SlideShare una empresa de Scribd logo
1 de 43
A Cross Country Analysis
on Top Income Inequality
Kyel Governor
6/9/2016
Thispaperusescross-countrypanel datafrom1950-2014 to investigateif the rise intopincome
inequalitycanbe explainedbycapital accumulationwheninnovationandcross-countrydifferencesare
takenintoaccount. PooledOLSandfixedeffectsmodelsare usedtoinvestigatethe relationship. This
paperfindssome evidence infavourof capital accumulationdriving topincome inequalityforhighly
competitivecountries.Howeverthere isstrongerevidence suggestingthatinnovationrentsdrivetop
income inequalityratherthancapital accumulation.
1
Contents
Introduction.......................................................................................................................................2
1 Theory........................................................................................................................................3
1.1 Piketty’s General Laws of Capitalism.....................................................................................3
1.2 Schumpeterian Growth Theory and Innovation Rents ............................................................4
2 Data and Measurement...............................................................................................................5
2.1 Income Inequality................................................................................................................6
2.2 Innovation...........................................................................................................................7
2.3 Piketty’s r > g and K over Output...........................................................................................9
2.4 Control Variables.................................................................................................................9
2.5 Global Competitiveness Index (GCI) ......................................................................................9
2.6 Table of Variables and Descriptions.......................................................................................9
3 Main Empirical Analysis:The effectof innovativenessandcapital accumulationonincome
distribution ......................................................................................................................................11
3.1 Estimation Strategy............................................................................................................11
3.2 Pikettyr> g and Capital Accumulation:ResultsfromPooledOLSandFixedEffectsRegressions
11
3.3 SchumpeterianGrowthTheoryandInnovationRents:ResultsfromPooledOLSandFixed
Effects Regressions .......................................................................................................................16
3.3.1 Measure of Innovation: Patents Granted .....................................................................16
3.3.2 Measure of Innovation: Book to Market.......................................................................20
3.4 Evidence of Piketty’sr>gand Schumpeter’sInnovationRents:Interactionof Countries
Grouped by the Global Competitiveness Index ...............................................................................23
3.4.1 GCI Grouping Table.....................................................................................................24
3.4.2 Results.......................................................................................................................24
4 Discussion.................................................................................................................................27
4.1 Reverse Causality –Innovation on Income Inequality OLS Regression and Panel VAR..............27
4.2 Pre and Post Y2K Regressions – Evidence of a New Economy................................................31
5 Conclusion................................................................................................................................33
References.......................................................................................................................................34
Appendix..........................................................................................................................................36
2
Introduction
Increasingtopincome inequality indevelopedcountries hasbecome aphenomenoninthe worldtoday
and yetthere isno general agreement onthe rootcauses behind the rise in topincome inequality.This
paperdebateswhethertopincome inequalityisdrivenby capital accumulationorinnovation.Thomas
Piketty hasbroughtthe issue of growthand income redistribution backtothe forefrontof economic
debateswithhis (2014) book, Capitalin the Twenty-FirstCentury,where he argues thatcapital
accumulationbythe wealthiesttop1%is drivingtopincome inequality. Inshort, he defendsthat wealth
growsfasterthan economicoutput,centeringhisargument onthe expressionr> g (where ris the rate
of returnto capital andg isthe economicgrowthrate),andso slowergrowthleadsto a rapidincrease in
income inequality. Piketty’smore thansix-hundredpage publication inlarge partignores cross-country
differencesand the endogenousevolutionof technology anditsrelationshipwith topincome inequality
leavinghis worksusceptibletodispute (Acemogluetal,2015). Infact in the UnitedStates patentingand
top 1% income share followparallel evolutions indicatingthatinnovationmayplayarole inexplaining
the rise in top income inequality (Aghionetal,2015). Thispaperwill use cross-countrypanel datafrom
1950-2014 to investigate if the rise in topincome inequality canbe explainedby capital accumulation
wheninnovationandcross-countrydifferences are takenintoaccount.Twodifferentmeasuresare used
for innovation;total patentsgranted andthe booktomarketratio. The capital share of national income
(CSNI) ismeasuredby twovariables;the capital stocktogrossdomesticproduct ratioand r minusg.
Lastly, top income inequalityismeasuredbythe share of income heldbythe top1% and the top 10%,
and the ParetoLorenzcoefficient.
In summary,thispaperfinds some evidence validatingPiketty’sargumentthatcapital accumulation
drivestopincome inequality forhighly competitive1
countries.Howeverthere isstrongerevidence
suggestingthatinnovationrentsdrivetopincome inequalityratherthancapital accumulation.Excluding
the leastcompetitivecountries, measuresof innovation show asignificantrelationshipwith topincome
inequality where higherlevelsof innovationleadto highertop 1% and top10% income shareswhich
reflectinnovationrentsandthus isconsistentwithSchumpeteriangrowththeory. The modelpredicts
that on average a 1% increase inpatentsgrantedleadstoa 16.6% increase in the top1% share of
income anda 14.7% increase inthe top 10% share of income,aresultconsistentwithAghionetal.
(2015) U.S. cross-state analysisoninnovationandtopincome inequality. The booktomarketratio
measure of innovation alsoshowsthatonaverage innovationincreases the top1% and top 10% share of
income.Aftercontrollingforfixedeffects,regressionsshow that innovationmeasuressuchas R&D
expenditure andbooktomarketratio are significantinpredictingtopincome inequality. Furthermore,
resultsshownosignificantrelationshipbetween innovation orCSNI andthe restof the income
distribution. Wheninvestigatingwhetherincome inequality causesinnovation pooledOLSregressions
showthat on average the top10% share of income increasesinnovation andapanel vectorauto
regressionsupportsthisresult.Finally the paperfindsthatbefore the year2000 there is no evidence of
a positive relationshipbetween CSNI andtopincome inequalityandnorelationshipbetweeninnovation
1 As measured by the Global Competitive Index (2015-2016).See Section 3.4.
3
and topincome inequality.However afterthe year2000 there isa positive relationship forbothCSNI
and innovationwithtopincome inequality.
ThisanalysishasbeenmotivatedbyAcemogluandRobinson(2015) where they criticize general lawsof
capitalismproposedinPiketty’s(2014) recentwork. Contrary to Piketty,theyshow thatr> g has no
relationship withtopincome inequality. Mypaperadds to theirworkontop income inequalityby
addingan analysisof innovationdriveninequality.ThispapercloselyfollowsAghion,Akcigit, Bergeaud,
Blundell and Hémous(2015) U.S. cross-state analysisof innovationontopincome inequality exceptthis
papertakesa cross-country perspectiveandfactorsin the ParetoLorenz Coefficientasa measuresof
income inequality.Like Aghionetal.(2015),the paper dependsonthe Schumpteriangrowthmodel
presentedinAghion,AkcigitandHowitt (2013) in where innovationledgrowthis dependenton
innovation rentswhichare representedbyincreasesintopincome inequality.Therefore this
endogenous growthmodelleadsthe analysisontopincome inequalityandinnovation. Thisgrowth
model also allowsthatanincrease in topincome inequality causesanincrease ininnovation as
mentionedby Tselosis(2011) and Zweimuller(2000).So our analysisinvestigatesthislikelyevent firstby
univariate regressions andthenby apanel vectorauto regression (VAR) oninnovationandtopincome
inequality motivatedbyAtemsandJones(2015).Finally the paperallows thatthere hasbeen arecent
shifttoa newdigital economy aspresentedin Stiroh’s(2004) review andsothe return on innovation
and humancapital has dramaticallyincreased.Thishypothesisissupportedby HémousandOlsen (2014)
inwhichtheyargue that the digital economybenefitshighskilledlabourers andinnovatorsbecause of
the automationof lowskilledlabour.Thispaperlookstoinvestigate thishypothesis.
The paper isorganizedasfollows. Section1of the paperreviewsPiketty’s general lawsof capitalismand
Schumpteriangrowththeorywhere topincome inequalityisdrivenbyquality-improvinginnovations.
Section2 introducesthe datainvolvedanddescribesthe measuresof innovationandincome inequality
usedinthe paper.Section3 presentsthe mainfindings.Section4isa discussionontwoextensionsof
the analysis:first,aninvestigationintoareverse causalitywhere topincome inequalitydrives
innovation;andsecond, alookat pre Y2K andpost Y2K regressions.Section5concludes.
1 Theory
1.1 Piketty’s General Laws of Capitalism
Pikettyproposesatheoryon the longrun proclivitiesof capitalismin Capitalin the Twenty-firstCentury
(2014). He reliesonMarxianeconomics aspremise tocome to the conclusionthat the future of
capitalismisone where there ismassownershipof capital bythe wealthiestandtheirmainincome
comesfromcapital rents.
To come to his predictions, Piketty firstpresents two“fundamental laws”;
𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑠ℎ𝑎𝑟𝑒 𝑜𝑓 𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑖𝑛𝑐𝑜𝑚𝑒 = 𝑟 𝑥 (
𝐾
𝑌
),
𝐾
𝑌
= 𝑠/𝑔,
4
where 𝑟 isthe real interestrate, 𝐾 isthe capital stock, 𝑌 isGDP, 𝑠 is the savingrate and 𝑔 isthe growth
rate of GDP. As derivedinaSolowmodel the growthrate of capital stock 𝐾 isequal to saving 𝑠𝑌 ina
closedeconomyandso the ratio
𝐾
𝑌
overtime will approach 𝑠/𝑔due to economicgrowth. Pikettythen
combinesthe twoequationstoget;
𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑠ℎ𝑎𝑟𝑒 𝑜𝑓 𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑖𝑛𝑐𝑜𝑚𝑒 = 𝑟 𝑥 (
𝑠
𝑔
).
Assuming 𝑟 and 𝑠 are approximatelyconstant derivesPiketty’sfirstgeneral law;
1. The capital share of national income ishigh whengrowth islow.
His secondgeneral lawstates;
2. r > g,
whichdefinesthatthe real interestrate r isalwaysgreaterthanthe growthrate g inan economy. This
lawholdswhena countryisdynamicallyefficient,meaningthatithas maximizedconsumption inevery
periodoveraninfinite timeline,otherwise itmaynotbe the case that the real interestrate exceedsthe
growthrate.
Piketty’s thirdgeneral law proposesthat;
3. Income inequalitytendsto rise wheneverr > g.
The logicbehindthispropositionisthat capital income growsapproximatelyatthe rate r where national
income growsat the rate g and since capital income is distributedmainlytothe wealthiest;their
incomesgrowfasterthanthe per capita income thusresultingincapital-drivenincome inequality.
Accordingto Piketty’stheorytopincome inequalityrisesasthe capital share of national income
increases because capital incomeismainlydistributedamongthe wealthiest. Assumingris
approximately constant,capital share of national incomewillrise asK/Yincreases(change in capital
exceedschange in GDP) and/oras g decreases makingthe distance between randg greater. Thispaper
testsbothhypotheses inordertosubstantiate Piketty’sgenerallawsof capitalism.
1.2 Schumpeterian Growth Theory and Innovation Rents
The baseline Schumpeteriangrowthmodel asdevelopedinAghionetal.(2015) offersa theoretical
frameworkforhowinnovationincreasestopincome inequality. The mainpredictionsof the model asit
relatestothispaperare;
 Innovation increasestopincome inequality.
 Higherinnovationrentleadstoincreasedlevelsof innovation.2
2 This resultis further explored in Section 4.2 where the marginal effect of innovation on income inequality is
stronger after the IT market crashed in 2000.
5
To come to these predictionsthe baseline Schumpeteriangrowthmodelisdevelopedasfollows. Given
discrete time,the economyispopulatedbytwotypesof individualseveryperiod;capital ownerswho
ownthe firmsand the productionworkers. Individualslive onlyforone period andinthe subsequent
perioda newgenerationisbornwiththose whoare bornto firmowners inheritingtheirparents’firm
and the restwork as productionworkersunlesstheysuccessfully innovate andtakeoveranincumbent.
The followingCobb-Douglastechnologyproductionfunction definesthe productionof finalgoods 𝑌𝑡:
𝑙𝑛𝑌𝑡 = ∫ 𝑙𝑛
1
0
𝑦𝑖𝑡 𝑑𝑖
𝑦𝑖𝑡 = 𝑞𝑖𝑡 𝑙𝑖𝑡
where 𝑦𝑖𝑡 isa vector of intermediateinputsusedinfinal production, 𝑙𝑖𝑡 isthe labourusedtoproduce
the intermediateinputs,and 𝑞𝑖𝑡 isthe labourproductivity; all attime t.Wheneverinnovationtakes
place inany sector 𝑖 inperiod 𝑡 the innovatorgainsa technologicalleadof 𝜂 𝐻 overthe competition and
so the innovatorproducesthe intermediateinput 𝑦𝑖𝑡 withthe productionfunction:
𝑦𝑖𝑡 = 𝜂 𝐻 𝑞𝑖,𝑡−1 𝑙𝑖𝑡
If no newinnovationsoccurinthe followingperiod,the innovator’stechnologyispartlyimitatedby
competitorsandsothe innovator’stechnological leaddecreasesto 𝜂 𝐿 where 1 < 𝜂 𝐿 < 𝜂 𝐻.
Solvingthe model yieldsthe twofollowingequationspertinenttothe paper’sanalysis;
𝑀𝐶𝑖𝑡 = 𝑤 𝑡/𝑞𝑖, 𝑡, and, 𝑝𝑖,𝑡 = 𝑤 𝑡 𝜂𝑖𝑡/𝑞𝑖, 𝑡
where 𝜂𝑖,𝑡 𝜖 {𝜂 𝐻, 𝜂 𝐿}, 𝑀𝐶𝑖𝑡 isthe marginal cost of producingthe intermediate input, 𝑤 𝑡 isthe wagespaid
to labourinvolvedinproducingthe intermediate input 𝑖,and 𝑝𝑖,𝑡 isthe price charged forthe
intermediate input;all attime t. Thusthe profitearnedbythe innovatorisgivenby;
𝜋𝑖𝑡 = ( 𝑝𝑖𝑡 − 𝑀𝐶𝑖𝑡) 𝑦𝑖𝑡 =
( 𝜂𝑖𝑡 − 1) 𝑌𝑡
𝜂𝑖𝑡
> 0
The profitsearnedbyinnovatorsinthismodel are innovationrents.Furthermore if the costof
innovationisallowedinthismodel then the costof innovationforentrantscanbe greaterthan the cost
of innovationforincumbents (Aghionetal.,2015). This leads toentrybarriersand thusreduces
innovative activity since entrantsreceive lessprofitfrominnovationthe higherthe costof innovationis
for them.
2 Data and Measurement
The paper uses cross-country unbalanced panel datacoveringthe period1950-2014. Missingvalues
betweendatapointsare approximatedbyalinearinterpolation.The 27countriescoveredare as
follows:Argentina,Australia,Canada,China,Colombia,Denmark,Finland,France,Germany,India,
6
Indonesia,Ireland,Italy,Japan,Korea,Malaysia,Netherlands,New Zealand,Norway,Portugal,
Singapore,SouthAfrica,Spain,Sweden,Switzerland,UnitedKingdom, and UnitedStates.
2.1 Income Inequality
The data on the share of income heldbythe top 1%,top 10%, and on the ParetoLorenzCoefficient was
obtained fromthe World Top Income Database (Alvaredo,Atkinson,Piketty,andSaez’sWorldTop
IncomesDatabase at http://topincomes.parisschoolofeconomics.eu/).These are ourtopincome
inequalitymeasures. Dataonthe share of income ownedbythe bottom40% was obtainedfromthe
WorldBank Databank (http://databank.worldbank.org/data).The share of income ownedbythe middle
40%-90% wasobtainedbysubtractingthe top10% and bottom40% share from100.
Figure1showsthe evolutionof top1%income inequality forFrance,Germany,SouthAfrica,Sweden,
and UnitedStates. The graphclearlyillustratessignificantcrosscountrydifferences howevera
significantupwardtrendisnoticeable forall countriesaroundafter1990. Figure 2 showsthe top 10%,
middle 40%-90%,andbottom 40% income sharesovertime forthe UnitedStates and itis evidentthat
top income sharesincrease atthe expenseof the middle class.
Figure 1
5
101520
1950 1960 1970 1980 1990 2000 2010 2020
Year
France Germany
South Africa Sweden
United States
Top 1% Income Share, 1960-2014
7
Figure 2
2.2 Innovation
We obtainpatentsgranted bythe USPTO and EPOto obtain total of patentsgranted percountryby
inventor’scountryof residence. Thisdatawasobtainedfromthe OECDdatabase
(http://stats.oecd.org/).The datacoversthe time period1999-2014. The bookto marketratio(b/m) was
createdby dividingcapital stockbymarketcapitalization. Capital stock (K) wasobtainedfromPenn
WorldTables version8.1 (www.ggdc.net/pwt) andmarketcapitalizationwasobtainedfrom the World
Bank Databank. Thisproxyfor the bookto marketratio workswell.The correlationbetweenthe actual
U.S. bookto market ratio withKovermarketcapitalizationisequal to0.9452. A low bookto market
ratiovalue impliesthatavaluationabove the worthof physical assets.A marketvaluationabove
physical assets impliesvaluableintangible assetssuchasinnovativeactivitiesandsomovementinthe
bookto marketratio isa validmeasure of innovation.
Figure 3 showsthe evolutionof patentsgrantedforFrance,Germany,SouthAfrica,Sweden,andUnited
Stateswhile Figure 4illustrates the dynamicrelationshipbetween booktomarketand patentsgranted
for the UnitedStates.Figure 4 gives evidence of arelationship between the twomeasuresof innovation
especiallywhencomparing the three yearlagof patentsgrantedwith the bookto marketratio.An
increase inpatentsgranted (lagged3years) leadstoa fall inthe book to marketratio.
1020304050
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
Year
Top 10% Share of Income Mid 40%-90% Share of Income
Bottom 40% Share of Income
United States Income Distribution 1950 - 2014
8
Figure 3
Figure 4
0
50
100150
2000 2005 2010 2015
Year
France Germany
South Africa Sweden
United States
Total Patents Granted, 1999-2014
.02.025.03.035.04
BooktoMarket
80
100120140160
2000 2005 2010 2015
Year
Granted patents (in 100s)
Granted patents (in 100s), Lagged 3 yearsBook to Market
United States, Book to Market vs Patents Granted
9
2.3 Piketty’s r > g and K over Output
Data on real interestrates (r) were retrievedfromthe WorldBankDatabankand data on GDP was
obtainedfromPennWorldTablesversion8.1and Madison’sdataset
(http://www.ggdc.net/maddison/maddison-project/data.htm).The GDPgrowthrate (g) iscomputed
and the variable r> g is computedbysimplysubtractinggfromr. Our capital to outputratio (K/Y) isalso
computedusingKand GDP data.
2.4 Control Variables
The paper controlsforthe size of the financial sector andthe size of the governmentsectorbecause
theyare likelyto have directeffectsoninnovationandtopincome share (Aghionetal,2015).
Furthermore the analysiscontrolsfor macroeconomicvariablessuchaspopulationgrowth,the
employmentrate, andhumancapital;andResearchand Development(R&D) variables suchasR&D
expenditure percapitaandnumberof R&D researchers. Controllingforthe employmentrate ismainly
to reduce time varyingcrosscountrydifferencesinordertohave more efficientresults. Humancapital is
controlled forsince thislikelyhasaneffectonincome inequalitymeasuresandinnovation. FinallyR&D
iscontrolledforsince itiscorrelatedwithpatentsgrantedandincome inequalitymeasures;alsoR&D
may be consideredameasure of innovation.
2.5 Global Competitiveness Index (GCI)
The paper groupscountriesbasedontheirGCI obtainedfromthe WorldEconomicForum Global
CompetitivenessReport (http://reports.weforum.org/global-competitiveness-report-2015-2016/) in
orderto see the relationshipof CSNIandinnovation ontopincome inequality acrossdifferenteconomic
stages.The GCI is a composite of sub-indexes suchasinstitutions,infrastructure, macroeconomic
environment,healthandprimaryeducation,highereducationand training, goodsmarketefficiency,
labourmarket,financial marketdevelopment,technological readiness,marketsize,business
sophistication,andinnovation. The table inSection3.4.1showshow countriesare grouped.
2.6 Table of Variables and Descriptions
Variable Number of
Observations
(annual)
Mean Standard
Deviation
Time
Span
Retrieval Description
Income Inequality Variables
Top 1% Share
of Income (%)
1177 12.057 3.425 1950-
2014
World Top
Incomes
Database
Share of national
income held by the
top 1% earners
Top 10%
Share of
Income (%)
962 36.6 6.809 1950-
2014
World Top
Incomes
Database
Share of national
income held by the
top 10% earners
Pareto Lorenz
Coefficient
1068 1.844 .224 1950-
2014
World Top
Incomes
Database
Measure of the
income distribution
Piketty and Innovation Variables
Patents
Granted
416 8.872 22.643 1999-
2014
OECD
Database
Number of patents
granted by the EPO
10
(in 1000s) and USPTO
Market
Capitalization
(% of GDP)
799 93.502 45.774 1975-
2014
World
Bank
Databank
Share price x number
of sharesoutstanding
for listed domestic
companies
r (%) 946 2.736 3.554 1961-
2014
World
Bank
Databank
Real interest rate
g (%) 1620 1.224 2.103 1951-
2010
Madison’s
dataset
GDP growth rate
K (in mil.
2005US$)
1644 8724524 1.27e+07 1950-
2011
Penn
World
Tables
Capital Stock
GDP (in mil.
2005US$)
1644 2856685 4178096 1950-
2011
Penn
World
Tables
Real GDP
Control Variables
R&D per
capita
expenditure (%
of GDP)
346 42289.59 20903.95 1996-
2010
World
Bank
Databank
Expenditures for
research and
development both
public and private
Number of
R&D
researchers
391 3345.405 1427.127 1996-
2013
World
Bank
Databank
Professionals
engaged in the
creation of new
knowledge,products,
processes, methods,
or systems
Index of
Human Capital
per person
1644 3.133 .337 1950-
2011
Penn
World
Tables
Based on years of
schoolingandreturns
to education
Share of
Government
Consumption
(%)
1644 .139 .034 1950-
2011
Penn
World
Tables
Share of government
consumption at
current PPPs
Services per
capita (% of
GDP)
1014 68.641 8.288 1960-
2014
World
Bank
Databank
Services correspond
to ISICdivisions50-99
(includes financial)
Employment
to population
ratio, 15+, total
(%) (national
estimate)
817 58.957 8.937 1980-
2014
World
Bank
Databank
Proportion of a
country's population
that is employed
Population
growth (annual
%)
1484 .989 .467 1950-
2011
Penn
World
Tables
Population growth
rate
Global N/A N/A N/A 2015-
2016
World
Economic
Integrates
macroeconomic and
11
Competitivene
ss Index
(1-7 (best))
Forum micro/business
aspects of
competitiveness into
a single index
3 Main Empirical Analysis:The effect of innovativeness and capital
accumulationon income distribution
3.1 Estimation Strategy
The main empirical analysis involveslookingatwhetherPiketty’svariables,r-g andcapital accumulation
measuredbyK/Y,or innovationmeasuredby b/mandtotal patentsgranted are responsible for
increasesintopincome shares. The mainestimatedequationis:
log( 𝑦𝑖𝑡) = 𝐴 + 𝐵𝑖 + 𝐵𝑡 + 𝐵1𝑗( 𝑃𝑖𝑘𝑒𝑡𝑡𝑦𝑗) + 𝐵2 𝑙𝑜𝑔(𝑖𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛) + 𝐵3𝑖 𝐺𝐶𝐼 𝑘−1 ∗ 𝑙𝑜𝑔(𝑖𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛)
+ 𝐵4𝑗𝑖 𝐺𝐶𝐼 𝑘−1 ∗ ( 𝑃𝑖𝑘𝑒𝑡𝑡𝑦𝑗) + 𝐵5 𝑋𝑖𝑡 + 𝜀𝑖𝑡
where 𝑦𝑖𝑡 isthe measure of inequality, 𝐵𝑖 isacountry fixedeffect, 𝐵𝑡 isa yearfixedeffect, 𝑃𝑖𝑘𝑒𝑡𝑡𝑦𝑗 isa
vectorof Pikettyvariableswhich include r-g(j=1) andK/Y(j=2), 𝑖𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛isinnovativeness measured
by b/mor total patentsgranted, 𝐺𝐶𝐼 𝑘−1 are GCI dummies and 𝑋 isa vector of control variables.
In Section 3.2 𝐵2, 𝐵3 and 𝐵4 are setto zero and a seriesof Pikettyregressionsare estimatedby pooled
OLS and fixedeffects.The restrictionon 𝐵2 isremoved inSection3.3and the model isestimatedagain
by pooledOLS andfixedeffects. InSection3.4all restrictionsare removedandthe model isestimated
by fixedeffects. All resultsare clusterbootstrappedand include time dummies.3
3.2 Piketty r > g and Capital Accumulation: Results from Pooled OLS and
Fixed Effects Regressions
The resultsfroma seriesof Pikettyregressionsare presentedbelowinTable 1.In (3) and(4) of Table 1 it
isassumedthat all capital marketsare openand cross countryvariationinr iszero andso r is
normalizedto0, therefore we have r-g=- g. The regressionsshow nosignificantrelationshipbetween
the top 1% share of income andr > g for pooledOLSand fixedeffectsestimates asseeninAcemogluand
Robinson(2015). ResultsfromTable 1 illustrate thatcrosscountrydifferences have astronginfluence
on Piketty’s r> g since fixedeffectscoefficientsare significantlylessthanpooledOLScoefficients.The
lastcolumnin Table 1 estimatesthatthe coefficientonPiketty’sK/Yissignificantbutnegativeinafixed
effectsregression.Thisimpliesthat anincrease inCSNIdecreasestopincome inequalitywhich
contradictsPiketty’s theory.
To continue withthe resultsfoundin(6) of Table 1 the nexttwotable shows the resultsfrom
regressions againstdifferentcontrolsandfordifferenttimespans. The resultsshow that K/Yremains
negative andsignificantafteraddingthe controls howeveroverthe period1999-2010 itbecomes
3 Additional regressionsarekept in the appendix.
12
insignificantasshownin(4) of Table 3. Overthe same periodin(3) of Table 3 r > g ispositive and
significant.
Table 1
Piketty's r > g and Capital Accumulation
1961-2010 1951-2010 1961-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
r - g 0.00921
(1.34)
-0.00246
(-1.02)
0.0103
(1.43)
0.00122
(0.48)
normalizing
r=0 (r-g=-g)
0.00818
(0.96)
-0.00155
(-0.79)
K/Y (Log) -0.125
(-0.64)
-0.536**
(-2.36)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 829 829 1339 1339 829 829
R2 0.284 0.202 0.293
R2 overall 0.233 0.194 0.173
R2 between 0.152 0.173 0.0217
R2 within 0.561 0.462 0.617
# of countries
in regression
27 27 27 27 27 27
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
13
Table 2
Piketty's r > g and Capital Accumulation with different controls
1961-2010
(1) (2) (3) (4)
Top 1% Share
of Income
(Log)
Top 1% Share
of Income
(Log)
Top 1% Share
of Income
(Log)
Top 1% Share
of Income
(Log)
r - g 0.00184
(0.77)
0.00221
(0.92)
0.00120
(0.47)
0.00259
(1.15)
K/Y (Log) -0.498**
(-2.46)
-0.487**
(-2.56)
-0.559**
(-2.45)
-0.506***
(-2.85)
Index of human
capital per person
(Log)
-0.752*
(-1.82)
-0.605
(-1.55)
Share of
Government
Consumption
-1.930***
(-3.22)
-1.762***
(-3.21)
Population growth
(annual %)
-0.0182
(-0.89)
-0.0369*
(-1.94)
Country Fixed
Effects
Yes Yes Yes Yes
Observations 829 829 829 829
R2 overall 0.193 0.358 0.155 0.331
R2 between 0.0544 0.214 0.0120 0.187
R2 within 0.643 0.651 0.618 0.672
# of countries in
regression
27 27 27 27
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
14
Table 3
Piketty's r > g and Capital Accumulation over different time spans
1980-2010 1999-2010
(1) (2) (3) (4)
Top 1% Share
of Income
(Log)
Top 1% Share
of Income
(Log)
Top 1% Share
of Income
(Log)
Top 1% Share
of Income
(Log)
r - g 0.0106
(1.34)
0.000222
(0.09)
0.0151*
(1.83)
-0.00308
(-1.52)
K/Y (Log) -0.0937
(-0.48)
-0.394**
(-2.21)
-0.0992
(-0.45)
-0.0203
(-0.09)
Country Fixed
Effects
No Yes No Yes
Observations 681 681 235 235
R2 0.284 0.101
R2 overall 0.200 0.0174
R2 between 0.0449 0.0110
R2 within 0.624 0.165
# of countries in
regression
27 27 27 27
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
In Table 4 all the control variables are includedandthe Pikettyvariablesare regressedonseveral
measuresof topincome inequality.The regressions show thatincreasingr-gincreasesthe top1% share
of income asseenin (1) of Table 4 butwhencountryfixedeffectsare includedthe coefficientbecomes
insignificant.In(6) of Table 4 K/Y has a significantandpositiverelationshipwiththe ParetoLorenz
coefficientandsoan increase inK/Ydecreasesincome inequality whichiswhatwasfoundinTable 1 to
3. This leadstoconflictingresultsintermsof analyzingPiketty’sargument.Favourable resultsforPiketty
are foundinthe pooledOLS regressionsbut notinfixedeffectsregressions.
15
Table 4
Piketty's r > g and Capital Accumulation – Top Income Inequality
1980-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
r - g 0.00904**
(2.06)
0.000327
(0.15)
0.00433
(1.03)
-0.000450
(-0.32)
-0.00850**
(-1.99)
-0.000450
(-0.32)
K/Y (Log) -0.116
(-0.67)
-0.380
(-1.38)
-0.110
(-0.65)
0.110
(0.46)
0.179
(1.13)
0.379**
(1.99)
Index of
human capital
per person
(Log)
-0.680*
(-1.75)
-0.673
(-1.06)
-0.293
(-0.89)
-0.132
(-0.24)
0.234
(0.69)
0.197
(0.42)
Share of
Government
Consumption
-4.549***
(-3.89)
-1.896***
(-2.68)
-2.050***
(-2.73)
-0.232
(-0.45)
2.592***
(3.43)
1.172**
(2.01)
Population
growth (annual
%)
0.0549
(1.05)
-0.0361
(-1.60)
-0.00665
(-0.13)
-0.0144
(-0.72)
-0.0144
(-0.33)
0.0125
(0.58)
Services per
capita (% of
GDP)
-0.00134
(-0.13)
-0.00357
(-0.08)
-0.00235
(-0.28)
-0.00361
(-0.07)
0.00128
(0.19)
-0.00387
(-0.20)
Employment to
population
ratio, 15+, total
(%) (national
estimate)
-0.00987*
(-1.75)
-0.00485
(-0.79)
-0.00943**
(-2.21)
-0.00218
(-0.76)
-0.00173
(-0.28)
0.00206
(0.40)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 502 502 444 444 450 450
R2 0.713 0.597 0.539
R2 overall 0.509 0.283 0.366
R2 between 0.468 0.233 0.227
R2 within 0.642 0.538 0.659
# of countries
in regression
26 26 23 23 24 24
t statistics in parentheses
16
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
3.3 Schumpeterian Growth Theory and Innovation Rents: Results from
Pooled OLS and Fixed Effects Regressions
So far our analysisprovides conflictingevidence forPiketty’sargumentonCSNI drivingtopincome
inequality. Basedonfixedeffectsthe analysisis inagreementwithAcemogluandRobinson(2015).In
thissectioninnovationisallowedinthe model andisfirstmeasuredby total patentsgranted andthen
by bookto market.The resultsmetfrombothmeasuressupport Schumpeteriangrowththeorybutdoes
not debunkPiketty’sargumentforcapital accumulationincreasingtopincome inequality.
3.3.1 MeasureofInnovation:PatentsGranted
Table 5 displays the results whentotal patentsgrantedis includedinthe regression. The pooledOLS
regression ontop10% share of income showssignificantresults findinga1% increase inpatentsgranted
increasesthe top10% share of income by6.6%. Alsointhe table,(1) shows r-gis positive andsignificant
and K/Y becomesinsignificantforall regressions givingevidence thatinnovationisanomittedvariable in
Section3.2 regressions.
The resultsinTable 5 give some evidence insupportof Schumpeteriangrowththeoryandinnovation
rentsand evidence againstPiketty’s supposition.Howeverthe regressionsinTable 5covera short time
span, 1999-2010, and so general inferencescannotbe drawn.
To ignore the insignificantcoefficientsonpatentsgrantedin (1) and(2) of Table 5 wouldbe unjust since
the top 1% share of income is one of the income inequalitymeasuresof interest. There isanapparent
omittedvariable whichis R&Dactivitysince thisisverylikelycorrelatedwithpatentsgrantedand can
likelybe correlatedwithourdependentvariables. R&Dactivityapproximatesthe total costof
innovations.AccordingtoSchumpeteriangrowththeoryasoutlinedinSection1.2,innovatorswill
produce an innovation if the expectedtotal profit4
fromaninnovation isequal toorgreaterthan the
expected costof innovation. If aninnovation,measuredbypatentsgranted,occurs thenthe innovator
has incurredthe cost of innovation whichisapproximatedbyR&Dactivity. Therefore R&Dactivity
affectspatentsgrantedwhere higherR&Dactivityleadstoan increase inpatentsgranted.R&Dactivity
can alsoaffectinnovationrents since the costof innovationhasanimpacton innovationrents. R&D
activitycan thushave a positive effectoninnovationrentsif the returnonR&D activityispositive anda
negative effectif the returnonR&D activityisnegative.
Table 6 controlsforR&D activityand providesasignificantand twice aslarge coefficientforpatents
grantedon the top 10% share of income pooledOLSregression thanwhatwaspresentedinTable 5,
4 Profitas defined by the equation 𝜋𝑖𝑡 = ( 𝑝𝑖𝑡 − 𝑀𝐶𝑖𝑡
) 𝑦𝑖𝑡 in Section 1.2. Expected total profitis the expected
present valueof the stream of all profits earned up to time T. See Aghion, Akcigitand Howitt (2013).
17
howeverregressionsonthe top1% share of income remaininsignificant.5
The keyresultinTable 6is in
(1) where r-gremains positive andsignificantinexplainingthe top 1% share of income.
Table 5 – Check appendix Table A1 for complete list of excluded countries.
Innovation Measure: Patents Granted – Top Income Inequality
1999-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
Patents
Granted (Log)
0.0243
(0.45)
0.0368
(0.55)
0.0664**
(2.19)
0.0410
(0.79)
-0.00404
(-0.15)
-0.0186
(-0.63)
r - g 0.0107*
(1.85)
-0.00403
(-1.29)
0.00422
(0.83)
-0.000130
(-0.05)
-0.00636
(-1.38)
0.00313**
(2.36)
K/Y (Log) -0.156
(-0.62)
-0.0816
(-0.19)
-0.234
(-1.07)
-0.0103
(-0.05)
0.119
(0.70)
-0.0280
(-0.11)
Index of
human capital
per person
(Log)
-0.811
(-0.99)
1.784
(1.05)
-0.594
(-0.98)
0.831
(1.01)
0.106
(0.23)
-0.665
(-1.02)
Share of
Government
Consumption
-3.994**
(-2.24)
-0.151
(-0.13)
-1.206
(-0.99)
-0.508
(-1.41)
2.163**
(2.25)
-0.370
(-0.79)
Population
growth
(annual %)
0.0954
(1.28)
-0.0113
(-0.28)
0.0551
(1.07)
-0.000113
(-0.01)
-0.0214
(-0.55)
-0.000475
(-0.02)
Services per
capita (% of
GDP)
-0.00186
(-0.09)
0.0961
(1.49)
0.00370
(0.31)
0.0391
(1.20)
0.00350
(0.39)
-0.0680***
(-3.05)
Employment
to population
ratio, 15+,
total (%)
(national
estimate)
-0.00779
(-0.95)
0.00377
(0.38)
-0.0100
(-1.56)
-0.00575
(-1.36)
-0.00516
(-0.72)
-0.00766
(-1.43)
5 By addingR&D activity to the regressions a few countries were dropped however Table 5 results were consistent
when the same countries were excluded and therefore the results arenot tainted by any selection bias.
18
Country Fixed
Effects
No Yes No Yes No Yes
Observations 220 220 202 202 202 202
R2 0.516 0.574 0.388
R2 overall 0.0623 0.00183 0.000000567
R2 between 0.0437 0.000400 0.00327
R2 within 0.255 0.270 0.370
# of countries
in regression
25 25 23 23 24 24
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
19
Table 6 - Check appendix Table A1 for complete list of excluded countries.
Innovation Measure: Patents Granted controlling for R&D – Top Income Inequality
1999-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
Patents
Granted (Log)
0.105
(1.39)
0.0367
(0.54)
0.122***
(2.90)
0.0447
(0.68)
0.0459
(0.94)
-0.00978
(-0.33)
R&D per
capita
expenditure
(Log)
-0.362
(-1.51)
0.136
(1.28)
-0.355***
(-2.58)
0.0360
(0.37)
-0.211
(-1.34)
-0.103
(-1.54)
R&D
researchers
(Log)
0.246
(1.00)
-0.108
(-1.18)
0.283**
(2.50)
-0.0496
(-0.75)
0.130
(0.86)
0.0746
(1.16)
r - g 0.0129**
(2.01)
-0.00294
(-0.89)
0.00455
(0.97)
-0.000283
(-0.10)
-0.00665
(-1.44)
0.00304*
(1.96)
K/Y (Log) -0.141
(-0.54)
-0.0495
(-0.12)
-0.196
(-1.03)
0.0442
(0.20)
0.131
(0.71)
-0.0704
(-0.28)
Index of
human capital
per person
(Log)
-0.950
(-0.91)
2.372
(1.60)
-0.582
(-1.03)
1.078
(1.05)
-0.0223
(-0.04)
-0.949
(-1.12)
Share of
Government
Consumption
-4.648***
(-2.74)
-0.289
(-0.55)
-1.801*
(-1.90)
-0.328
(-0.92)
1.910**
(2.00)
-0.456
(-0.75)
Population
growth (annual
%)
0.107
(1.34)
-0.0106
(-0.27)
0.0654
(1.40)
-0.000541
(-0.03)
-0.0164
(-0.46)
-0.00186
(-0.09)
Services per
capita (% of
GDP)
0.00965
(0.44)
0.0937*
(1.78)
0.00941
(0.77)
0.0383
(1.09)
0.0102
(0.84)
-0.0650**
(-2.47)
Employment
to population
ratio, 15+,
total (%)
-0.00521
(-0.51)
-0.00369
(-0.60)
-0.00973
(-1.50)
-0.00499
(-1.22)
-0.00343
(-0.44)
-0.00421
(-0.89)
20
(national
estimate)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 209 209 197 197 192 192
R2 0.625 0.707 0.484
R2 overall 0.0597 0.00287 0.0153
R2 between 0.0231 0.000901 0.0176
R2 within 0.359 0.274 0.411
# of countries
in regression
24 24 23 23 23 23
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
3.3.2 MeasureofInnovation:Bookto Market
As statedinthe previoussubsection,there isashorttime spanof available dataforpatentsgrantedand
so the advantage gainedfromusingthe booktomarket ratioas a measure of innovationisalongertime
span. The disadvantage,however,isthat b/misan indirectmeasure of innovative activity since itdoes
not directly capture innovative activitylike the more commonlyusedmeasure patentsgranted, but
insteaditapproximates the aggregate value of innovation. AccordingtoSchumpeteriangrowth theory
as presentedinSection1.2,the value of an innovationdecreasesovertime whichcorrespondstoan
increasingb/m.If a newinnovation occurs ina givenperiod thenthe value of thatinnovation isgreater
than the one that precededit andso there will be acorrespondingdecreasein the b/m.
The results fromregressingincome inequalitymeasureson b/mare displayedin Table 7.The regressions
inTable 7 coverthe period1980-2010 whichisthe same periodcoveredinthe Pikettyregressions in
Section3.2. The table predicts thatboth r-g andb/m contribute to the top 1% share of income however
r-g affectsthe top1% share of income ina pooledOLSregressionwhereasb/maffectsthe top1% share
of income ina fixedeffectsregression.Betweenthe resultsinSection3.3.1and Table 7 there is some
evidence thatinnovation increasestopincome inequalityandalsothatina pooledOLSthere isa
positive andsignificantrelationship betweenr-gandthe top 1% share of income.
Kitchensinkregressions6
ontopincome inequalityare performedandthe resultsare capturedinTable
8. The mostsignificantresultsare foundfor pooledOLS regressionon the top1% and top10% share of
income.Bothinnovationmeasuresare significantinthe correctdirectionindicatingthatinnovation
increasestopincome inequalityandmore sofor the top 1% share of income thanthe top 10% share of
income. Furthermore the Pikettyvariable r-gbecomesinsignificant. Table 8providessome more
evidence thatinnovationrentsexplain topincome inequalityratherthanthe Pikettyvariables.However
the resultsare not robustto countryfixedeffects.
6 Kitchen sink regression includes all variables of interestand controls.
21
Table 7 - Check appendix Table A1 for complete list of excluded countries.
Innovation Measure: Book to Market – Top Income Inequality
1980-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
Book to
Market (Log)
-0.0447
(-0.89)
-0.0578*
(-1.78)
-0.0542
(-1.43)
-0.0149
(-0.80)
0.000460
(0.01)
0.00975
(0.51)
r - g 0.00763*
(1.65)
-0.000391
(-0.11)
0.00436
(1.08)
0.00150
(0.67)
-0.00202
(-0.45)
0.00163
(0.54)
Index of
human capital
per person
(Log)
-1.051**
(-2.36)
-1.220**
(-2.04)
-0.299
(-1.00)
-0.340
(-0.71)
0.215
(1.00)
0.517
(1.12)
Share of
Government
Consumption
-3.669***
(-3.18)
-1.340
(-1.50)
-1.765**
(-2.08)
-0.0889
(-0.18)
1.852***
(3.06)
1.320
(1.40)
Population
growth
(annual %)
0.0436
(0.91)
-0.0137
(-0.55)
-0.0215
(-0.41)
-0.0109
(-0.56)
-0.0186
(-0.50)
0.000551
(0.02)
Services per
capita (% of
GDP)
-0.00677
(-0.93)
-0.0146
(-0.29)
-0.00271
(-0.39)
-0.00653
(-0.12)
0.00373
(0.75)
0.00459
(0.20)
Employment
to population
ratio, 15+,
total (%)
(national
estimate)
-0.00382
(-0.77)
-0.00816
(-1.26)
-0.00966**
(-1.98)
-0.00413*
(-1.78)
-0.00723**
(-1.99)
0.00440
(0.76)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 433 433 408 408 388 388
R2 0.723 0.593 0.600
R2 overall 0.582 0.430 0.370
R2 between 0.444 0.362 0.155
R2 within 0.670 0.550 0.648
# of countries
in regression
25 25 23 23 23 23
22
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
Table 8 - Check appendix Table A1 for complete list of excluded countries.
Kitchen Sink – Top Income Inequality
1999-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
Book to
Market (Log)
-0.138*
(-1.72)
-0.0608
(-0.89)
-0.116*
(-1.93)
-0.0338
(-1.07)
0.0521
(0.90)
-0.0184
(-0.43)
Patents
Granted (Log)
0.166**
(2.33)
0.0408
(0.50)
0.147***
(3.93)
0.0562
(0.85)
0.0199
(0.40)
0.00391
(0.11)
R&D per
capita
expenditure
(Log)
-0.659***
(-2.66)
0.0508
(0.24)
-0.517***
(-4.04)
-0.0274
(-0.24)
-0.0921
(-0.55)
-0.189*
(-1.75)
R&D
researchers
(Log)
0.413*
(1.90)
-0.0785
(-0.64)
0.368***
(3.69)
-0.0358
(-0.56)
0.0778
(0.56)
0.0833
(0.95)
r - g 0.0111
(1.48)
-0.00428
(-0.78)
0.00280
(0.67)
-0.0000555
(-0.02)
-0.00373
(-0.69)
0.00365
(1.61)
Index of
human capital
per person
(Log)
-0.873
(-1.06)
2.269
(1.26)
-0.256
(-0.51)
0.952
(1.02)
-0.103
(-0.23)
-0.828
(-0.70)
Share of
Government
Consumption
-3.661**
(-2.15)
-0.144
(-0.23)
-1.400
(-1.37)
-0.265
(-0.72)
1.362*
(1.67)
-0.0669
(-0.11)
Population
growth (annual
%)
0.0658
(0.92)
-0.00620
(-0.18)
0.0324
(0.70)
0.00180
(0.11)
0.00152
(0.05)
0.00275
(0.13)
23
Services per
capita (% of
GDP)
0.0162
(0.98)
0.113**
(2.15)
0.0152
(1.43)
0.0475
(1.61)
0.00586
(0.62)
-0.0610**
(-2.01)
Employment
to population
ratio, 15+,
total (%)
(national
estimate)
0.00337
(0.45)
-0.00219
(-0.27)
-0.00837
(-1.45)
-0.00390
(-0.82)
-0.00734
(-1.18)
-0.00424
(-0.81)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 195 195 189 189 178 178
R2 0.666 0.744 0.463
R2 overall 0.0505 0.0166 0.00676
R2 between 0.0324 0.00713 0.0112
R2 within 0.368 0.286 0.442
# of countries
in regression
24 24 23 23 23 23
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
3.4 Evidence of Piketty’s r>g and Schumpeter’s Innovation Rents:
Interaction of Countries Grouped by the Global Competitiveness Index
The countriesinthe panel have verydifferenteconomies andsosuch differencesshouldbe considered.
The fixedeffectsestimatorisappropriate forcapturingthese differences.Howeveritisimportantto
recognize thatdifferenteconomieswillreactdifferentlytochangesininnovationandPikettyvariables.
Ratherthan interactingeachcountrywitheachmeasure of innovationandeachPikettyvariable the
countriesare firstgroupedintermsof theircompetitivenessasmeasuredbythe GCI and then the
interactionbetweenthe variablesof interestandthe GCIgroupsare estimated.Fourgroupsare created
as seeninthe table belowwhere GCI0isthe group withthe mostcompetitive countriesandGCI3isthe
groupwiththe leastcompetitive countries.
24
3.4.1 GCIGroupingTable
Global Competitive Index (2015-2016) Categories
Label Value Countries Total
GCI0 GCI ≥ 5.5 Switzerland, Singapore, United States, Germany,
Netherlands, Japan, Finland
7
GCI1 5.4 ≥ GCI ≥ 5 Sweden, United Kingdom, Norway, Denmark,
Canada, New Zealand, Malaysia, Australia, France,
Ireland, Korea Rep.
11
GCI2 4.9 ≥ GCI ≥ 4.5 China, Spain, Indonesia, Portugal, Italy 5
GCI3 4.4 ≥ GCI South Africa, India, Columbia, Argentina 4
3.4.2 Results
Table 9 displays resultsfromfixedeffectsregressions7
showingthatanincrease in r>g increases the top
1% share of income forGCI0 countriesand thatthe effectof anincrease inr>g is nearlyzero or negative
for mostnon-GCI0countries.Alsothe table showsthat the innovationmeasure patentsgranted hasno
effectontopincome inequalitybutR&Dexpenditure does.ExcludingpatentsgrantedandtreatingR&D
as a measure of innovation,(5) and(6) show that the coefficienton R&D expenditure percapitais
significantandpositive forthe top1% and top10% share of income forGCI0 countriesbutfor the top
10% share of income is nearlyzeroornegative forall othercountries.
These resultsshowthat aftercontrollingforfixedeffects andinnovation the Pikettyvariable r>g isnot
significantinexplainingtopincome inequalityforthe most competitive economies.Thisisevidence
againstPiketty’stheoryoncapital driventopincome inequality. Furthermore the resultsgive some
justificationforSchumpeteriangrowththeory andinnovationrents.
7 Pooled OLS regressions areincluded in Appendix Table A6.
25
Table 9 - Check appendix Table A1 for complete list of excluded countries.
Global Competitive Index Interaction Results – GCI0 is base
Piketty: r>g, K/Y
1980-2010
Patents Granted
1999-2010
R&D: Expenditure
1996-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
r - g 0.0110*
(1.88)
0.00121
(0.46)
0.00141
(0.23)
0.00425
(0.93)
0.000675
(0.20)
0.000542
(0.19)
GCI1 x r - g -0.0131*
(-1.67)
-0.00312
(-0.82)
-0.00830
(-0.97)
-0.00704
(-1.35)
-0.00926*
(-1.65)
-0.00542
(-1.47)
GCI2 x r - g -0.0146**
(-2.38)
-0.00193
(-0.51)
-0.00354
(-0.19)
-0.00251
(-0.25)
-0.00247
(-0.49)
-0.00217
(-0.64)
GCI3 x r - g -0.0104
(-1.49)
0.00176
(0.57)
-0.00243
(-0.36)
-0.00352
(-0.63)
-0.00221
(-0.70)
-0.00384
(-1.29)
Patents
Granted (Log)
0.0951
(0.67)
0.111
(1.28)
GCI1 x Patents
Granted (Log)
-0.0752
(-0.48)
-0.104
(-1.17)
GCI2 x Patents
Granted (Log)
-0.121
(-0.70)
-0.114
(-1.18)
GCI3 x Patents
Granted (Log)
-0.101
(-0.69)
-0.0783
(-0.90)
R&D per
capita
expenditure
(Log)
0.168*
(1.67)
0.0993
(1.01)
0.571***
(2.72)
0.357***
(2.69)
GCI1 x R&D
per capita
expenditure
(Log)
-0.363
(-1.59)
-0.326**
(-2.42)
GCI2 x R&D
per capita
expenditure
-0.519*
(-1.94)
-0.377**
(-2.21)
26
(Log)
GCI3 x R&D
per capita
expenditure
(Log)
-0.334
(-1.41)
-0.407***
(-3.82)
K/Y (Log) -0.717
(-1.39)
0.216
(0.77)
0.0334
(0.07)
0.106
(0.47)
0.344
(0.96)
0.353
(1.64)
GCI1 x K/Y
(Log)
0.373
(0.60)
-0.107
(-0.20)
GCI2 x K/Y
(Log)
0.467
(0.57)
-0.656
(-1.03)
GCI3 x K/Y
(Log)
-0.232
(-0.23)
0.0525
(0.16)
R&D
researchers
(Log)
-0.0977
(-1.12)
-0.0333
(-0.53)
-0.121
(-0.99)
-0.0380
(-0.55)
Index of
human capital
per person
(Log)
-0.756
(-1.14)
-0.0911
(-0.17)
2.286
(1.47)
0.942
(0.95)
0.835
(0.82)
1.048*
(1.74)
Share of
Government
Consumption
-1.722**
(-1.97)
-0.287
(-0.61)
-0.444
(-0.69)
-0.561
(-1.50)
-1.108
(-1.64)
-0.327
(-0.79)
Population
growth (annual
%)
-0.0411*
(-1.66)
-0.0112
(-0.64)
-0.00980
(-0.26)
-0.00161
(-0.11)
-0.0399
(-1.34)
-0.0222*
(-1.69)
Services per
capita (% of
GDP)
-0.00545
(-0.13)
-0.00182
(-0.03)
0.0972*
(1.89)
0.0401
(1.04)
0.0673
(1.09)
0.0431
(1.01)
Employment to
population
ratio, 15+, total
(%) (national
estimate)
-0.00515
(-0.79)
-0.00183
(-0.72)
-0.00368
(-0.56)
-0.00576
(-1.40)
0.00617
(1.09)
0.000102
(0.02)
Country Fixed Yes Yes Yes Yes Yes Yes
27
Effects
Observations 502 444 209 197 278 255
R2 overall 0.0403 0.167 0.0433 0.000163 0.00318 0.00781
R2 between 0.000164 0.0887 0.0153 0.00104 0.0127 0.00819
R2 within 0.662 0.558 0.382 0.351 0.540 0.543
# of countries
in regression
26 23 24 23 25 23
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
4 Discussion
4.1 Reverse Causality –Innovation on Income Inequality OLS Regression and
Panel VAR
It isworthwhile toinvestigatethe regressionof innovationonincome inequality inordertotruly
investigateSchumpeteriangrowththeory.Accordingtotheory,highertopincome inequalityshouldlead
to greaterinnovative activitysince inventors receive largerinnovationrentsandthus have a greater
incentive toinnovate. If this isthe case then the coefficientonpatentsgrantedin ourregressionsontop
income inequalitymeasures inSection3.3.1can likelysufferfromendogeneitycaused byreverse
causalitysince patentsgrantedisameasure of innovative activity. The regressionsonbooktomarketin
Section3.3.2 avoid some of this problembecause booktomarketis a measure of the value of innovative
activitiesratherthaninnovative activity.
Table 10 displaysthe resultswhenpatentsgrantedisregressedonincome inequalitymeasures. An
increase inthe top10% share of income significantlyincreasespatentsgrantedasshownin (3) of Table
10. Thisresultsuggeststhathighertop 10% share of income increasesinnovativeactivity.However,the
mostsignificantresultfromthe table isthatR&D expenditure percapitahasa significantlystrong
positive effectonpatentsgranted whichremainsrobustafteraccountingfor countryfixedeffects.
28
Table 10 - Check appendix Table A1 for complete list of excluded countries.
Reverse Causality – Patents Granted on Top Income Inequality
1999-2010
(1) (2) (3) (4) (5) (6)
Patents
Granted
(Log)
Patents
Granted
(Log)
Patents
Granted
(Log)
Patents
Granted
(Log)
Patents
Granted
(Log)
Patents
Granted
(Log)
Top 1% Share
of Income
(Log)
0.950
(1.55)
0.157
(0.69)
Top 10% Share
of Income
(Log)
2.817**
(2.43)
0.492
(0.96)
Pareto Lorenz
Coefficient
(log)
1.450
(0.94)
-0.137
(-0.55)
R&D per capita
expenditure
(Log)
2.138***
(3.00)
0.447
(1.33)
2.279***
(3.61)
0.779***
(2.77)
2.184***
(2.76)
0.562*
(1.68)
R&D
researchers
(Log)
-0.567
(-0.76)
0.443*
(1.73)
-1.058*
(-1.74)
0.334
(1.62)
-0.549
(-0.60)
0.413
(1.51)
r - g 0.00561
(0.32)
-0.00712*
(-1.71)
0.00865
(0.40)
-0.00606
(-0.94)
0.0266
(1.30)
-0.00836*
(-1.75)
K/Y (Log) 0.00361
(0.00)
0.986
(1.04)
0.471
(0.56)
0.522
(0.74)
-0.289
(-0.26)
1.325
(1.32)
Index of human
capital per
person (Log)
1.281
(0.39)
1.307
(0.43)
1.718
(0.62)
2.140
(0.56)
0.393
(0.11)
2.737
(0.80)
Share of
Government
Consumption
1.413
(0.30)
2.711
(1.06)
2.328
(0.54)
1.768
(0.69)
-5.480
(-0.97)
2.885
(0.86)
Population
growth (annual
%)
-0.270
(-0.96)
0.0272
(0.36)
-0.338*
(-1.83)
-0.00215
(-0.03)
-0.169
(-0.58)
0.0380
(0.46)
Services per -0.186*** -0.0567 -0.154*** -0.0490 -0.196*** -0.0676
29
capita (% of
GDP)
(-3.89) (-0.51) (-3.24) (-0.57) (-3.35) (-0.59)
Employment to
population
ratio, 15+, total
(%) (national
estimate)
-0.00384
(-0.13)
-0.0236
(-1.04)
0.0242
(0.70)
-0.0132
(-0.60)
-0.00461
(-0.12)
-0.0325
(-1.37)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 209 209 197 197 192 192
R2 0.925 0.930 0.923
R2 overall 0.793 0.707 0.806
R2 between 0.744 0.685 0.765
R2 within 0.642 0.764 0.651
# of countries
in regression
24 24 23 23 23 23
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
To furtherinvestigatethe causal relationshipbetween topincome inequality andinnovation the paper
conductstwo panel VARestimation;the firstbetweenthe top1% share of income and b/m, andthe
secondbetweenthe top10%share of income and b/m.The numberof lagschosenisfour,basedon
whatwas done inAtemsand Jones(2015). The resultsfromgrangercausalitytests showninFigure A1in
the appendix support whatwas foundinthe precedingregressions;increasesinthe top10% share of
income causesincreasesininnovation. Figure 5and Figure 6 show selected impulseresponse functions
fromboth multivariate regressions.
30
Figure 5 – Impulse Response Function: Impulse (Book to Market), Response (Top 1% Share of Income)
Figure 6 - Impulse Response Function: Impulse (Top 10% Share of Income), Response (Book to Market)
-2
-1.5
-1
-.5
0
0 5 10
bm : top1inc
IRF
step
impulse : response
-.005
0
0 5 10
top10inc : bm
IRF
step
impulse : response
31
4.2 Pre and Post Y2K Regressions – Evidence of a New Economy
So far the analysishasconsistentlygivendifferentresultsbetweenregressionswithtime spansstarting
fromand after1997 to 2010 and regressionswherethe time periodcoveredextendfromthe 1980s to
2010. Therefore thissectionfurtherexploresthe difference inresults.Table 11displaysresultsfor
regressionoverthe full period, pre-2000 andpost-2000.
The year 2000 can be consideredaturningpointineconomichistory.The endof the dot-combubble
had begunandit putout many incumbentsinthe relativelyyounginformationtechnology(IT) industry.
Between1990 and 2000 there washeavyinvestmentintothe ITindustrywhichledtoan IT stock bubble
whichiscommonlyreferred toasthe dot-combubble.8
The endof the dot-combubble resultedinless
incumbentfirmsinthe ITindustryandexcessITinfrastructure,whichgave rise tolowerentrybarriers
and thusincreasingcompetitionandinnovativeactivitiesinthe ITindustry. The post2000 IT industry
differsfromotherindustriesbecauseitrequireslittle capital,littlelabour,andhighabilityworkers(high
levelsof humancapital) forfirmstobe successful.The dot-combubbleenableddigitalizationwhichhas
transformed the advancedeconomiesintoadigital economy(Stiroh,2004). HémousandOlsen(2014)
suggestthatthe digital economycreateslargerhumancapital andinnovationrents resultinginhigher
income inequality.
Before 2000 innovationplaysnorole inexplainingincreases intopincome inequality andhumancapital
issignificantandnegativelycorrelatedwiththe top1% share of income asshowninTable 11. However
after2000 Table 11 showsthat innovationincreasestop1% and top10% share of income forGCI0
countries while humancapital becomesinsignificantwithapositivecoefficient.Table 11also showsthat
the coefficienton r-ghas become significantandpositiveforGCI0countriespost2000 whichgivessome
indicationforPiketty’shypothesis thatthe future of capitalismtendsto capital accumulationdriven
income inequality.The increasingsignificanceof innovationandhumancapital inTable 11 provides
evidence thatincreasesininnovationandhumancapital rents explainsrecenttrendsintopincome
inequality.
8 See “The Boom and Bust in Information Technology Investment” by Mark Doms. http://www.frbsf.org/economic-
research/files/er19-34bk.pdf.
32
Table 11 - Check appendix Table A1 for complete list of excluded countries.
Y2K Regressions: Book to Market pre and post 2000 – GCI0 is base
1980-2010 1980-2000 2000-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Book to
Market (Log)
-0.0812
(-1.10)
-0.00325
(-0.14)
-0.0534
(-0.66)
0.00927
(0.26)
-0.116**
(-2.03)
-0.0465*
(-1.70)
GCI1 x Book
to Market
(Log)
0.0228
(0.29)
-0.0242
(-1.12)
0.0208
(0.22)
-0.0142
(-0.44)
0.0380
(0.51)
-0.00547
(-0.20)
GCI2 x Book
to Market
(Log)
0.100
(0.67)
-0.000403
(-0.00)
0.0941
(0.35)
0.0140
(0.14)
0.00673
(0.02)
-0.0175
(-0.15)
GCI3 x Book
to Market
(Log)
-0.0920
(-1.10)
0.0488
(1.34)
0.131
(0.94)
-0.0409
(-0.72)
0.0340
(0.31)
0.0986***
(2.58)
r - g 0.0102**
(2.38)
0.00288
(1.21)
0.00590
(0.97)
-0.00126
(-0.57)
0.00517
(1.02)
0.00670*
(1.74)
GCI1 x r - g -0.0144**
(-2.07)
-0.00281
(-0.79)
-0.0119
(-1.56)
-0.000886
(-0.28)
-0.0129
(-1.43)
-0.00878*
(-1.94)
GCI2 x r - g -0.0162
(-0.76)
-0.00102
(-0.06)
-0.0206
(-0.55)
-0.000654
(-0.04)
-0.00319
(-0.06)
-0.000171
(-0.01)
GCI3 x r - g -0.0110**
(-2.31)
-0.0000623
(-0.02)
-0.00789
(-0.99)
-0.000938
(-0.30)
-0.00707
(-1.46)
-0.00882***
(-2.86)
Index of
human capital
per person
(Log)
-1.210**
(-1.99)
-0.349
(-0.70)
-1.723**
(-2.42)
-0.572
(-1.51)
1.391
(0.91)
0.544
(0.73)
Share of
Government
Consumption
-1.344
(-1.59)
0.0252
(0.05)
-0.568
(-0.45)
0.338
(0.47)
-0.351
(-0.64)
-0.495
(-1.47)
Population
growth (annual
-0.0136
(-0.59)
-0.0115
(-0.59)
0.0154
(0.13)
-0.00795
(-0.22)
-0.00565
(-0.22)
-0.00421
(-0.53)
33
%)
Services per
capita (% of
GDP)
-0.0132
(-0.30)
-0.00716
(-0.12)
-0.00184
(-0.03)
0.00168
(0.04)
0.121**
(2.21)
0.0414*
(1.83)
Employment
to population
ratio, 15+,
total (%)
(national
estimate)
-0.00741
(-1.16)
-0.00374*
(-1.70)
-0.0106
(-1.27)
-0.00385
(-1.37)
0.000607
(0.12)
-0.00283
(-0.89)
Country Fixed
Effects
Yes Yes Yes Yes Yes Yes
Observations 433 408 271 256 183 172
R2 overall 0.606 0.0668 0.502 0.264 0.0470 0.171
R2 between 0.572 0.0191 0.263 0.212 0.0418 0.103
R2 within 0.692 0.567 0.589 0.479 0.435 0.391
# of countries
in regression
25 23 22 21 24 22
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
5 Conclusion
Thispaperhas analyzedthe effectof CSNI andinnovationon topincome inequality.The resultsshow
that neitherCSNI norinnovationaffects the restof the income distributionasmeasuredbythe Pareto
Lorenzcoefficient.Itisfoundthatinnovation issignificantinexplainingtopincome inequality where an
increase ininnovationincreasesthe top1% andtop 10% share of income aspredictedinSchumpeter’s
growthmodel;howeverthese resultsdonotholdforthe leastcompetitive economies. Furthermore
regressions predictthatincreasesinthe top10% share of income leadtoincreases ininnovativeactivity
whichfurthervalidatesSchumpeteriangrowththeory. Insupportof Pikettythe paperfinds evidence
that as an economybecomesmore competitive r-gbecomesmore significantinexplainingtopincome
inequality trends.Lastly,resultsindicate that humancapital rents mightplay asignificantrole in
explainingtopincome inequalitytrends inthe future.
The limitations realizedinthispaperare a few.The paperwas limited toarelativelysmall numberof
observationsforthe analysisbutthisissue wasmitigatedbyusingpairs-clusterbootstrapprocedure
whichhas betterfinite sample propertiesthanclusteredstandarderrors.Alsointermsof observations,
the paperwas limitedto27 countrieswithmostof the observationscomingfrom OECDcountriesin
Europe and NorthAmericaandthus presentingsignificantselectionbias,yetthe analysisislargelybased
on topincome inequalityof advancedeconomies andtherefore beingrestrictedtothese countriesis
34
acceptable.Lastlythe time spanpresented significantchallengeswhichthe paperattemptedto
overcome byintroducingbooktomarketas a measure of innovation.
References
Acemoglu, J., Robinson, J. A. (2015).The Rise and Decline of General Laws of Capitalism Journal of
Economic Perspectives, 29(1), 3-28. Retrieved from http://economics.mit.edu/files/11348
Aghion, P., Akcigit, U., Bergeaud, A., Blundell, R. W., & Hemous, D. (2015).Innovation and top income
inequality C.E.P.R. Discussion Papers, CEPR Discussion Papers: 10659. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/1700663907?accountid=11233
Aghion, P., Akcigit, U., & Howitt, P. (2013).What Do We Learn From Schumpeterian Growth
Theory? National Bureau of Economic Research, Inc, NBER Working Papers: 18824. Retrieved
from http://www.nber.org/papers/w18824
Atems, B., & Jones, J. (2015). Income inequality and economic growth: A panel VAR
approach. Empirical Economics, 48(4), 1541-1561. doi:http://dx.doi.org/10.1007/s00181-014-
0841-7
Falkinger, J., & Zweimuller, J. (1997). The impact of income inequality on product diversity and
economic growth. Metroeconomica, 48(3), 211-237. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/56703414?accountid=11233
Hémous, D., & Olsen, M. (2014). The rise of the machines: Automation, horizontal innovation and
income inequality C.E.P.R. Discussion Papers, CEPR Discussion Papers: 10244. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/1660008993?accountid=11233
Hsing, Y. (2005). Economic growth and income inequality: The case of the US.International Journal of
Social Economics, 32(7), 639-647. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/56597962?accountid=11233
Lee, N., & Rodriguez-Pose, A. (2013). Innovation and spatial inequality in europe and USA. Journal of
Economic Geography, 13(1), 1-22. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/1314321790?accountid=11233
35
Raff, Daniel M G (Review of: Scherer,Frederic M., & Perlman, M., eds). (1993). Review of:
Entrepreneurship, technological innovation, and economic growth: Studies in the schumpeterian
tradition. Journal of Economic Literature, 31(4), 2017-2018. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/56619242?accountid=11233
Stiroh,Kevin (Review of: Greenan, Nathalie, L'Horty, Y., & Mairesse, J., eds). (2004). Review of:
Productivity, inequality, and the digital economy: A transatlantic perspective. Journal of Economic
Literature, 42(1), 218-219. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/56460277?accountid=11233
Tselios, V. (2011). Is inequality good for innovation? International Regional Science Review, 34(1),
75-101. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/855184089?accountid=11233
Voitchovsky, S. (2005). Does the profile of income inequality matter for economic growth?:
Distinguishing between the effects of inequality in different parts of the income
distribution. Journal of Economic Growth,10(3), 273-296. doi:http://dx.doi.org/10.1007/s10887-
005-3535-3
Zweimuller, J. (2000). Schumpeterian entrepreneurs meet engel's law: The impact of inequality on
innovation-driven growth. Journal of Economic Growth, 5(2), 185-206. Retrieved from
http://sfx.scholarsportal.info/guelph/docview/57019108?accountid=11233
36
Appendix
Table A1
Tables Excluded Countries
Table 1 None
Table 2 None
Table 3 None
Table 4 (1)-(2) India
(3)-(4) Argentina, Columbia, India, Indonesia
(5)-(6) Argentina, Finland, India
Table 5 (1)-(2) India, Argentina
(3)-(4) Argentina, Columbia, India, Indonesia
(5)-(6) Argentina, Finland, India
Table 6 (1)-(2) India, Argentina, Indonesia
(3)-(4) Argentina, Columbia, India, Indonesia
(5)-(6) Argentina, Finland, India, Indonesia
Table 7 (1)-(2) India, Indonesia
(3)-(4) Argentina, Columbia, India, Indonesia
(5)-(6) Argentina, Finland, India, Indonesia
Table 8 (1)-(2) India, Argentina, Indonesia
(3)-(4) Argentina, Columbia, India, Indonesia
(5)-(6) Argentina, Finland, India, Indonesia
Table 9 (1) India
(2) Argentina, Columbia, India, Indonesia
(3) India, Argentina, Indonesia
(4) Argentina, Columbia, India, Indonesia
(5) India, Indonesia
(6) Argentina, Columbia, India, Indonesia
Table 10 (1)-(2) India, Argentina, Indonesia
(3)-(4) Argentina, Columbia, India, Indonesia
(5)-(6) Argentina, Finland, India, Indonesia
Table 11 (1) India, Indonesia
(2) Argentina, Columbia, India, Indonesia
(3) India, Indonesia, Canada, Columbia, China
(4) India, Argentina, Indonesia, Canada, Columbia,
(5) India, Indonesia, Portugal
(6) Argentina, Columbia, India, Indonesia, Portugal
37
Table A2
Innovation Measure: Patents Granted
1999-2014
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
Patents
Granted (Log)
-0.00792
(-0.25)
0.0428
(0.71)
0.0317
(1.58)
0.0354
(0.88)
-0.00540
(-0.40)
-0.0320
(-1.02)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 303 303 284 284 276 276
R2 0.0320 0.113 0.0600
R2 overall 0.000542 0.110 0.0281
R2 between 0.000126 0.0913 0.000743
R2 within 0.150 0.130 0.233
# of countries
in regression
26 26 23 23 25 25
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
Table A3
Innovation Measure: Patents Granted
1999-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
Patents
Granted (Log)
-0.0120
(-0.29)
0.0495
(0.84)
0.0456*
(1.95)
0.0507
(1.30)
-0.00219
(-0.12)
-0.0266
(-0.75)
r - g 0.0199*
(1.89)
-0.00269
(-0.98)
0.00872
(1.13)
0.000323
(0.14)
-0.00735
(-1.10)
0.00171*
(1.68)
K/Y (Log) -0.140
(-0.59)
-0.161
(-0.45)
-0.243
(-1.11)
-0.0480
(-0.26)
0.123
(1.00)
0.0507
(0.22)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 229 229 210 210 211 211
R2 0.131 0.248 0.103
38
R2 overall 0.000312 0.156 0.00727
R2 between 0.0000748 0.103 0.00183
R2 within 0.164 0.152 0.223
# of countries
in regression
26 26 23 23 25 25
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
Table A4
Innovation Measure: Book to Market
1980-2014
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
Book to
Market (Log)
-0.0395
(-0.46)
-0.0139
(-0.51)
-0.101*
(-1.90)
-0.00175
(-0.14)
0.0150
(0.42)
0.00322
(0.16)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 685 685 640 640 621 621
R2 0.225 0.256 0.302
R2 overall 0.216 0.0911 0.295
R2 between 0.128 0.00617 0.0537
R2 within 0.225 0.525 0.426 0.535
# of countries
in regression
25 25 23 23 23 23
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
39
Table A5
Innovation Measure: Book to Market
1980-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Pareto
Lorenz
Coefficient
(log)
Pareto
Lorenz
Coefficient
(log)
Book to
Market (Log)
-0.121**
(-2.18)
-0.0251
(-0.81)
-0.107*
(-1.91)
-0.00251
(-0.18)
0.0143
(0.43)
0.00401
(0.24)
r - g 0.00489
(0.77)
-0.00121
(-0.32)
0.00334
(0.68)
0.000753
(0.38)
0.00150
(0.34)
0.00239
(0.97)
Country Fixed
Effects
No Yes No Yes No Yes
Observations 595 595 570 570 550 550
R2 0.351 0.268 0.324
R2 overall 0.294 0.0835 0.319
R2 between 0.164 0.000296 0.0331
R2 within 0.591 0.511 0.574
# of countries
in regression
25 25 23 23 23 23
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
Table A6
Global Competitive Index Interaction Results – GCI0 is base
Piketty: r>g, K/Y
1980-2010
Patents Granted
1999-2010
R&D: Expenditure
1996-2010
(1) (2) (3) (4) (5) (6)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
Top 1%
Share of
Income
(Log)
Top 10%
Share of
Income
(Log)
r - g 0.00770
(0.95)
-0.00149
(-0.30)
0.0179*
(1.85)
0.00484
(0.73)
0.0184
(1.39)
0.00694
(0.91)
GCI1 x r - g -0.00442
(-0.42)
0.00360
(0.57)
-0.0187
(-1.46)
-0.00863
(-0.96)
-0.0230
(-1.63)
-0.00849
(-1.08)
GCI2 x r - g -0.000829
(-0.08)
0.000689
(0.09)
-0.00864
(-0.26)
-0.00378
(-0.30)
-0.0155
(-0.95)
-0.00931
(-0.86)
40
GCI3 x r - g -0.00816
(-0.94)
0.00431
(0.64)
-0.0174
(-1.26)
-0.0155*
(-1.82)
-0.0140
(-0.95)
-0.00329
(-0.33)
Patents
Granted (Log)
0.101
(0.86)
0.115**
(2.35)
GCI1 x Patents
Granted (Log)
0.0126
(0.12)
0.0545
(1.02)
GCI2 x Patents
Granted (Log)
-0.00923
(-0.05)
-0.0609
(-0.81)
GCI3 x Patents
Granted (Log)
-0.142
(-0.89)
-0.229**
(-2.52)
R&D per
capita
expenditure
(Log)
-0.234
(-0.85)
-0.271**
(-1.98)
-0.0521
(-0.30)
-0.103
(-0.78)
GCI1 x R&D
per capita
expenditure
(Log)
0.00147
(0.08)
-0.00572
(-0.60)
GCI2 x R&D
per capita
expenditure
(Log)
-0.0257
(-0.68)
-0.00836
(-0.39)
GCI3 x R&D
per capita
expenditure
(Log)
0.0259
(0.50)
0.0412
(1.19)
K/Y (Log) -0.0854
(-0.46)
-0.00443
(-0.03)
-0.121
(-0.38)
0.0256
(0.13)
-0.0316
(-0.13)
-0.117
(-0.58)
GCI1 x K/Y
(Log)
-0.0255
(-0.19)
-0.0882
(-1.07)
GCI2 x K/Y
(Log)
-0.275*
(-1.73)
-0.181**
(-2.17)
GCI3 x K/Y 0.186 0.328
41
(Log) (0.92) (1.14)
R&D
researchers
(Log)
0.188
(0.65)
0.155
(1.31)
0.0935
(0.48)
0.227*
(1.79)
Index of
human capital
per person
(Log)
-0.913**
(-2.16)
-0.304
(-0.79)
-0.601
(-0.47)
0.00836
(0.02)
-0.857
(-0.77)
-0.278
(-0.44)
Share of
Government
Consumption
-4.068***
(-3.06)
-1.684**
(-2.07)
-3.354*
(-1.93)
-1.339
(-1.63)
-4.414**
(-2.32)
-1.792
(-1.62)
Population
growth (annual
%)
0.0276
(0.47)
-0.0221
(-0.46)
0.137*
(1.87)
0.0605
(1.62)
0.0564
(0.91)
0.0311
(0.65)
Services per
capita (% of
GDP)
-0.00345
(-0.33)
-0.00414
(-0.61)
0.0110
(0.32)
0.0195
(1.22)
-0.00774
(-0.53)
-0.0131
(-1.37)
Employment to
population
ratio, 15+, total
(%) (national
estimate)
-0.00971*
(-1.74)
-0.00850**
(-2.26)
-0.00523
(-0.57)
-0.00649
(-1.32)
-0.00746
(-0.81)
-0.00715
(-1.24)
Country Fixed
Effects
No No No No No No
Observations 502 444 209 197 278 255
R2 0.783 0.717 0.695 0.839 0.686 0.726
# of countries
in regression
26 23 24 23 25 23
t statistics in parentheses
Cluster Bootstrapped Standard Errors
* p < 0.1, ** p < 0.05, *** p < 0.01
42
Figure A1 – Granger Causality Tables
ALL 14.438 4 0.006
bm 14.438 4 0.006
top1inc
ALL 6.120 4 0.190
top1inc 6.120 4 0.190
bm
Equation  Excluded chi2 df Prob > chi2
ALL 6.517 4 0.164
bm 6.517 4 0.164
top10inc
ALL 11.512 4 0.021
top10inc 11.512 4 0.021
bm
Equation  Excluded chi2 df Prob > chi2

Más contenido relacionado

La actualidad más candente

2011 - Charting international labor comparisons
2011 - Charting international labor comparisons2011 - Charting international labor comparisons
2011 - Charting international labor comparisons
Richard Han
 
Trade Shocks, Taxes, and Inequality
Trade Shocks, Taxes, and InequalityTrade Shocks, Taxes, and Inequality
Trade Shocks, Taxes, and Inequality
Stockholm Institute of Transition Economics
 
Microeconomics Essay Draft
Microeconomics Essay Draft Microeconomics Essay Draft
Microeconomics Essay Draft
Declan Boyle
 
4Q_2015_RetailReport_email
4Q_2015_RetailReport_email4Q_2015_RetailReport_email
4Q_2015_RetailReport_email
Patrick Anthon
 
2016-us-goodwill-impairment-study
2016-us-goodwill-impairment-study2016-us-goodwill-impairment-study
2016-us-goodwill-impairment-study
Dave Wright
 
Lecture 7 - Endogenous growth theory
Lecture 7 - Endogenous growth theoryLecture 7 - Endogenous growth theory
Lecture 7 - Endogenous growth theory
UNU.MERIT
 
Economic Base Assessment presentation
Economic Base Assessment presentationEconomic Base Assessment presentation
Economic Base Assessment presentation
MVRPC
 

La actualidad más candente (20)

C&W-Marketbeat-U.S. Industrial-Q3-2018 #CRE #REALESTATE
C&W-Marketbeat-U.S. Industrial-Q3-2018 #CRE #REALESTATEC&W-Marketbeat-U.S. Industrial-Q3-2018 #CRE #REALESTATE
C&W-Marketbeat-U.S. Industrial-Q3-2018 #CRE #REALESTATE
 
2011 - Charting international labor comparisons
2011 - Charting international labor comparisons2011 - Charting international labor comparisons
2011 - Charting international labor comparisons
 
Decon 02
Decon 02Decon 02
Decon 02
 
Misallocations go a long way. Firm-level evidence from Poland
Misallocations go a long way. Firm-level evidence from PolandMisallocations go a long way. Firm-level evidence from Poland
Misallocations go a long way. Firm-level evidence from Poland
 
Wpti dec 2017_lkr-no comments
Wpti dec 2017_lkr-no commentsWpti dec 2017_lkr-no comments
Wpti dec 2017_lkr-no comments
 
Trade Shocks, Taxes, and Inequality
Trade Shocks, Taxes, and InequalityTrade Shocks, Taxes, and Inequality
Trade Shocks, Taxes, and Inequality
 
A CONTRIBUTION TO THE EMPIRICS OF ECONOMIC GROWTH (N.GREGORY MANKIW, DAVID ...
A CONTRIBUTION TO THE  EMPIRICS OF ECONOMIC  GROWTH (N.GREGORY MANKIW, DAVID ...A CONTRIBUTION TO THE  EMPIRICS OF ECONOMIC  GROWTH (N.GREGORY MANKIW, DAVID ...
A CONTRIBUTION TO THE EMPIRICS OF ECONOMIC GROWTH (N.GREGORY MANKIW, DAVID ...
 
TVA p3 The GLOBAL SOCIO ENVIRO ECONOMIC System
TVA p3 The GLOBAL SOCIO ENVIRO ECONOMIC SystemTVA p3 The GLOBAL SOCIO ENVIRO ECONOMIC System
TVA p3 The GLOBAL SOCIO ENVIRO ECONOMIC System
 
Microeconomics Essay Draft
Microeconomics Essay Draft Microeconomics Essay Draft
Microeconomics Essay Draft
 
4Q_2015_RetailReport_email
4Q_2015_RetailReport_email4Q_2015_RetailReport_email
4Q_2015_RetailReport_email
 
An evaluation of_the_relevance_of_the_ak
An evaluation of_the_relevance_of_the_akAn evaluation of_the_relevance_of_the_ak
An evaluation of_the_relevance_of_the_ak
 
9789740328255
97897403282559789740328255
9789740328255
 
2016-us-goodwill-impairment-study
2016-us-goodwill-impairment-study2016-us-goodwill-impairment-study
2016-us-goodwill-impairment-study
 
Unempolyment in india
Unempolyment in indiaUnempolyment in india
Unempolyment in india
 
Lecture 7 - Endogenous growth theory
Lecture 7 - Endogenous growth theoryLecture 7 - Endogenous growth theory
Lecture 7 - Endogenous growth theory
 
Economic Base Assessment presentation
Economic Base Assessment presentationEconomic Base Assessment presentation
Economic Base Assessment presentation
 
Pertemuan 4 capm, single index model, apt
Pertemuan 4 capm, single index model, aptPertemuan 4 capm, single index model, apt
Pertemuan 4 capm, single index model, apt
 
C&W MARKETBEAT- U.S. Office Q4 2018
C&W MARKETBEAT- U.S. Office Q4 2018C&W MARKETBEAT- U.S. Office Q4 2018
C&W MARKETBEAT- U.S. Office Q4 2018
 
2016 09-29 gci summit v10
2016 09-29 gci summit v102016 09-29 gci summit v10
2016 09-29 gci summit v10
 
Greater Charlotte in the global economy: Benchmarking the region's global com...
Greater Charlotte in the global economy: Benchmarking the region's global com...Greater Charlotte in the global economy: Benchmarking the region's global com...
Greater Charlotte in the global economy: Benchmarking the region's global com...
 

Similar a kyel research paper draft final

Dissertation
DissertationDissertation
Dissertation
Interdax
 
Final Paper-Development Economics
Final Paper-Development EconomicsFinal Paper-Development Economics
Final Paper-Development Economics
Gang Wu
 
Comparative Advantage and Intellectual Property Rights: Some Evidences from C...
Comparative Advantage and Intellectual Property Rights: Some Evidences from C...Comparative Advantage and Intellectual Property Rights: Some Evidences from C...
Comparative Advantage and Intellectual Property Rights: Some Evidences from C...
Peter Zhen Ye
 
Estimating the stock of public capital in 170 countries May 2021.pdf
Estimating the stock of public capital in 170 countries May 2021.pdfEstimating the stock of public capital in 170 countries May 2021.pdf
Estimating the stock of public capital in 170 countries May 2021.pdf
AbdelmalekBOUMDIR1
 
test-bank-ch-chapter-questions chap 5.pdf
test-bank-ch-chapter-questions chap 5.pdftest-bank-ch-chapter-questions chap 5.pdf
test-bank-ch-chapter-questions chap 5.pdf
HiLinh29
 
Disparity in growth rates among countries
Disparity in growth rates among countriesDisparity in growth rates among countries
Disparity in growth rates among countries
Sparsh Banga
 
1 10The Real Economy in the Long RunECO372The.docx
1     10The Real Economy in the Long RunECO372The.docx1     10The Real Economy in the Long RunECO372The.docx
1 10The Real Economy in the Long RunECO372The.docx
honey725342
 

Similar a kyel research paper draft final (20)

Dissertation
DissertationDissertation
Dissertation
 
Final Paper-Development Economics
Final Paper-Development EconomicsFinal Paper-Development Economics
Final Paper-Development Economics
 
Fiscal space and the composition of public finances - Jean-Marc Fournier, OECD
Fiscal space and the composition of public finances - Jean-Marc Fournier, OECDFiscal space and the composition of public finances - Jean-Marc Fournier, OECD
Fiscal space and the composition of public finances - Jean-Marc Fournier, OECD
 
wp13135
wp13135wp13135
wp13135
 
Presentation
Presentation Presentation
Presentation
 
How Much Should We Trust China's GDP.pdf
How Much Should We Trust China's GDP.pdfHow Much Should We Trust China's GDP.pdf
How Much Should We Trust China's GDP.pdf
 
Comparative Advantage and Intellectual Property Rights: Some Evidences from C...
Comparative Advantage and Intellectual Property Rights: Some Evidences from C...Comparative Advantage and Intellectual Property Rights: Some Evidences from C...
Comparative Advantage and Intellectual Property Rights: Some Evidences from C...
 
劉遵義 Income inequality under economic globalisation(final) 20150414
劉遵義 Income inequality under economic globalisation(final) 20150414劉遵義 Income inequality under economic globalisation(final) 20150414
劉遵義 Income inequality under economic globalisation(final) 20150414
 
Jonathan D. Ostry - Fondo Monetario Internacional (FMI).
Jonathan D. Ostry - Fondo Monetario Internacional (FMI). Jonathan D. Ostry - Fondo Monetario Internacional (FMI).
Jonathan D. Ostry - Fondo Monetario Internacional (FMI).
 
How has Income Inequality and Wage Disparity Between Native and Foreign Sub-p...
How has Income Inequality and Wage Disparity Between Native and Foreign Sub-p...How has Income Inequality and Wage Disparity Between Native and Foreign Sub-p...
How has Income Inequality and Wage Disparity Between Native and Foreign Sub-p...
 
The drivers of income inequality
The drivers of income inequalityThe drivers of income inequality
The drivers of income inequality
 
Medef - La France peut-elle retrouver le chemin d'une croissance forte ?
Medef  - La France peut-elle retrouver le chemin d'une croissance forte ?Medef  - La France peut-elle retrouver le chemin d'une croissance forte ?
Medef - La France peut-elle retrouver le chemin d'une croissance forte ?
 
Chang Report Full.pdf
Chang Report Full.pdfChang Report Full.pdf
Chang Report Full.pdf
 
La diseguaglianza economica dal 1900 ad oggi
La diseguaglianza economica dal 1900 ad oggiLa diseguaglianza economica dal 1900 ad oggi
La diseguaglianza economica dal 1900 ad oggi
 
Session 8 c diewert discussion of feenstra inklaar and timmer
Session 8 c diewert discussion of feenstra inklaar and timmerSession 8 c diewert discussion of feenstra inklaar and timmer
Session 8 c diewert discussion of feenstra inklaar and timmer
 
Estimating the stock of public capital in 170 countries May 2021.pdf
Estimating the stock of public capital in 170 countries May 2021.pdfEstimating the stock of public capital in 170 countries May 2021.pdf
Estimating the stock of public capital in 170 countries May 2021.pdf
 
test-bank-ch-chapter-questions chap 5.pdf
test-bank-ch-chapter-questions chap 5.pdftest-bank-ch-chapter-questions chap 5.pdf
test-bank-ch-chapter-questions chap 5.pdf
 
Disparity in growth rates among countries
Disparity in growth rates among countriesDisparity in growth rates among countries
Disparity in growth rates among countries
 
1 10The Real Economy in the Long RunECO372The.docx
1     10The Real Economy in the Long RunECO372The.docx1     10The Real Economy in the Long RunECO372The.docx
1 10The Real Economy in the Long RunECO372The.docx
 
IMF Working Paper
IMF Working PaperIMF Working Paper
IMF Working Paper
 

kyel research paper draft final

  • 1. A Cross Country Analysis on Top Income Inequality Kyel Governor 6/9/2016 Thispaperusescross-countrypanel datafrom1950-2014 to investigateif the rise intopincome inequalitycanbe explainedbycapital accumulationwheninnovationandcross-countrydifferencesare takenintoaccount. PooledOLSandfixedeffectsmodelsare usedtoinvestigatethe relationship. This paperfindssome evidence infavourof capital accumulationdriving topincome inequalityforhighly competitivecountries.Howeverthere isstrongerevidence suggestingthatinnovationrentsdrivetop income inequalityratherthancapital accumulation.
  • 2. 1 Contents Introduction.......................................................................................................................................2 1 Theory........................................................................................................................................3 1.1 Piketty’s General Laws of Capitalism.....................................................................................3 1.2 Schumpeterian Growth Theory and Innovation Rents ............................................................4 2 Data and Measurement...............................................................................................................5 2.1 Income Inequality................................................................................................................6 2.2 Innovation...........................................................................................................................7 2.3 Piketty’s r > g and K over Output...........................................................................................9 2.4 Control Variables.................................................................................................................9 2.5 Global Competitiveness Index (GCI) ......................................................................................9 2.6 Table of Variables and Descriptions.......................................................................................9 3 Main Empirical Analysis:The effectof innovativenessandcapital accumulationonincome distribution ......................................................................................................................................11 3.1 Estimation Strategy............................................................................................................11 3.2 Pikettyr> g and Capital Accumulation:ResultsfromPooledOLSandFixedEffectsRegressions 11 3.3 SchumpeterianGrowthTheoryandInnovationRents:ResultsfromPooledOLSandFixed Effects Regressions .......................................................................................................................16 3.3.1 Measure of Innovation: Patents Granted .....................................................................16 3.3.2 Measure of Innovation: Book to Market.......................................................................20 3.4 Evidence of Piketty’sr>gand Schumpeter’sInnovationRents:Interactionof Countries Grouped by the Global Competitiveness Index ...............................................................................23 3.4.1 GCI Grouping Table.....................................................................................................24 3.4.2 Results.......................................................................................................................24 4 Discussion.................................................................................................................................27 4.1 Reverse Causality –Innovation on Income Inequality OLS Regression and Panel VAR..............27 4.2 Pre and Post Y2K Regressions – Evidence of a New Economy................................................31 5 Conclusion................................................................................................................................33 References.......................................................................................................................................34 Appendix..........................................................................................................................................36
  • 3. 2 Introduction Increasingtopincome inequality indevelopedcountries hasbecome aphenomenoninthe worldtoday and yetthere isno general agreement onthe rootcauses behind the rise in topincome inequality.This paperdebateswhethertopincome inequalityisdrivenby capital accumulationorinnovation.Thomas Piketty hasbroughtthe issue of growthand income redistribution backtothe forefrontof economic debateswithhis (2014) book, Capitalin the Twenty-FirstCentury,where he argues thatcapital accumulationbythe wealthiesttop1%is drivingtopincome inequality. Inshort, he defendsthat wealth growsfasterthan economicoutput,centeringhisargument onthe expressionr> g (where ris the rate of returnto capital andg isthe economicgrowthrate),andso slowergrowthleadsto a rapidincrease in income inequality. Piketty’smore thansix-hundredpage publication inlarge partignores cross-country differencesand the endogenousevolutionof technology anditsrelationshipwith topincome inequality leavinghis worksusceptibletodispute (Acemogluetal,2015). Infact in the UnitedStates patentingand top 1% income share followparallel evolutions indicatingthatinnovationmayplayarole inexplaining the rise in top income inequality (Aghionetal,2015). Thispaperwill use cross-countrypanel datafrom 1950-2014 to investigate if the rise in topincome inequality canbe explainedby capital accumulation wheninnovationandcross-countrydifferences are takenintoaccount.Twodifferentmeasuresare used for innovation;total patentsgranted andthe booktomarketratio. The capital share of national income (CSNI) ismeasuredby twovariables;the capital stocktogrossdomesticproduct ratioand r minusg. Lastly, top income inequalityismeasuredbythe share of income heldbythe top1% and the top 10%, and the ParetoLorenzcoefficient. In summary,thispaperfinds some evidence validatingPiketty’sargumentthatcapital accumulation drivestopincome inequality forhighly competitive1 countries.Howeverthere isstrongerevidence suggestingthatinnovationrentsdrivetopincome inequalityratherthancapital accumulation.Excluding the leastcompetitivecountries, measuresof innovation show asignificantrelationshipwith topincome inequality where higherlevelsof innovationleadto highertop 1% and top10% income shareswhich reflectinnovationrentsandthus isconsistentwithSchumpeteriangrowththeory. The modelpredicts that on average a 1% increase inpatentsgrantedleadstoa 16.6% increase in the top1% share of income anda 14.7% increase inthe top 10% share of income,aresultconsistentwithAghionetal. (2015) U.S. cross-state analysisoninnovationandtopincome inequality. The booktomarketratio measure of innovation alsoshowsthatonaverage innovationincreases the top1% and top 10% share of income.Aftercontrollingforfixedeffects,regressionsshow that innovationmeasuressuchas R&D expenditure andbooktomarketratio are significantinpredictingtopincome inequality. Furthermore, resultsshownosignificantrelationshipbetween innovation orCSNI andthe restof the income distribution. Wheninvestigatingwhetherincome inequality causesinnovation pooledOLSregressions showthat on average the top10% share of income increasesinnovation andapanel vectorauto regressionsupportsthisresult.Finally the paperfindsthatbefore the year2000 there is no evidence of a positive relationshipbetween CSNI andtopincome inequalityandnorelationshipbetweeninnovation 1 As measured by the Global Competitive Index (2015-2016).See Section 3.4.
  • 4. 3 and topincome inequality.However afterthe year2000 there isa positive relationship forbothCSNI and innovationwithtopincome inequality. ThisanalysishasbeenmotivatedbyAcemogluandRobinson(2015) where they criticize general lawsof capitalismproposedinPiketty’s(2014) recentwork. Contrary to Piketty,theyshow thatr> g has no relationship withtopincome inequality. Mypaperadds to theirworkontop income inequalityby addingan analysisof innovationdriveninequality.ThispapercloselyfollowsAghion,Akcigit, Bergeaud, Blundell and Hémous(2015) U.S. cross-state analysisof innovationontopincome inequality exceptthis papertakesa cross-country perspectiveandfactorsin the ParetoLorenz Coefficientasa measuresof income inequality.Like Aghionetal.(2015),the paper dependsonthe Schumpteriangrowthmodel presentedinAghion,AkcigitandHowitt (2013) in where innovationledgrowthis dependenton innovation rentswhichare representedbyincreasesintopincome inequality.Therefore this endogenous growthmodelleadsthe analysisontopincome inequalityandinnovation. Thisgrowth model also allowsthatanincrease in topincome inequality causesanincrease ininnovation as mentionedby Tselosis(2011) and Zweimuller(2000).So our analysisinvestigatesthislikelyevent firstby univariate regressions andthenby apanel vectorauto regression (VAR) oninnovationandtopincome inequality motivatedbyAtemsandJones(2015).Finally the paperallows thatthere hasbeen arecent shifttoa newdigital economy aspresentedin Stiroh’s(2004) review andsothe return on innovation and humancapital has dramaticallyincreased.Thishypothesisissupportedby HémousandOlsen (2014) inwhichtheyargue that the digital economybenefitshighskilledlabourers andinnovatorsbecause of the automationof lowskilledlabour.Thispaperlookstoinvestigate thishypothesis. The paper isorganizedasfollows. Section1of the paperreviewsPiketty’s general lawsof capitalismand Schumpteriangrowththeorywhere topincome inequalityisdrivenbyquality-improvinginnovations. Section2 introducesthe datainvolvedanddescribesthe measuresof innovationandincome inequality usedinthe paper.Section3 presentsthe mainfindings.Section4isa discussionontwoextensionsof the analysis:first,aninvestigationintoareverse causalitywhere topincome inequalitydrives innovation;andsecond, alookat pre Y2K andpost Y2K regressions.Section5concludes. 1 Theory 1.1 Piketty’s General Laws of Capitalism Pikettyproposesatheoryon the longrun proclivitiesof capitalismin Capitalin the Twenty-firstCentury (2014). He reliesonMarxianeconomics aspremise tocome to the conclusionthat the future of capitalismisone where there ismassownershipof capital bythe wealthiestandtheirmainincome comesfromcapital rents. To come to his predictions, Piketty firstpresents two“fundamental laws”; 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑠ℎ𝑎𝑟𝑒 𝑜𝑓 𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑖𝑛𝑐𝑜𝑚𝑒 = 𝑟 𝑥 ( 𝐾 𝑌 ), 𝐾 𝑌 = 𝑠/𝑔,
  • 5. 4 where 𝑟 isthe real interestrate, 𝐾 isthe capital stock, 𝑌 isGDP, 𝑠 is the savingrate and 𝑔 isthe growth rate of GDP. As derivedinaSolowmodel the growthrate of capital stock 𝐾 isequal to saving 𝑠𝑌 ina closedeconomyandso the ratio 𝐾 𝑌 overtime will approach 𝑠/𝑔due to economicgrowth. Pikettythen combinesthe twoequationstoget; 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑠ℎ𝑎𝑟𝑒 𝑜𝑓 𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑖𝑛𝑐𝑜𝑚𝑒 = 𝑟 𝑥 ( 𝑠 𝑔 ). Assuming 𝑟 and 𝑠 are approximatelyconstant derivesPiketty’sfirstgeneral law; 1. The capital share of national income ishigh whengrowth islow. His secondgeneral lawstates; 2. r > g, whichdefinesthatthe real interestrate r isalwaysgreaterthanthe growthrate g inan economy. This lawholdswhena countryisdynamicallyefficient,meaningthatithas maximizedconsumption inevery periodoveraninfinite timeline,otherwise itmaynotbe the case that the real interestrate exceedsthe growthrate. Piketty’s thirdgeneral law proposesthat; 3. Income inequalitytendsto rise wheneverr > g. The logicbehindthispropositionisthat capital income growsapproximatelyatthe rate r where national income growsat the rate g and since capital income is distributedmainlytothe wealthiest;their incomesgrowfasterthanthe per capita income thusresultingincapital-drivenincome inequality. Accordingto Piketty’stheorytopincome inequalityrisesasthe capital share of national income increases because capital incomeismainlydistributedamongthe wealthiest. Assumingris approximately constant,capital share of national incomewillrise asK/Yincreases(change in capital exceedschange in GDP) and/oras g decreases makingthe distance between randg greater. Thispaper testsbothhypotheses inordertosubstantiate Piketty’sgenerallawsof capitalism. 1.2 Schumpeterian Growth Theory and Innovation Rents The baseline Schumpeteriangrowthmodel asdevelopedinAghionetal.(2015) offersa theoretical frameworkforhowinnovationincreasestopincome inequality. The mainpredictionsof the model asit relatestothispaperare;  Innovation increasestopincome inequality.  Higherinnovationrentleadstoincreasedlevelsof innovation.2 2 This resultis further explored in Section 4.2 where the marginal effect of innovation on income inequality is stronger after the IT market crashed in 2000.
  • 6. 5 To come to these predictionsthe baseline Schumpeteriangrowthmodelisdevelopedasfollows. Given discrete time,the economyispopulatedbytwotypesof individualseveryperiod;capital ownerswho ownthe firmsand the productionworkers. Individualslive onlyforone period andinthe subsequent perioda newgenerationisbornwiththose whoare bornto firmowners inheritingtheirparents’firm and the restwork as productionworkersunlesstheysuccessfully innovate andtakeoveranincumbent. The followingCobb-Douglastechnologyproductionfunction definesthe productionof finalgoods 𝑌𝑡: 𝑙𝑛𝑌𝑡 = ∫ 𝑙𝑛 1 0 𝑦𝑖𝑡 𝑑𝑖 𝑦𝑖𝑡 = 𝑞𝑖𝑡 𝑙𝑖𝑡 where 𝑦𝑖𝑡 isa vector of intermediateinputsusedinfinal production, 𝑙𝑖𝑡 isthe labourusedtoproduce the intermediateinputs,and 𝑞𝑖𝑡 isthe labourproductivity; all attime t.Wheneverinnovationtakes place inany sector 𝑖 inperiod 𝑡 the innovatorgainsa technologicalleadof 𝜂 𝐻 overthe competition and so the innovatorproducesthe intermediateinput 𝑦𝑖𝑡 withthe productionfunction: 𝑦𝑖𝑡 = 𝜂 𝐻 𝑞𝑖,𝑡−1 𝑙𝑖𝑡 If no newinnovationsoccurinthe followingperiod,the innovator’stechnologyispartlyimitatedby competitorsandsothe innovator’stechnological leaddecreasesto 𝜂 𝐿 where 1 < 𝜂 𝐿 < 𝜂 𝐻. Solvingthe model yieldsthe twofollowingequationspertinenttothe paper’sanalysis; 𝑀𝐶𝑖𝑡 = 𝑤 𝑡/𝑞𝑖, 𝑡, and, 𝑝𝑖,𝑡 = 𝑤 𝑡 𝜂𝑖𝑡/𝑞𝑖, 𝑡 where 𝜂𝑖,𝑡 𝜖 {𝜂 𝐻, 𝜂 𝐿}, 𝑀𝐶𝑖𝑡 isthe marginal cost of producingthe intermediate input, 𝑤 𝑡 isthe wagespaid to labourinvolvedinproducingthe intermediate input 𝑖,and 𝑝𝑖,𝑡 isthe price charged forthe intermediate input;all attime t. Thusthe profitearnedbythe innovatorisgivenby; 𝜋𝑖𝑡 = ( 𝑝𝑖𝑡 − 𝑀𝐶𝑖𝑡) 𝑦𝑖𝑡 = ( 𝜂𝑖𝑡 − 1) 𝑌𝑡 𝜂𝑖𝑡 > 0 The profitsearnedbyinnovatorsinthismodel are innovationrents.Furthermore if the costof innovationisallowedinthismodel then the costof innovationforentrantscanbe greaterthan the cost of innovationforincumbents (Aghionetal.,2015). This leads toentrybarriersand thusreduces innovative activity since entrantsreceive lessprofitfrominnovationthe higherthe costof innovationis for them. 2 Data and Measurement The paper uses cross-country unbalanced panel datacoveringthe period1950-2014. Missingvalues betweendatapointsare approximatedbyalinearinterpolation.The 27countriescoveredare as follows:Argentina,Australia,Canada,China,Colombia,Denmark,Finland,France,Germany,India,
  • 7. 6 Indonesia,Ireland,Italy,Japan,Korea,Malaysia,Netherlands,New Zealand,Norway,Portugal, Singapore,SouthAfrica,Spain,Sweden,Switzerland,UnitedKingdom, and UnitedStates. 2.1 Income Inequality The data on the share of income heldbythe top 1%,top 10%, and on the ParetoLorenzCoefficient was obtained fromthe World Top Income Database (Alvaredo,Atkinson,Piketty,andSaez’sWorldTop IncomesDatabase at http://topincomes.parisschoolofeconomics.eu/).These are ourtopincome inequalitymeasures. Dataonthe share of income ownedbythe bottom40% was obtainedfromthe WorldBank Databank (http://databank.worldbank.org/data).The share of income ownedbythe middle 40%-90% wasobtainedbysubtractingthe top10% and bottom40% share from100. Figure1showsthe evolutionof top1%income inequality forFrance,Germany,SouthAfrica,Sweden, and UnitedStates. The graphclearlyillustratessignificantcrosscountrydifferences howevera significantupwardtrendisnoticeable forall countriesaroundafter1990. Figure 2 showsthe top 10%, middle 40%-90%,andbottom 40% income sharesovertime forthe UnitedStates and itis evidentthat top income sharesincrease atthe expenseof the middle class. Figure 1 5 101520 1950 1960 1970 1980 1990 2000 2010 2020 Year France Germany South Africa Sweden United States Top 1% Income Share, 1960-2014
  • 8. 7 Figure 2 2.2 Innovation We obtainpatentsgranted bythe USPTO and EPOto obtain total of patentsgranted percountryby inventor’scountryof residence. Thisdatawasobtainedfromthe OECDdatabase (http://stats.oecd.org/).The datacoversthe time period1999-2014. The bookto marketratio(b/m) was createdby dividingcapital stockbymarketcapitalization. Capital stock (K) wasobtainedfromPenn WorldTables version8.1 (www.ggdc.net/pwt) andmarketcapitalizationwasobtainedfrom the World Bank Databank. Thisproxyfor the bookto marketratio workswell.The correlationbetweenthe actual U.S. bookto market ratio withKovermarketcapitalizationisequal to0.9452. A low bookto market ratiovalue impliesthatavaluationabove the worthof physical assets.A marketvaluationabove physical assets impliesvaluableintangible assetssuchasinnovativeactivitiesandsomovementinthe bookto marketratio isa validmeasure of innovation. Figure 3 showsthe evolutionof patentsgrantedforFrance,Germany,SouthAfrica,Sweden,andUnited Stateswhile Figure 4illustrates the dynamicrelationshipbetween booktomarketand patentsgranted for the UnitedStates.Figure 4 gives evidence of arelationship between the twomeasuresof innovation especiallywhencomparing the three yearlagof patentsgrantedwith the bookto marketratio.An increase inpatentsgranted (lagged3years) leadstoa fall inthe book to marketratio. 1020304050 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Year Top 10% Share of Income Mid 40%-90% Share of Income Bottom 40% Share of Income United States Income Distribution 1950 - 2014
  • 9. 8 Figure 3 Figure 4 0 50 100150 2000 2005 2010 2015 Year France Germany South Africa Sweden United States Total Patents Granted, 1999-2014 .02.025.03.035.04 BooktoMarket 80 100120140160 2000 2005 2010 2015 Year Granted patents (in 100s) Granted patents (in 100s), Lagged 3 yearsBook to Market United States, Book to Market vs Patents Granted
  • 10. 9 2.3 Piketty’s r > g and K over Output Data on real interestrates (r) were retrievedfromthe WorldBankDatabankand data on GDP was obtainedfromPennWorldTablesversion8.1and Madison’sdataset (http://www.ggdc.net/maddison/maddison-project/data.htm).The GDPgrowthrate (g) iscomputed and the variable r> g is computedbysimplysubtractinggfromr. Our capital to outputratio (K/Y) isalso computedusingKand GDP data. 2.4 Control Variables The paper controlsforthe size of the financial sector andthe size of the governmentsectorbecause theyare likelyto have directeffectsoninnovationandtopincome share (Aghionetal,2015). Furthermore the analysiscontrolsfor macroeconomicvariablessuchaspopulationgrowth,the employmentrate, andhumancapital;andResearchand Development(R&D) variables suchasR&D expenditure percapitaandnumberof R&D researchers. Controllingforthe employmentrate ismainly to reduce time varyingcrosscountrydifferencesinordertohave more efficientresults. Humancapital is controlled forsince thislikelyhasaneffectonincome inequalitymeasuresandinnovation. FinallyR&D iscontrolledforsince itiscorrelatedwithpatentsgrantedandincome inequalitymeasures;alsoR&D may be consideredameasure of innovation. 2.5 Global Competitiveness Index (GCI) The paper groupscountriesbasedontheirGCI obtainedfromthe WorldEconomicForum Global CompetitivenessReport (http://reports.weforum.org/global-competitiveness-report-2015-2016/) in orderto see the relationshipof CSNIandinnovation ontopincome inequality acrossdifferenteconomic stages.The GCI is a composite of sub-indexes suchasinstitutions,infrastructure, macroeconomic environment,healthandprimaryeducation,highereducationand training, goodsmarketefficiency, labourmarket,financial marketdevelopment,technological readiness,marketsize,business sophistication,andinnovation. The table inSection3.4.1showshow countriesare grouped. 2.6 Table of Variables and Descriptions Variable Number of Observations (annual) Mean Standard Deviation Time Span Retrieval Description Income Inequality Variables Top 1% Share of Income (%) 1177 12.057 3.425 1950- 2014 World Top Incomes Database Share of national income held by the top 1% earners Top 10% Share of Income (%) 962 36.6 6.809 1950- 2014 World Top Incomes Database Share of national income held by the top 10% earners Pareto Lorenz Coefficient 1068 1.844 .224 1950- 2014 World Top Incomes Database Measure of the income distribution Piketty and Innovation Variables Patents Granted 416 8.872 22.643 1999- 2014 OECD Database Number of patents granted by the EPO
  • 11. 10 (in 1000s) and USPTO Market Capitalization (% of GDP) 799 93.502 45.774 1975- 2014 World Bank Databank Share price x number of sharesoutstanding for listed domestic companies r (%) 946 2.736 3.554 1961- 2014 World Bank Databank Real interest rate g (%) 1620 1.224 2.103 1951- 2010 Madison’s dataset GDP growth rate K (in mil. 2005US$) 1644 8724524 1.27e+07 1950- 2011 Penn World Tables Capital Stock GDP (in mil. 2005US$) 1644 2856685 4178096 1950- 2011 Penn World Tables Real GDP Control Variables R&D per capita expenditure (% of GDP) 346 42289.59 20903.95 1996- 2010 World Bank Databank Expenditures for research and development both public and private Number of R&D researchers 391 3345.405 1427.127 1996- 2013 World Bank Databank Professionals engaged in the creation of new knowledge,products, processes, methods, or systems Index of Human Capital per person 1644 3.133 .337 1950- 2011 Penn World Tables Based on years of schoolingandreturns to education Share of Government Consumption (%) 1644 .139 .034 1950- 2011 Penn World Tables Share of government consumption at current PPPs Services per capita (% of GDP) 1014 68.641 8.288 1960- 2014 World Bank Databank Services correspond to ISICdivisions50-99 (includes financial) Employment to population ratio, 15+, total (%) (national estimate) 817 58.957 8.937 1980- 2014 World Bank Databank Proportion of a country's population that is employed Population growth (annual %) 1484 .989 .467 1950- 2011 Penn World Tables Population growth rate Global N/A N/A N/A 2015- 2016 World Economic Integrates macroeconomic and
  • 12. 11 Competitivene ss Index (1-7 (best)) Forum micro/business aspects of competitiveness into a single index 3 Main Empirical Analysis:The effect of innovativeness and capital accumulationon income distribution 3.1 Estimation Strategy The main empirical analysis involveslookingatwhetherPiketty’svariables,r-g andcapital accumulation measuredbyK/Y,or innovationmeasuredby b/mandtotal patentsgranted are responsible for increasesintopincome shares. The mainestimatedequationis: log( 𝑦𝑖𝑡) = 𝐴 + 𝐵𝑖 + 𝐵𝑡 + 𝐵1𝑗( 𝑃𝑖𝑘𝑒𝑡𝑡𝑦𝑗) + 𝐵2 𝑙𝑜𝑔(𝑖𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛) + 𝐵3𝑖 𝐺𝐶𝐼 𝑘−1 ∗ 𝑙𝑜𝑔(𝑖𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛) + 𝐵4𝑗𝑖 𝐺𝐶𝐼 𝑘−1 ∗ ( 𝑃𝑖𝑘𝑒𝑡𝑡𝑦𝑗) + 𝐵5 𝑋𝑖𝑡 + 𝜀𝑖𝑡 where 𝑦𝑖𝑡 isthe measure of inequality, 𝐵𝑖 isacountry fixedeffect, 𝐵𝑡 isa yearfixedeffect, 𝑃𝑖𝑘𝑒𝑡𝑡𝑦𝑗 isa vectorof Pikettyvariableswhich include r-g(j=1) andK/Y(j=2), 𝑖𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛isinnovativeness measured by b/mor total patentsgranted, 𝐺𝐶𝐼 𝑘−1 are GCI dummies and 𝑋 isa vector of control variables. In Section 3.2 𝐵2, 𝐵3 and 𝐵4 are setto zero and a seriesof Pikettyregressionsare estimatedby pooled OLS and fixedeffects.The restrictionon 𝐵2 isremoved inSection3.3and the model isestimatedagain by pooledOLS andfixedeffects. InSection3.4all restrictionsare removedandthe model isestimated by fixedeffects. All resultsare clusterbootstrappedand include time dummies.3 3.2 Piketty r > g and Capital Accumulation: Results from Pooled OLS and Fixed Effects Regressions The resultsfroma seriesof Pikettyregressionsare presentedbelowinTable 1.In (3) and(4) of Table 1 it isassumedthat all capital marketsare openand cross countryvariationinr iszero andso r is normalizedto0, therefore we have r-g=- g. The regressionsshow nosignificantrelationshipbetween the top 1% share of income andr > g for pooledOLSand fixedeffectsestimates asseeninAcemogluand Robinson(2015). ResultsfromTable 1 illustrate thatcrosscountrydifferences have astronginfluence on Piketty’s r> g since fixedeffectscoefficientsare significantlylessthanpooledOLScoefficients.The lastcolumnin Table 1 estimatesthatthe coefficientonPiketty’sK/Yissignificantbutnegativeinafixed effectsregression.Thisimpliesthat anincrease inCSNIdecreasestopincome inequalitywhich contradictsPiketty’s theory. To continue withthe resultsfoundin(6) of Table 1 the nexttwotable shows the resultsfrom regressions againstdifferentcontrolsandfordifferenttimespans. The resultsshow that K/Yremains negative andsignificantafteraddingthe controls howeveroverthe period1999-2010 itbecomes 3 Additional regressionsarekept in the appendix.
  • 13. 12 insignificantasshownin(4) of Table 3. Overthe same periodin(3) of Table 3 r > g ispositive and significant. Table 1 Piketty's r > g and Capital Accumulation 1961-2010 1951-2010 1961-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 1% Share of Income (Log) r - g 0.00921 (1.34) -0.00246 (-1.02) 0.0103 (1.43) 0.00122 (0.48) normalizing r=0 (r-g=-g) 0.00818 (0.96) -0.00155 (-0.79) K/Y (Log) -0.125 (-0.64) -0.536** (-2.36) Country Fixed Effects No Yes No Yes No Yes Observations 829 829 1339 1339 829 829 R2 0.284 0.202 0.293 R2 overall 0.233 0.194 0.173 R2 between 0.152 0.173 0.0217 R2 within 0.561 0.462 0.617 # of countries in regression 27 27 27 27 27 27 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01
  • 14. 13 Table 2 Piketty's r > g and Capital Accumulation with different controls 1961-2010 (1) (2) (3) (4) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 1% Share of Income (Log) r - g 0.00184 (0.77) 0.00221 (0.92) 0.00120 (0.47) 0.00259 (1.15) K/Y (Log) -0.498** (-2.46) -0.487** (-2.56) -0.559** (-2.45) -0.506*** (-2.85) Index of human capital per person (Log) -0.752* (-1.82) -0.605 (-1.55) Share of Government Consumption -1.930*** (-3.22) -1.762*** (-3.21) Population growth (annual %) -0.0182 (-0.89) -0.0369* (-1.94) Country Fixed Effects Yes Yes Yes Yes Observations 829 829 829 829 R2 overall 0.193 0.358 0.155 0.331 R2 between 0.0544 0.214 0.0120 0.187 R2 within 0.643 0.651 0.618 0.672 # of countries in regression 27 27 27 27 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01
  • 15. 14 Table 3 Piketty's r > g and Capital Accumulation over different time spans 1980-2010 1999-2010 (1) (2) (3) (4) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 1% Share of Income (Log) r - g 0.0106 (1.34) 0.000222 (0.09) 0.0151* (1.83) -0.00308 (-1.52) K/Y (Log) -0.0937 (-0.48) -0.394** (-2.21) -0.0992 (-0.45) -0.0203 (-0.09) Country Fixed Effects No Yes No Yes Observations 681 681 235 235 R2 0.284 0.101 R2 overall 0.200 0.0174 R2 between 0.0449 0.0110 R2 within 0.624 0.165 # of countries in regression 27 27 27 27 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 In Table 4 all the control variables are includedandthe Pikettyvariablesare regressedonseveral measuresof topincome inequality.The regressions show thatincreasingr-gincreasesthe top1% share of income asseenin (1) of Table 4 butwhencountryfixedeffectsare includedthe coefficientbecomes insignificant.In(6) of Table 4 K/Y has a significantandpositiverelationshipwiththe ParetoLorenz coefficientandsoan increase inK/Ydecreasesincome inequality whichiswhatwasfoundinTable 1 to 3. This leadstoconflictingresultsintermsof analyzingPiketty’sargument.Favourable resultsforPiketty are foundinthe pooledOLS regressionsbut notinfixedeffectsregressions.
  • 16. 15 Table 4 Piketty's r > g and Capital Accumulation – Top Income Inequality 1980-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) r - g 0.00904** (2.06) 0.000327 (0.15) 0.00433 (1.03) -0.000450 (-0.32) -0.00850** (-1.99) -0.000450 (-0.32) K/Y (Log) -0.116 (-0.67) -0.380 (-1.38) -0.110 (-0.65) 0.110 (0.46) 0.179 (1.13) 0.379** (1.99) Index of human capital per person (Log) -0.680* (-1.75) -0.673 (-1.06) -0.293 (-0.89) -0.132 (-0.24) 0.234 (0.69) 0.197 (0.42) Share of Government Consumption -4.549*** (-3.89) -1.896*** (-2.68) -2.050*** (-2.73) -0.232 (-0.45) 2.592*** (3.43) 1.172** (2.01) Population growth (annual %) 0.0549 (1.05) -0.0361 (-1.60) -0.00665 (-0.13) -0.0144 (-0.72) -0.0144 (-0.33) 0.0125 (0.58) Services per capita (% of GDP) -0.00134 (-0.13) -0.00357 (-0.08) -0.00235 (-0.28) -0.00361 (-0.07) 0.00128 (0.19) -0.00387 (-0.20) Employment to population ratio, 15+, total (%) (national estimate) -0.00987* (-1.75) -0.00485 (-0.79) -0.00943** (-2.21) -0.00218 (-0.76) -0.00173 (-0.28) 0.00206 (0.40) Country Fixed Effects No Yes No Yes No Yes Observations 502 502 444 444 450 450 R2 0.713 0.597 0.539 R2 overall 0.509 0.283 0.366 R2 between 0.468 0.233 0.227 R2 within 0.642 0.538 0.659 # of countries in regression 26 26 23 23 24 24 t statistics in parentheses
  • 17. 16 Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 3.3 Schumpeterian Growth Theory and Innovation Rents: Results from Pooled OLS and Fixed Effects Regressions So far our analysisprovides conflictingevidence forPiketty’sargumentonCSNI drivingtopincome inequality. Basedonfixedeffectsthe analysisis inagreementwithAcemogluandRobinson(2015).In thissectioninnovationisallowedinthe model andisfirstmeasuredby total patentsgranted andthen by bookto market.The resultsmetfrombothmeasuressupport Schumpeteriangrowththeorybutdoes not debunkPiketty’sargumentforcapital accumulationincreasingtopincome inequality. 3.3.1 MeasureofInnovation:PatentsGranted Table 5 displays the results whentotal patentsgrantedis includedinthe regression. The pooledOLS regression ontop10% share of income showssignificantresults findinga1% increase inpatentsgranted increasesthe top10% share of income by6.6%. Alsointhe table,(1) shows r-gis positive andsignificant and K/Y becomesinsignificantforall regressions givingevidence thatinnovationisanomittedvariable in Section3.2 regressions. The resultsinTable 5 give some evidence insupportof Schumpeteriangrowththeoryandinnovation rentsand evidence againstPiketty’s supposition.Howeverthe regressionsinTable 5covera short time span, 1999-2010, and so general inferencescannotbe drawn. To ignore the insignificantcoefficientsonpatentsgrantedin (1) and(2) of Table 5 wouldbe unjust since the top 1% share of income is one of the income inequalitymeasuresof interest. There isanapparent omittedvariable whichis R&Dactivitysince thisisverylikelycorrelatedwithpatentsgrantedand can likelybe correlatedwithourdependentvariables. R&Dactivityapproximatesthe total costof innovations.AccordingtoSchumpeteriangrowththeoryasoutlinedinSection1.2,innovatorswill produce an innovation if the expectedtotal profit4 fromaninnovation isequal toorgreaterthan the expected costof innovation. If aninnovation,measuredbypatentsgranted,occurs thenthe innovator has incurredthe cost of innovation whichisapproximatedbyR&Dactivity. Therefore R&Dactivity affectspatentsgrantedwhere higherR&Dactivityleadstoan increase inpatentsgranted.R&Dactivity can alsoaffectinnovationrents since the costof innovationhasanimpacton innovationrents. R&D activitycan thushave a positive effectoninnovationrentsif the returnonR&D activityispositive anda negative effectif the returnonR&D activityisnegative. Table 6 controlsforR&D activityand providesasignificantand twice aslarge coefficientforpatents grantedon the top 10% share of income pooledOLSregression thanwhatwaspresentedinTable 5, 4 Profitas defined by the equation 𝜋𝑖𝑡 = ( 𝑝𝑖𝑡 − 𝑀𝐶𝑖𝑡 ) 𝑦𝑖𝑡 in Section 1.2. Expected total profitis the expected present valueof the stream of all profits earned up to time T. See Aghion, Akcigitand Howitt (2013).
  • 18. 17 howeverregressionsonthe top1% share of income remaininsignificant.5 The keyresultinTable 6is in (1) where r-gremains positive andsignificantinexplainingthe top 1% share of income. Table 5 – Check appendix Table A1 for complete list of excluded countries. Innovation Measure: Patents Granted – Top Income Inequality 1999-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) Patents Granted (Log) 0.0243 (0.45) 0.0368 (0.55) 0.0664** (2.19) 0.0410 (0.79) -0.00404 (-0.15) -0.0186 (-0.63) r - g 0.0107* (1.85) -0.00403 (-1.29) 0.00422 (0.83) -0.000130 (-0.05) -0.00636 (-1.38) 0.00313** (2.36) K/Y (Log) -0.156 (-0.62) -0.0816 (-0.19) -0.234 (-1.07) -0.0103 (-0.05) 0.119 (0.70) -0.0280 (-0.11) Index of human capital per person (Log) -0.811 (-0.99) 1.784 (1.05) -0.594 (-0.98) 0.831 (1.01) 0.106 (0.23) -0.665 (-1.02) Share of Government Consumption -3.994** (-2.24) -0.151 (-0.13) -1.206 (-0.99) -0.508 (-1.41) 2.163** (2.25) -0.370 (-0.79) Population growth (annual %) 0.0954 (1.28) -0.0113 (-0.28) 0.0551 (1.07) -0.000113 (-0.01) -0.0214 (-0.55) -0.000475 (-0.02) Services per capita (% of GDP) -0.00186 (-0.09) 0.0961 (1.49) 0.00370 (0.31) 0.0391 (1.20) 0.00350 (0.39) -0.0680*** (-3.05) Employment to population ratio, 15+, total (%) (national estimate) -0.00779 (-0.95) 0.00377 (0.38) -0.0100 (-1.56) -0.00575 (-1.36) -0.00516 (-0.72) -0.00766 (-1.43) 5 By addingR&D activity to the regressions a few countries were dropped however Table 5 results were consistent when the same countries were excluded and therefore the results arenot tainted by any selection bias.
  • 19. 18 Country Fixed Effects No Yes No Yes No Yes Observations 220 220 202 202 202 202 R2 0.516 0.574 0.388 R2 overall 0.0623 0.00183 0.000000567 R2 between 0.0437 0.000400 0.00327 R2 within 0.255 0.270 0.370 # of countries in regression 25 25 23 23 24 24 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01
  • 20. 19 Table 6 - Check appendix Table A1 for complete list of excluded countries. Innovation Measure: Patents Granted controlling for R&D – Top Income Inequality 1999-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) Patents Granted (Log) 0.105 (1.39) 0.0367 (0.54) 0.122*** (2.90) 0.0447 (0.68) 0.0459 (0.94) -0.00978 (-0.33) R&D per capita expenditure (Log) -0.362 (-1.51) 0.136 (1.28) -0.355*** (-2.58) 0.0360 (0.37) -0.211 (-1.34) -0.103 (-1.54) R&D researchers (Log) 0.246 (1.00) -0.108 (-1.18) 0.283** (2.50) -0.0496 (-0.75) 0.130 (0.86) 0.0746 (1.16) r - g 0.0129** (2.01) -0.00294 (-0.89) 0.00455 (0.97) -0.000283 (-0.10) -0.00665 (-1.44) 0.00304* (1.96) K/Y (Log) -0.141 (-0.54) -0.0495 (-0.12) -0.196 (-1.03) 0.0442 (0.20) 0.131 (0.71) -0.0704 (-0.28) Index of human capital per person (Log) -0.950 (-0.91) 2.372 (1.60) -0.582 (-1.03) 1.078 (1.05) -0.0223 (-0.04) -0.949 (-1.12) Share of Government Consumption -4.648*** (-2.74) -0.289 (-0.55) -1.801* (-1.90) -0.328 (-0.92) 1.910** (2.00) -0.456 (-0.75) Population growth (annual %) 0.107 (1.34) -0.0106 (-0.27) 0.0654 (1.40) -0.000541 (-0.03) -0.0164 (-0.46) -0.00186 (-0.09) Services per capita (% of GDP) 0.00965 (0.44) 0.0937* (1.78) 0.00941 (0.77) 0.0383 (1.09) 0.0102 (0.84) -0.0650** (-2.47) Employment to population ratio, 15+, total (%) -0.00521 (-0.51) -0.00369 (-0.60) -0.00973 (-1.50) -0.00499 (-1.22) -0.00343 (-0.44) -0.00421 (-0.89)
  • 21. 20 (national estimate) Country Fixed Effects No Yes No Yes No Yes Observations 209 209 197 197 192 192 R2 0.625 0.707 0.484 R2 overall 0.0597 0.00287 0.0153 R2 between 0.0231 0.000901 0.0176 R2 within 0.359 0.274 0.411 # of countries in regression 24 24 23 23 23 23 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 3.3.2 MeasureofInnovation:Bookto Market As statedinthe previoussubsection,there isashorttime spanof available dataforpatentsgrantedand so the advantage gainedfromusingthe booktomarket ratioas a measure of innovationisalongertime span. The disadvantage,however,isthat b/misan indirectmeasure of innovative activity since itdoes not directly capture innovative activitylike the more commonlyusedmeasure patentsgranted, but insteaditapproximates the aggregate value of innovation. AccordingtoSchumpeteriangrowth theory as presentedinSection1.2,the value of an innovationdecreasesovertime whichcorrespondstoan increasingb/m.If a newinnovation occurs ina givenperiod thenthe value of thatinnovation isgreater than the one that precededit andso there will be acorrespondingdecreasein the b/m. The results fromregressingincome inequalitymeasureson b/mare displayedin Table 7.The regressions inTable 7 coverthe period1980-2010 whichisthe same periodcoveredinthe Pikettyregressions in Section3.2. The table predicts thatboth r-g andb/m contribute to the top 1% share of income however r-g affectsthe top1% share of income ina pooledOLSregressionwhereasb/maffectsthe top1% share of income ina fixedeffectsregression.Betweenthe resultsinSection3.3.1and Table 7 there is some evidence thatinnovation increasestopincome inequalityandalsothatina pooledOLSthere isa positive andsignificantrelationship betweenr-gandthe top 1% share of income. Kitchensinkregressions6 ontopincome inequalityare performedandthe resultsare capturedinTable 8. The mostsignificantresultsare foundfor pooledOLS regressionon the top1% and top10% share of income.Bothinnovationmeasuresare significantinthe correctdirectionindicatingthatinnovation increasestopincome inequalityandmore sofor the top 1% share of income thanthe top 10% share of income. Furthermore the Pikettyvariable r-gbecomesinsignificant. Table 8providessome more evidence thatinnovationrentsexplain topincome inequalityratherthanthe Pikettyvariables.However the resultsare not robustto countryfixedeffects. 6 Kitchen sink regression includes all variables of interestand controls.
  • 22. 21 Table 7 - Check appendix Table A1 for complete list of excluded countries. Innovation Measure: Book to Market – Top Income Inequality 1980-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) Book to Market (Log) -0.0447 (-0.89) -0.0578* (-1.78) -0.0542 (-1.43) -0.0149 (-0.80) 0.000460 (0.01) 0.00975 (0.51) r - g 0.00763* (1.65) -0.000391 (-0.11) 0.00436 (1.08) 0.00150 (0.67) -0.00202 (-0.45) 0.00163 (0.54) Index of human capital per person (Log) -1.051** (-2.36) -1.220** (-2.04) -0.299 (-1.00) -0.340 (-0.71) 0.215 (1.00) 0.517 (1.12) Share of Government Consumption -3.669*** (-3.18) -1.340 (-1.50) -1.765** (-2.08) -0.0889 (-0.18) 1.852*** (3.06) 1.320 (1.40) Population growth (annual %) 0.0436 (0.91) -0.0137 (-0.55) -0.0215 (-0.41) -0.0109 (-0.56) -0.0186 (-0.50) 0.000551 (0.02) Services per capita (% of GDP) -0.00677 (-0.93) -0.0146 (-0.29) -0.00271 (-0.39) -0.00653 (-0.12) 0.00373 (0.75) 0.00459 (0.20) Employment to population ratio, 15+, total (%) (national estimate) -0.00382 (-0.77) -0.00816 (-1.26) -0.00966** (-1.98) -0.00413* (-1.78) -0.00723** (-1.99) 0.00440 (0.76) Country Fixed Effects No Yes No Yes No Yes Observations 433 433 408 408 388 388 R2 0.723 0.593 0.600 R2 overall 0.582 0.430 0.370 R2 between 0.444 0.362 0.155 R2 within 0.670 0.550 0.648 # of countries in regression 25 25 23 23 23 23
  • 23. 22 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 Table 8 - Check appendix Table A1 for complete list of excluded countries. Kitchen Sink – Top Income Inequality 1999-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) Book to Market (Log) -0.138* (-1.72) -0.0608 (-0.89) -0.116* (-1.93) -0.0338 (-1.07) 0.0521 (0.90) -0.0184 (-0.43) Patents Granted (Log) 0.166** (2.33) 0.0408 (0.50) 0.147*** (3.93) 0.0562 (0.85) 0.0199 (0.40) 0.00391 (0.11) R&D per capita expenditure (Log) -0.659*** (-2.66) 0.0508 (0.24) -0.517*** (-4.04) -0.0274 (-0.24) -0.0921 (-0.55) -0.189* (-1.75) R&D researchers (Log) 0.413* (1.90) -0.0785 (-0.64) 0.368*** (3.69) -0.0358 (-0.56) 0.0778 (0.56) 0.0833 (0.95) r - g 0.0111 (1.48) -0.00428 (-0.78) 0.00280 (0.67) -0.0000555 (-0.02) -0.00373 (-0.69) 0.00365 (1.61) Index of human capital per person (Log) -0.873 (-1.06) 2.269 (1.26) -0.256 (-0.51) 0.952 (1.02) -0.103 (-0.23) -0.828 (-0.70) Share of Government Consumption -3.661** (-2.15) -0.144 (-0.23) -1.400 (-1.37) -0.265 (-0.72) 1.362* (1.67) -0.0669 (-0.11) Population growth (annual %) 0.0658 (0.92) -0.00620 (-0.18) 0.0324 (0.70) 0.00180 (0.11) 0.00152 (0.05) 0.00275 (0.13)
  • 24. 23 Services per capita (% of GDP) 0.0162 (0.98) 0.113** (2.15) 0.0152 (1.43) 0.0475 (1.61) 0.00586 (0.62) -0.0610** (-2.01) Employment to population ratio, 15+, total (%) (national estimate) 0.00337 (0.45) -0.00219 (-0.27) -0.00837 (-1.45) -0.00390 (-0.82) -0.00734 (-1.18) -0.00424 (-0.81) Country Fixed Effects No Yes No Yes No Yes Observations 195 195 189 189 178 178 R2 0.666 0.744 0.463 R2 overall 0.0505 0.0166 0.00676 R2 between 0.0324 0.00713 0.0112 R2 within 0.368 0.286 0.442 # of countries in regression 24 24 23 23 23 23 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 3.4 Evidence of Piketty’s r>g and Schumpeter’s Innovation Rents: Interaction of Countries Grouped by the Global Competitiveness Index The countriesinthe panel have verydifferenteconomies andsosuch differencesshouldbe considered. The fixedeffectsestimatorisappropriate forcapturingthese differences.Howeveritisimportantto recognize thatdifferenteconomieswillreactdifferentlytochangesininnovationandPikettyvariables. Ratherthan interactingeachcountrywitheachmeasure of innovationandeachPikettyvariable the countriesare firstgroupedintermsof theircompetitivenessasmeasuredbythe GCI and then the interactionbetweenthe variablesof interestandthe GCIgroupsare estimated.Fourgroupsare created as seeninthe table belowwhere GCI0isthe group withthe mostcompetitive countriesandGCI3isthe groupwiththe leastcompetitive countries.
  • 25. 24 3.4.1 GCIGroupingTable Global Competitive Index (2015-2016) Categories Label Value Countries Total GCI0 GCI ≥ 5.5 Switzerland, Singapore, United States, Germany, Netherlands, Japan, Finland 7 GCI1 5.4 ≥ GCI ≥ 5 Sweden, United Kingdom, Norway, Denmark, Canada, New Zealand, Malaysia, Australia, France, Ireland, Korea Rep. 11 GCI2 4.9 ≥ GCI ≥ 4.5 China, Spain, Indonesia, Portugal, Italy 5 GCI3 4.4 ≥ GCI South Africa, India, Columbia, Argentina 4 3.4.2 Results Table 9 displays resultsfromfixedeffectsregressions7 showingthatanincrease in r>g increases the top 1% share of income forGCI0 countriesand thatthe effectof anincrease inr>g is nearlyzero or negative for mostnon-GCI0countries.Alsothe table showsthat the innovationmeasure patentsgranted hasno effectontopincome inequalitybutR&Dexpenditure does.ExcludingpatentsgrantedandtreatingR&D as a measure of innovation,(5) and(6) show that the coefficienton R&D expenditure percapitais significantandpositive forthe top1% and top10% share of income forGCI0 countriesbutfor the top 10% share of income is nearlyzeroornegative forall othercountries. These resultsshowthat aftercontrollingforfixedeffects andinnovation the Pikettyvariable r>g isnot significantinexplainingtopincome inequalityforthe most competitive economies.Thisisevidence againstPiketty’stheoryoncapital driventopincome inequality. Furthermore the resultsgive some justificationforSchumpeteriangrowththeory andinnovationrents. 7 Pooled OLS regressions areincluded in Appendix Table A6.
  • 26. 25 Table 9 - Check appendix Table A1 for complete list of excluded countries. Global Competitive Index Interaction Results – GCI0 is base Piketty: r>g, K/Y 1980-2010 Patents Granted 1999-2010 R&D: Expenditure 1996-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) r - g 0.0110* (1.88) 0.00121 (0.46) 0.00141 (0.23) 0.00425 (0.93) 0.000675 (0.20) 0.000542 (0.19) GCI1 x r - g -0.0131* (-1.67) -0.00312 (-0.82) -0.00830 (-0.97) -0.00704 (-1.35) -0.00926* (-1.65) -0.00542 (-1.47) GCI2 x r - g -0.0146** (-2.38) -0.00193 (-0.51) -0.00354 (-0.19) -0.00251 (-0.25) -0.00247 (-0.49) -0.00217 (-0.64) GCI3 x r - g -0.0104 (-1.49) 0.00176 (0.57) -0.00243 (-0.36) -0.00352 (-0.63) -0.00221 (-0.70) -0.00384 (-1.29) Patents Granted (Log) 0.0951 (0.67) 0.111 (1.28) GCI1 x Patents Granted (Log) -0.0752 (-0.48) -0.104 (-1.17) GCI2 x Patents Granted (Log) -0.121 (-0.70) -0.114 (-1.18) GCI3 x Patents Granted (Log) -0.101 (-0.69) -0.0783 (-0.90) R&D per capita expenditure (Log) 0.168* (1.67) 0.0993 (1.01) 0.571*** (2.72) 0.357*** (2.69) GCI1 x R&D per capita expenditure (Log) -0.363 (-1.59) -0.326** (-2.42) GCI2 x R&D per capita expenditure -0.519* (-1.94) -0.377** (-2.21)
  • 27. 26 (Log) GCI3 x R&D per capita expenditure (Log) -0.334 (-1.41) -0.407*** (-3.82) K/Y (Log) -0.717 (-1.39) 0.216 (0.77) 0.0334 (0.07) 0.106 (0.47) 0.344 (0.96) 0.353 (1.64) GCI1 x K/Y (Log) 0.373 (0.60) -0.107 (-0.20) GCI2 x K/Y (Log) 0.467 (0.57) -0.656 (-1.03) GCI3 x K/Y (Log) -0.232 (-0.23) 0.0525 (0.16) R&D researchers (Log) -0.0977 (-1.12) -0.0333 (-0.53) -0.121 (-0.99) -0.0380 (-0.55) Index of human capital per person (Log) -0.756 (-1.14) -0.0911 (-0.17) 2.286 (1.47) 0.942 (0.95) 0.835 (0.82) 1.048* (1.74) Share of Government Consumption -1.722** (-1.97) -0.287 (-0.61) -0.444 (-0.69) -0.561 (-1.50) -1.108 (-1.64) -0.327 (-0.79) Population growth (annual %) -0.0411* (-1.66) -0.0112 (-0.64) -0.00980 (-0.26) -0.00161 (-0.11) -0.0399 (-1.34) -0.0222* (-1.69) Services per capita (% of GDP) -0.00545 (-0.13) -0.00182 (-0.03) 0.0972* (1.89) 0.0401 (1.04) 0.0673 (1.09) 0.0431 (1.01) Employment to population ratio, 15+, total (%) (national estimate) -0.00515 (-0.79) -0.00183 (-0.72) -0.00368 (-0.56) -0.00576 (-1.40) 0.00617 (1.09) 0.000102 (0.02) Country Fixed Yes Yes Yes Yes Yes Yes
  • 28. 27 Effects Observations 502 444 209 197 278 255 R2 overall 0.0403 0.167 0.0433 0.000163 0.00318 0.00781 R2 between 0.000164 0.0887 0.0153 0.00104 0.0127 0.00819 R2 within 0.662 0.558 0.382 0.351 0.540 0.543 # of countries in regression 26 23 24 23 25 23 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 4 Discussion 4.1 Reverse Causality –Innovation on Income Inequality OLS Regression and Panel VAR It isworthwhile toinvestigatethe regressionof innovationonincome inequality inordertotruly investigateSchumpeteriangrowththeory.Accordingtotheory,highertopincome inequalityshouldlead to greaterinnovative activitysince inventors receive largerinnovationrentsandthus have a greater incentive toinnovate. If this isthe case then the coefficientonpatentsgrantedin ourregressionsontop income inequalitymeasures inSection3.3.1can likelysufferfromendogeneitycaused byreverse causalitysince patentsgrantedisameasure of innovative activity. The regressionsonbooktomarketin Section3.3.2 avoid some of this problembecause booktomarketis a measure of the value of innovative activitiesratherthaninnovative activity. Table 10 displaysthe resultswhenpatentsgrantedisregressedonincome inequalitymeasures. An increase inthe top10% share of income significantlyincreasespatentsgrantedasshownin (3) of Table 10. Thisresultsuggeststhathighertop 10% share of income increasesinnovativeactivity.However,the mostsignificantresultfromthe table isthatR&D expenditure percapitahasa significantlystrong positive effectonpatentsgranted whichremainsrobustafteraccountingfor countryfixedeffects.
  • 29. 28 Table 10 - Check appendix Table A1 for complete list of excluded countries. Reverse Causality – Patents Granted on Top Income Inequality 1999-2010 (1) (2) (3) (4) (5) (6) Patents Granted (Log) Patents Granted (Log) Patents Granted (Log) Patents Granted (Log) Patents Granted (Log) Patents Granted (Log) Top 1% Share of Income (Log) 0.950 (1.55) 0.157 (0.69) Top 10% Share of Income (Log) 2.817** (2.43) 0.492 (0.96) Pareto Lorenz Coefficient (log) 1.450 (0.94) -0.137 (-0.55) R&D per capita expenditure (Log) 2.138*** (3.00) 0.447 (1.33) 2.279*** (3.61) 0.779*** (2.77) 2.184*** (2.76) 0.562* (1.68) R&D researchers (Log) -0.567 (-0.76) 0.443* (1.73) -1.058* (-1.74) 0.334 (1.62) -0.549 (-0.60) 0.413 (1.51) r - g 0.00561 (0.32) -0.00712* (-1.71) 0.00865 (0.40) -0.00606 (-0.94) 0.0266 (1.30) -0.00836* (-1.75) K/Y (Log) 0.00361 (0.00) 0.986 (1.04) 0.471 (0.56) 0.522 (0.74) -0.289 (-0.26) 1.325 (1.32) Index of human capital per person (Log) 1.281 (0.39) 1.307 (0.43) 1.718 (0.62) 2.140 (0.56) 0.393 (0.11) 2.737 (0.80) Share of Government Consumption 1.413 (0.30) 2.711 (1.06) 2.328 (0.54) 1.768 (0.69) -5.480 (-0.97) 2.885 (0.86) Population growth (annual %) -0.270 (-0.96) 0.0272 (0.36) -0.338* (-1.83) -0.00215 (-0.03) -0.169 (-0.58) 0.0380 (0.46) Services per -0.186*** -0.0567 -0.154*** -0.0490 -0.196*** -0.0676
  • 30. 29 capita (% of GDP) (-3.89) (-0.51) (-3.24) (-0.57) (-3.35) (-0.59) Employment to population ratio, 15+, total (%) (national estimate) -0.00384 (-0.13) -0.0236 (-1.04) 0.0242 (0.70) -0.0132 (-0.60) -0.00461 (-0.12) -0.0325 (-1.37) Country Fixed Effects No Yes No Yes No Yes Observations 209 209 197 197 192 192 R2 0.925 0.930 0.923 R2 overall 0.793 0.707 0.806 R2 between 0.744 0.685 0.765 R2 within 0.642 0.764 0.651 # of countries in regression 24 24 23 23 23 23 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 To furtherinvestigatethe causal relationshipbetween topincome inequality andinnovation the paper conductstwo panel VARestimation;the firstbetweenthe top1% share of income and b/m, andthe secondbetweenthe top10%share of income and b/m.The numberof lagschosenisfour,basedon whatwas done inAtemsand Jones(2015). The resultsfromgrangercausalitytests showninFigure A1in the appendix support whatwas foundinthe precedingregressions;increasesinthe top10% share of income causesincreasesininnovation. Figure 5and Figure 6 show selected impulseresponse functions fromboth multivariate regressions.
  • 31. 30 Figure 5 – Impulse Response Function: Impulse (Book to Market), Response (Top 1% Share of Income) Figure 6 - Impulse Response Function: Impulse (Top 10% Share of Income), Response (Book to Market) -2 -1.5 -1 -.5 0 0 5 10 bm : top1inc IRF step impulse : response -.005 0 0 5 10 top10inc : bm IRF step impulse : response
  • 32. 31 4.2 Pre and Post Y2K Regressions – Evidence of a New Economy So far the analysishasconsistentlygivendifferentresultsbetweenregressionswithtime spansstarting fromand after1997 to 2010 and regressionswherethe time periodcoveredextendfromthe 1980s to 2010. Therefore thissectionfurtherexploresthe difference inresults.Table 11displaysresultsfor regressionoverthe full period, pre-2000 andpost-2000. The year 2000 can be consideredaturningpointineconomichistory.The endof the dot-combubble had begunandit putout many incumbentsinthe relativelyyounginformationtechnology(IT) industry. Between1990 and 2000 there washeavyinvestmentintothe ITindustrywhichledtoan IT stock bubble whichiscommonlyreferred toasthe dot-combubble.8 The endof the dot-combubble resultedinless incumbentfirmsinthe ITindustryandexcessITinfrastructure,whichgave rise tolowerentrybarriers and thusincreasingcompetitionandinnovativeactivitiesinthe ITindustry. The post2000 IT industry differsfromotherindustriesbecauseitrequireslittle capital,littlelabour,andhighabilityworkers(high levelsof humancapital) forfirmstobe successful.The dot-combubbleenableddigitalizationwhichhas transformed the advancedeconomiesintoadigital economy(Stiroh,2004). HémousandOlsen(2014) suggestthatthe digital economycreateslargerhumancapital andinnovationrents resultinginhigher income inequality. Before 2000 innovationplaysnorole inexplainingincreases intopincome inequality andhumancapital issignificantandnegativelycorrelatedwiththe top1% share of income asshowninTable 11. However after2000 Table 11 showsthat innovationincreasestop1% and top10% share of income forGCI0 countries while humancapital becomesinsignificantwithapositivecoefficient.Table 11also showsthat the coefficienton r-ghas become significantandpositiveforGCI0countriespost2000 whichgivessome indicationforPiketty’shypothesis thatthe future of capitalismtendsto capital accumulationdriven income inequality.The increasingsignificanceof innovationandhumancapital inTable 11 provides evidence thatincreasesininnovationandhumancapital rents explainsrecenttrendsintopincome inequality. 8 See “The Boom and Bust in Information Technology Investment” by Mark Doms. http://www.frbsf.org/economic- research/files/er19-34bk.pdf.
  • 33. 32 Table 11 - Check appendix Table A1 for complete list of excluded countries. Y2K Regressions: Book to Market pre and post 2000 – GCI0 is base 1980-2010 1980-2000 2000-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Book to Market (Log) -0.0812 (-1.10) -0.00325 (-0.14) -0.0534 (-0.66) 0.00927 (0.26) -0.116** (-2.03) -0.0465* (-1.70) GCI1 x Book to Market (Log) 0.0228 (0.29) -0.0242 (-1.12) 0.0208 (0.22) -0.0142 (-0.44) 0.0380 (0.51) -0.00547 (-0.20) GCI2 x Book to Market (Log) 0.100 (0.67) -0.000403 (-0.00) 0.0941 (0.35) 0.0140 (0.14) 0.00673 (0.02) -0.0175 (-0.15) GCI3 x Book to Market (Log) -0.0920 (-1.10) 0.0488 (1.34) 0.131 (0.94) -0.0409 (-0.72) 0.0340 (0.31) 0.0986*** (2.58) r - g 0.0102** (2.38) 0.00288 (1.21) 0.00590 (0.97) -0.00126 (-0.57) 0.00517 (1.02) 0.00670* (1.74) GCI1 x r - g -0.0144** (-2.07) -0.00281 (-0.79) -0.0119 (-1.56) -0.000886 (-0.28) -0.0129 (-1.43) -0.00878* (-1.94) GCI2 x r - g -0.0162 (-0.76) -0.00102 (-0.06) -0.0206 (-0.55) -0.000654 (-0.04) -0.00319 (-0.06) -0.000171 (-0.01) GCI3 x r - g -0.0110** (-2.31) -0.0000623 (-0.02) -0.00789 (-0.99) -0.000938 (-0.30) -0.00707 (-1.46) -0.00882*** (-2.86) Index of human capital per person (Log) -1.210** (-1.99) -0.349 (-0.70) -1.723** (-2.42) -0.572 (-1.51) 1.391 (0.91) 0.544 (0.73) Share of Government Consumption -1.344 (-1.59) 0.0252 (0.05) -0.568 (-0.45) 0.338 (0.47) -0.351 (-0.64) -0.495 (-1.47) Population growth (annual -0.0136 (-0.59) -0.0115 (-0.59) 0.0154 (0.13) -0.00795 (-0.22) -0.00565 (-0.22) -0.00421 (-0.53)
  • 34. 33 %) Services per capita (% of GDP) -0.0132 (-0.30) -0.00716 (-0.12) -0.00184 (-0.03) 0.00168 (0.04) 0.121** (2.21) 0.0414* (1.83) Employment to population ratio, 15+, total (%) (national estimate) -0.00741 (-1.16) -0.00374* (-1.70) -0.0106 (-1.27) -0.00385 (-1.37) 0.000607 (0.12) -0.00283 (-0.89) Country Fixed Effects Yes Yes Yes Yes Yes Yes Observations 433 408 271 256 183 172 R2 overall 0.606 0.0668 0.502 0.264 0.0470 0.171 R2 between 0.572 0.0191 0.263 0.212 0.0418 0.103 R2 within 0.692 0.567 0.589 0.479 0.435 0.391 # of countries in regression 25 23 22 21 24 22 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 5 Conclusion Thispaperhas analyzedthe effectof CSNI andinnovationon topincome inequality.The resultsshow that neitherCSNI norinnovationaffects the restof the income distributionasmeasuredbythe Pareto Lorenzcoefficient.Itisfoundthatinnovation issignificantinexplainingtopincome inequality where an increase ininnovationincreasesthe top1% andtop 10% share of income aspredictedinSchumpeter’s growthmodel;howeverthese resultsdonotholdforthe leastcompetitive economies. Furthermore regressions predictthatincreasesinthe top10% share of income leadtoincreases ininnovativeactivity whichfurthervalidatesSchumpeteriangrowththeory. Insupportof Pikettythe paperfinds evidence that as an economybecomesmore competitive r-gbecomesmore significantinexplainingtopincome inequality trends.Lastly,resultsindicate that humancapital rents mightplay asignificantrole in explainingtopincome inequalitytrends inthe future. The limitations realizedinthispaperare a few.The paperwas limited toarelativelysmall numberof observationsforthe analysisbutthisissue wasmitigatedbyusingpairs-clusterbootstrapprocedure whichhas betterfinite sample propertiesthanclusteredstandarderrors.Alsointermsof observations, the paperwas limitedto27 countrieswithmostof the observationscomingfrom OECDcountriesin Europe and NorthAmericaandthus presentingsignificantselectionbias,yetthe analysisislargelybased on topincome inequalityof advancedeconomies andtherefore beingrestrictedtothese countriesis
  • 35. 34 acceptable.Lastlythe time spanpresented significantchallengeswhichthe paperattemptedto overcome byintroducingbooktomarketas a measure of innovation. References Acemoglu, J., Robinson, J. A. (2015).The Rise and Decline of General Laws of Capitalism Journal of Economic Perspectives, 29(1), 3-28. Retrieved from http://economics.mit.edu/files/11348 Aghion, P., Akcigit, U., Bergeaud, A., Blundell, R. W., & Hemous, D. (2015).Innovation and top income inequality C.E.P.R. Discussion Papers, CEPR Discussion Papers: 10659. Retrieved from http://sfx.scholarsportal.info/guelph/docview/1700663907?accountid=11233 Aghion, P., Akcigit, U., & Howitt, P. (2013).What Do We Learn From Schumpeterian Growth Theory? National Bureau of Economic Research, Inc, NBER Working Papers: 18824. Retrieved from http://www.nber.org/papers/w18824 Atems, B., & Jones, J. (2015). Income inequality and economic growth: A panel VAR approach. Empirical Economics, 48(4), 1541-1561. doi:http://dx.doi.org/10.1007/s00181-014- 0841-7 Falkinger, J., & Zweimuller, J. (1997). The impact of income inequality on product diversity and economic growth. Metroeconomica, 48(3), 211-237. Retrieved from http://sfx.scholarsportal.info/guelph/docview/56703414?accountid=11233 Hémous, D., & Olsen, M. (2014). The rise of the machines: Automation, horizontal innovation and income inequality C.E.P.R. Discussion Papers, CEPR Discussion Papers: 10244. Retrieved from http://sfx.scholarsportal.info/guelph/docview/1660008993?accountid=11233 Hsing, Y. (2005). Economic growth and income inequality: The case of the US.International Journal of Social Economics, 32(7), 639-647. Retrieved from http://sfx.scholarsportal.info/guelph/docview/56597962?accountid=11233 Lee, N., & Rodriguez-Pose, A. (2013). Innovation and spatial inequality in europe and USA. Journal of Economic Geography, 13(1), 1-22. Retrieved from http://sfx.scholarsportal.info/guelph/docview/1314321790?accountid=11233
  • 36. 35 Raff, Daniel M G (Review of: Scherer,Frederic M., & Perlman, M., eds). (1993). Review of: Entrepreneurship, technological innovation, and economic growth: Studies in the schumpeterian tradition. Journal of Economic Literature, 31(4), 2017-2018. Retrieved from http://sfx.scholarsportal.info/guelph/docview/56619242?accountid=11233 Stiroh,Kevin (Review of: Greenan, Nathalie, L'Horty, Y., & Mairesse, J., eds). (2004). Review of: Productivity, inequality, and the digital economy: A transatlantic perspective. Journal of Economic Literature, 42(1), 218-219. Retrieved from http://sfx.scholarsportal.info/guelph/docview/56460277?accountid=11233 Tselios, V. (2011). Is inequality good for innovation? International Regional Science Review, 34(1), 75-101. Retrieved from http://sfx.scholarsportal.info/guelph/docview/855184089?accountid=11233 Voitchovsky, S. (2005). Does the profile of income inequality matter for economic growth?: Distinguishing between the effects of inequality in different parts of the income distribution. Journal of Economic Growth,10(3), 273-296. doi:http://dx.doi.org/10.1007/s10887- 005-3535-3 Zweimuller, J. (2000). Schumpeterian entrepreneurs meet engel's law: The impact of inequality on innovation-driven growth. Journal of Economic Growth, 5(2), 185-206. Retrieved from http://sfx.scholarsportal.info/guelph/docview/57019108?accountid=11233
  • 37. 36 Appendix Table A1 Tables Excluded Countries Table 1 None Table 2 None Table 3 None Table 4 (1)-(2) India (3)-(4) Argentina, Columbia, India, Indonesia (5)-(6) Argentina, Finland, India Table 5 (1)-(2) India, Argentina (3)-(4) Argentina, Columbia, India, Indonesia (5)-(6) Argentina, Finland, India Table 6 (1)-(2) India, Argentina, Indonesia (3)-(4) Argentina, Columbia, India, Indonesia (5)-(6) Argentina, Finland, India, Indonesia Table 7 (1)-(2) India, Indonesia (3)-(4) Argentina, Columbia, India, Indonesia (5)-(6) Argentina, Finland, India, Indonesia Table 8 (1)-(2) India, Argentina, Indonesia (3)-(4) Argentina, Columbia, India, Indonesia (5)-(6) Argentina, Finland, India, Indonesia Table 9 (1) India (2) Argentina, Columbia, India, Indonesia (3) India, Argentina, Indonesia (4) Argentina, Columbia, India, Indonesia (5) India, Indonesia (6) Argentina, Columbia, India, Indonesia Table 10 (1)-(2) India, Argentina, Indonesia (3)-(4) Argentina, Columbia, India, Indonesia (5)-(6) Argentina, Finland, India, Indonesia Table 11 (1) India, Indonesia (2) Argentina, Columbia, India, Indonesia (3) India, Indonesia, Canada, Columbia, China (4) India, Argentina, Indonesia, Canada, Columbia, (5) India, Indonesia, Portugal (6) Argentina, Columbia, India, Indonesia, Portugal
  • 38. 37 Table A2 Innovation Measure: Patents Granted 1999-2014 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) Patents Granted (Log) -0.00792 (-0.25) 0.0428 (0.71) 0.0317 (1.58) 0.0354 (0.88) -0.00540 (-0.40) -0.0320 (-1.02) Country Fixed Effects No Yes No Yes No Yes Observations 303 303 284 284 276 276 R2 0.0320 0.113 0.0600 R2 overall 0.000542 0.110 0.0281 R2 between 0.000126 0.0913 0.000743 R2 within 0.150 0.130 0.233 # of countries in regression 26 26 23 23 25 25 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 Table A3 Innovation Measure: Patents Granted 1999-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) Patents Granted (Log) -0.0120 (-0.29) 0.0495 (0.84) 0.0456* (1.95) 0.0507 (1.30) -0.00219 (-0.12) -0.0266 (-0.75) r - g 0.0199* (1.89) -0.00269 (-0.98) 0.00872 (1.13) 0.000323 (0.14) -0.00735 (-1.10) 0.00171* (1.68) K/Y (Log) -0.140 (-0.59) -0.161 (-0.45) -0.243 (-1.11) -0.0480 (-0.26) 0.123 (1.00) 0.0507 (0.22) Country Fixed Effects No Yes No Yes No Yes Observations 229 229 210 210 211 211 R2 0.131 0.248 0.103
  • 39. 38 R2 overall 0.000312 0.156 0.00727 R2 between 0.0000748 0.103 0.00183 R2 within 0.164 0.152 0.223 # of countries in regression 26 26 23 23 25 25 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 Table A4 Innovation Measure: Book to Market 1980-2014 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) Book to Market (Log) -0.0395 (-0.46) -0.0139 (-0.51) -0.101* (-1.90) -0.00175 (-0.14) 0.0150 (0.42) 0.00322 (0.16) Country Fixed Effects No Yes No Yes No Yes Observations 685 685 640 640 621 621 R2 0.225 0.256 0.302 R2 overall 0.216 0.0911 0.295 R2 between 0.128 0.00617 0.0537 R2 within 0.225 0.525 0.426 0.535 # of countries in regression 25 25 23 23 23 23 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01
  • 40. 39 Table A5 Innovation Measure: Book to Market 1980-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 10% Share of Income (Log) Pareto Lorenz Coefficient (log) Pareto Lorenz Coefficient (log) Book to Market (Log) -0.121** (-2.18) -0.0251 (-0.81) -0.107* (-1.91) -0.00251 (-0.18) 0.0143 (0.43) 0.00401 (0.24) r - g 0.00489 (0.77) -0.00121 (-0.32) 0.00334 (0.68) 0.000753 (0.38) 0.00150 (0.34) 0.00239 (0.97) Country Fixed Effects No Yes No Yes No Yes Observations 595 595 570 570 550 550 R2 0.351 0.268 0.324 R2 overall 0.294 0.0835 0.319 R2 between 0.164 0.000296 0.0331 R2 within 0.591 0.511 0.574 # of countries in regression 25 25 23 23 23 23 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01 Table A6 Global Competitive Index Interaction Results – GCI0 is base Piketty: r>g, K/Y 1980-2010 Patents Granted 1999-2010 R&D: Expenditure 1996-2010 (1) (2) (3) (4) (5) (6) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) Top 1% Share of Income (Log) Top 10% Share of Income (Log) r - g 0.00770 (0.95) -0.00149 (-0.30) 0.0179* (1.85) 0.00484 (0.73) 0.0184 (1.39) 0.00694 (0.91) GCI1 x r - g -0.00442 (-0.42) 0.00360 (0.57) -0.0187 (-1.46) -0.00863 (-0.96) -0.0230 (-1.63) -0.00849 (-1.08) GCI2 x r - g -0.000829 (-0.08) 0.000689 (0.09) -0.00864 (-0.26) -0.00378 (-0.30) -0.0155 (-0.95) -0.00931 (-0.86)
  • 41. 40 GCI3 x r - g -0.00816 (-0.94) 0.00431 (0.64) -0.0174 (-1.26) -0.0155* (-1.82) -0.0140 (-0.95) -0.00329 (-0.33) Patents Granted (Log) 0.101 (0.86) 0.115** (2.35) GCI1 x Patents Granted (Log) 0.0126 (0.12) 0.0545 (1.02) GCI2 x Patents Granted (Log) -0.00923 (-0.05) -0.0609 (-0.81) GCI3 x Patents Granted (Log) -0.142 (-0.89) -0.229** (-2.52) R&D per capita expenditure (Log) -0.234 (-0.85) -0.271** (-1.98) -0.0521 (-0.30) -0.103 (-0.78) GCI1 x R&D per capita expenditure (Log) 0.00147 (0.08) -0.00572 (-0.60) GCI2 x R&D per capita expenditure (Log) -0.0257 (-0.68) -0.00836 (-0.39) GCI3 x R&D per capita expenditure (Log) 0.0259 (0.50) 0.0412 (1.19) K/Y (Log) -0.0854 (-0.46) -0.00443 (-0.03) -0.121 (-0.38) 0.0256 (0.13) -0.0316 (-0.13) -0.117 (-0.58) GCI1 x K/Y (Log) -0.0255 (-0.19) -0.0882 (-1.07) GCI2 x K/Y (Log) -0.275* (-1.73) -0.181** (-2.17) GCI3 x K/Y 0.186 0.328
  • 42. 41 (Log) (0.92) (1.14) R&D researchers (Log) 0.188 (0.65) 0.155 (1.31) 0.0935 (0.48) 0.227* (1.79) Index of human capital per person (Log) -0.913** (-2.16) -0.304 (-0.79) -0.601 (-0.47) 0.00836 (0.02) -0.857 (-0.77) -0.278 (-0.44) Share of Government Consumption -4.068*** (-3.06) -1.684** (-2.07) -3.354* (-1.93) -1.339 (-1.63) -4.414** (-2.32) -1.792 (-1.62) Population growth (annual %) 0.0276 (0.47) -0.0221 (-0.46) 0.137* (1.87) 0.0605 (1.62) 0.0564 (0.91) 0.0311 (0.65) Services per capita (% of GDP) -0.00345 (-0.33) -0.00414 (-0.61) 0.0110 (0.32) 0.0195 (1.22) -0.00774 (-0.53) -0.0131 (-1.37) Employment to population ratio, 15+, total (%) (national estimate) -0.00971* (-1.74) -0.00850** (-2.26) -0.00523 (-0.57) -0.00649 (-1.32) -0.00746 (-0.81) -0.00715 (-1.24) Country Fixed Effects No No No No No No Observations 502 444 209 197 278 255 R2 0.783 0.717 0.695 0.839 0.686 0.726 # of countries in regression 26 23 24 23 25 23 t statistics in parentheses Cluster Bootstrapped Standard Errors * p < 0.1, ** p < 0.05, *** p < 0.01
  • 43. 42 Figure A1 – Granger Causality Tables ALL 14.438 4 0.006 bm 14.438 4 0.006 top1inc ALL 6.120 4 0.190 top1inc 6.120 4 0.190 bm Equation Excluded chi2 df Prob > chi2 ALL 6.517 4 0.164 bm 6.517 4 0.164 top10inc ALL 11.512 4 0.021 top10inc 11.512 4 0.021 bm Equation Excluded chi2 df Prob > chi2