowSkillindustries[1.
24]Highskilllaborshare(t-1)0.
000160.
000150.
00039[0.
79][0.
78][2.
44]**ICTcapitalshare(t-1)0.
003890.
00454[0.
31][0.
51]Observations448144814481389238922527R-squared0.
0020.
0020.
020.
020.
020.
03CountryfixedeffectsNONOYESYESYESYESIndustry-timefixedeffectsYESYESYESYESYESYESRobusttstatisticsinbrackets,observationsareclusteredbycountry*significantat10%;**significantat5%;***significantat1%Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicated36Table8.
DisentanglingtheEffectsofProductandLaborMarketInstitutionsonMFPGrowthDependentvariable:annualMFPgrowth(1)(2)(3)(4)(5)(6)(7)1990-20031990-2003TFPlevel,relativetoleader(t-1)-0.
0029-0.
0030-0.
0002-0.
0112-0.
0115-0.
0037-0.
0023[1.
46][1.
27][0.
08][3.
19]***[3.
29]***[0.
88][-0.
98]TFPlevel,relativetoleader(t-1)0.
0100.
006InteractedwithPMR[2.
25]**[1.
31]TFPlevel,relativetoleader(t-1)0.
0150.
0160.
0160.
0160.
015InteractedwithPMRinICTusingindustries[2.
45]**[2.
03]**[2.
68]***[2.
11]**[2.
47]**TFPlevel,relativetoleader(t-1)-0.
0040.
011-0.
0070.
002-0.
005InteractedwithPMRinICTproducingindustries[0.
36][0.
57][0.
73][0.
12][-0.
47]TFPlevel,relativetoleader(t-1)0.
0110.
0070.
0040.
0030.
009InteractedwithPMRinotherindustries[1.
74]*[0.
91][0.
52][0.
38][1.
46]TFPlevel,relativetoleader(t-1)-0.
028Australiaspecific[-1.
52]Australiarelativetoaverage0.
007[0.
7]EPL-0.
001-0.
001-0.
001[1.
37][1.
36][0.
71]EPL-0.
002-0.
002-0.
002-0.
002Highskillindustries[3.
70]***[3.
85]***[4.
04]***[-3.
48]***Highskillshare(t-1)0.
000030.
000030.
000070.
000160.
000190.
000450.
000054[0.
25][0.
28][0.
57][1.
17][1.
37][2.
75]***[0.
56]ICTshare(t-1)0.
0010.
001-0.
0040.
0040.
0040.
004-0.
002[0.
09][0.
14][0.
43][0.
62][0.
59][0.
48][-0.
29]Observations3679367923773679367923773679R-squared0.
010.
010.
010.
020.
030.
040.
01CountryfixedeffectsNONONOYESYESYESNOIndustry-timefixedeffectsYESYESYESYESYESYESYESRobusttstatisticsinbrackets,observationsareclusteredbycountrywhencountryfixedeffectsareomitted,andbycountry-yearinothercases*significantat10%;**significantat5%;***significantat1%Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicated37Table9.
ImpactofLaborandProductMarketInstitutionsonICTCapitalDeepeningDependentvariable:annualgrowthinICTcapitallaborratio(1)(2)(3)(4)(5)(6)(7)(8)ICTcapital-laborratio,relativetoleader(t-1)0.
0060.
004-0.
0030.
0040.
002-0.
002-0.
0020.
004[4.
2]***[3.
3]***[0.
73][3.
31]***[2.
31]**[-1.
21][-1.
49][1.
8]*ShareofICTcapitalintotalcapital(t-1)-0.
006-0.
0080.
003-0.
009-0.
0130.
003-0.
008-0.
008[-2.
92]***[-5.
08]***[0.
69][-4.
98]***[-8.
4]***[1.
29][-3.
98]***[-3.
35]***Australiarelativetoaverage0.
015[1.
43]SectoralPMR-0.
038-0.
009-0.
0130.
0700.
025[-5.
55]***[-1.
29][-1.
9]*[3.
12]***[1.
59]EPL-0.
022-0.
023-0.
025-0.
021[-10.
82]***[-10.
39]***[-11.
18]***[-3.
62]***EPL-0.
021Highskillindustries[3.
47]***EPL-0.
024Lowskillindustries[3.
77]***Observations41164378437841164116411641164378R-squared0.
020.
130.
130.
140.
270.
140.
280.
13CountryfixedeffectsNONONONOYESNOYESNOIndustry-timefixedeffectsNONOYESNONOYESYESYESRobusttstatisticsinbrackets,observationsareclusteredbycountrywhencountryfixedeffectsareomitted,andbycountry-yearinothercases*significantat10%;**significantat5%;***significantat1%Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicated38Table10.
PredictedImpactofProductandLaborMarketReforms1980-19901991-2003ChangeinannualMFP:ActualPredicted1/1/Predictedchangeisbasedonestimatesofregression(6)inTable6.
TFPgapsareaveragesover1980-82and1991-93respectivelyfor1980-1990and1991-2003Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicatedBottomquartileEPL1/TopquartileEPL1/DifferenceinannualMFPgrowth:ActualPredicted2/2/Predicteddifferenceisbasedonregression(7)ofTable7Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicated0.
58%PanelB.
PredictedimpactoflabormarketflexibilityonMFPgrowthinhighskillindustriesacrossOECDcountries0.
593.
611.
13%0.
68%(1990-2003).
0.
9%AnnualMFPgrowthEPLindexPMR(periodend)0.
220.
180.
4%0.
10%1/CountriesareatthebottomquartileEPLareAustralia,Canada,theUnitedKingdomandtheUnitedStates;countriesatthetopquartileEPLareGermany,ItalyandSpain.
.
.
42.
7%47.
0%0.
07%0.
17%.
1.
2%0.
44%PanelA.
PredictedimpactofproductmarketreformsonMFPgrowthinAustralianICTusingindustriesImpactofPMRinICTusingsectorsImpactof"passive"MFPconvergenceDistancetoMFPleader(beg.
period)AnnualMFPgrowth39Table11.
ControllingforOtherDeterminantsofMFPGrowthDependentvariable:annualMFPgrowth(1)(2)(3)(4)(5)(6)(7)1990-2003TFPlevel,relativetoleader(t-1)-0.
00878-0.
01983-0.
02009-0.
01174-0.
0043-0.
01163-0.
01326[1.
96]*[3.
97]***[4.
04]***[3.
36]***[1.
02][3.
31]***[2.
16]**TFPlevel,relativetoleader(t-1)0.
0340.
0280.
0280.
0160.
0180.
0160.
038InteractedwithPMRinICTusingindustries[3.
16]***[2.
59]**[2.
59]**[2.
80]***[2.
29]**[2.
62]***[3.
41]***TFPlevel,relativetoleader(t-1)0.
017730.
028330.
02974-0.
006870.
00362-0.
007320.
05778InteractedwithPMRinICTproducingindustries[0.
85][1.
35][1.
43][0.
68][0.
18][0.
72][1.
72]*TFPlevel,relativetoleader(t-1)0.
00377-0.
01321-0.
013660.
003870.
004490.
00351-0.
00709InteractedwithPMRinotherindustries[0.
50][1.
60][1.
67]*[0.
55][0.
50][0.
48][0.
75]EPLinhighskillindustries-0.
0019-0.
0013-0.
0013-0.
0018-0.
0023-0.
0018-0.
0016[3.
37]***[2.
40]**[2.
45]**[3.
91]***[4.
14]***[3.
94]***[2.
61]**Highskillshare(t-1)0.
000420.
00040.
000410.
00020.
000460.
000190.
00053[2.
30]**[2.
21]**[2.
28]**[1.
45][2.
85]***[1.
45][2.
55]**R&Dintensity(t-1)0.
001440.
00028[2.
21]**[0.
41]TFPlevel,relativetoleader(t-1)-0.
002-0.
002-0.
002InteractedwithR&Dintensity[5.
59]***[6.
13]***[4.
59]***Exportorientation(t-1)0.
001050.
002680.
001040.
00233[0.
81][1.
72]*[0.
79][1.
37]TradeweightedUSR&Dintensity(t-1)-0.
00007[0.
29]Observations2398239823983679237736791657R-squared0.
030.
040.
040.
030.
040.
030.
06CountryfixedeffectsYESYESYESYESYESYESYESIndustry-timefixedeffectsYESYESYESYESYESYESYESRobusttstatisticsinbrackets,observationsareclusteredbycountry-year*significantat10%;**significantat5%;***significantat1%Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicated40AppendixI.
TheImpactofProductandLaborMarketInstitutionsonLaborProductivityDependentvariable:annuallaborproductivitygrowth(1)(2)(3)(4)(5)(6)(7)(8)(9)1990-20031990-2003Laborproductivitylevel,relativetoleader(t-1)-0.
015-0.
004-0.
014-0.
019-0.
019-0.
017-0.
018-0.
016-0.
018[-5.
37]***[1.
32][5.
12]***[-6.
28]***[6.
27]***[4.
09]***[5.
52]***[3.
93]***[-5.
86]***Laborproductivitylevel,relativetoleader(t-1)0.
0110.
0100.
013InteractedwithPMR[2.
14]**[1.
50][1.
23]Laborproductivitylevel,relativetoleader(t-1)0.
0290.
0260.
029InteractedwithPMRinICTusingindustries[4.
41]***[3.
08]***[2.
76]***Laborproductivitylevel,relativetoleader(t-1)-0.
008-0.
005-0.
010InteractedwithPMRinICTproducingindustries[0.
79][0.
25][-0.
95]Laborproductivitylevel,relativetoleader(t-1)0.
0090.
0090.
008InteractedwithPMRinotherindustries[1.
23][0.
91][0.
98]EPL-0.
002-0.
002-0.
002-0.
002-0.
002Highskillindustries[1.
47][3.
62]***[3.
16]***[4.
63]***[-3.
51]***EPL0.
0006Lowskillindustries[0.
48]Highskillshare(t-1)0.
00030.
00020.
00040.
00030.
00040.
00060.
00040.
00060.
0004[2.
06]**[0.
93][2.
12]*[2.
2]**[2.
49]**[3.
28]***[1.
96]*[2.
26]**[2.
19]**ICTshare(t-1)0.
0150.
0130.
0120.
0170.
0170.
0240.
0140.
0230.
014[1.
98]*[0.
97][0.
86][2.
25]**[2.
29]**[2.
51]**[1.
00][1.
80]*[1.
14]Observations394239423942372437242397372423973724R-squared0.
030.
010.
040.
040.
040.
040.
040.
040.
04CountryfixedeffectsYESNOYESYESYESYESYESYESYESIndustry-timefixedeffectsYESYESYESYESYESYESYESYESYESRobusttstatisticsinbrackets,observationsareclusteredbycountry-year*significantat10%;**significantat5%;***significantat1%Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicated41AppendixII.
DroppingCountriesOnebyOne(1)(2)(3)(4)(5)TFPlevel,relativetoleader(t-1)-0.
010-0.
013-0.
014-0.
01093-0.
0112[3.
11]***[-3.
77]***[-3.
82]***[3.
12]***[3.
19]***TFPlevel,relativetoleader(t-1)0.
0090.
006InteractedwithPMR[1.
87]*[1.
31]TFPlevel,relativetoleader(t-1)0.
018InteractedwithPMRinICTusingindustries[3.
04]***TFPlevel,relativetoleader(t-1)0.
006InteractedwithPMRinICTproducingindustries[0.
65]TFPlevel,relativetoleader(t-1)0.
005InteractedwithPMRinotherindustries[0.
65]EPL-0.
001-0.
002Highskillindustries[3.
16]***[3.
70]***Highskillshare(t-1)0.
00020.
00030.
000150.
00016[1.
27][1.
52][0.
78][1.
17]ICTcapitalshare(t-1)0.
01250.
0120.
003890.
004[1.
92]*[1.
85]*[0.
31][0.
62]Observations44813539353938923679R-squared0.
020.
030.
030.
020.
02CountryfixedeffectsYESYESYESYESYESIndustry-timefixedeffectsYESYESYESYESYESEPLTFPlevel*PMRHighestp-valueHighskillind.
ICTusingind.
Coefficient-0.
0080.
0070.
012-0.
001-0.
002-0.
0020.
013T-stat[-2.
44]**[1.
32][1.
95]*[-2.
29]**[-2.
78]***[-2.
98]***[2.
3]**CountrydroppedGermanyAustraliaDenmarkItalyItalyItalyDenmarkSmallestp-valueCoefficient-0.
0110.
0140.
021-0.
002-0.
002-0.
0020.
019T-stat[-3.
41]***[2.
27]**[3.
52]***[-4.
76]***[-4.
43]***[-4.
54]***[3.
18]***CountrydroppedUnitedStatesJapanBelgiumFranceFranceFranceBelgiumRobusttstatisticsinbrackets,observationsareclusteredbycountry-year*significantat10%;**significantat5%;***significantat1%Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicatedThetablereports,foreachvariableofinterest,thecoefficient,tstatistics,andthecountrydropped,whenthecoeficientistheleastsignificant(highestpvalue)andthemostsignificant(lowestpvalue).
-0.
0115(6)0.
016[3.
29]***0.
000190.
004YESYES[1.
37][0.
59]36790.
03[2.
68]***[0.
73][0.
52][3.
85]***-0.
0070.
004-0.
00242AppendixIII.
Regressionswith3-yearAveragesDependentvariable:3yearsaverageMFPgrowth(overlappingpanel)(1)(2)(3)(6)(7)(8)1990-2003TFPlevel,relativetoleader(t-2)-0.
016-0.
017-0.
018-0.
018-0.
018-0.
014[4.
28]***[4.
18]***[4.
21]***[4.
30]***[4.
34]***[2.
85]***TFPlevel,relativetoleader(t-2)0.
0100.
009InteractedwithPMR[1.
78]*[1.
57]SectoralPMR(t-2)0.
00050.
0010-0.
0012-0.
0006-0.
0032[0.
04][0.
08][0.
10][0.
05][0.
21]TFPlevel,relativetoleader(t-2)0.
0130.
0140.
017InteractedwithPMRinICTusingindustries[1.
76]*[1.
79]*[1.
81]*TFPlevel,relativetoleader(t-2)0.
0030.
002-0.
010InteractedwithPMRinICTproducingindustries[0.
21][0.
12][0.
54]TFPlevel,relativetoleader(t-2)0.
0090.
0080.
010InteractedwithPMRinotherindustries[1.
59][1.
27][1.
30]EPL-0.
0016-0.
0017-0.
0018Highskillindustries[2.
98]***[3.
02]***[2.
52]**Observations425540024002400240022607R-squared0.
050.
050.
050.
050.
050.
07CountryfixedeffectsYESYESYESYESYESYESIndustry-timefixedeffectsYESYESYESYESYESYESDependentvariableisthe3-yearmovingaveragegrowthrateofMFPgrowth(yearst+1,tandt-1)Robusttstatisticsinbrackets,observationsareclusteredbycountry-industry*significantat10%;**significantat5%;***significantat1%Note:TFPlevelsrelativetoleaderareexpressedinLogs.
Regressionsareonthefullsample(1980-2003),unlessindicated