technology gaps, trade and income
TRANSCRIPT
Technology Gaps, Trade and Income
Thomas Sampson
London School of Economics
May 2019
Thomas Sampson (LSE) Technology Gaps May 2019
Technology gaps
Productivity differences generate cross-country income inequality(Caselli 2005) and determine Ricardian comparative advantage(Ricardo 1817, Costinot, Donaldson & Komunjer 2012)
Sources of productivity variation
Allocative efficiency (Acemoglu & Zilibotti 2001, Hsieh & Klenow2009)Technology gaps (Parente & Prescott 1994, Comin & Mestieri 2018)
Thomas Sampson (LSE) Technology Gaps May 2019
Technology gaps
Productivity differences generate cross-country income inequality(Caselli 2005) and determine Ricardian comparative advantage(Ricardo 1817, Costinot, Donaldson & Komunjer 2012)
Sources of productivity variation
Allocative efficiency (Acemoglu & Zilibotti 2001, Hsieh & Klenow2009)Technology gaps (Parente & Prescott 1994, Comin & Mestieri 2018)
Thomas Sampson (LSE) Technology Gaps May 2019
This paper
1 Theory of equilibrium technology gaps
Model how firm-level innovation and learning jointly shape theglobal productivity distribution and Ricardian comparativeadvantageNational innovation systems determine R&D efficiency (Nelson1993)Steady state technology gaps larger in more innovation-dependentindustries with lower advantage of backwardness and greaterlocalization of knowledge spillovers
2 How quantitatively important are technology gaps in explainingcross-country wage and income inequality?
Use R&D and bilateral trade data to estimate model and quantifywage and income variation due to technology gapsTechnology gaps account for around one-quarter to one-third ofnominal wage variation within OECD
Thomas Sampson (LSE) Technology Gaps May 2019
This paper
1 Theory of equilibrium technology gaps
Model how firm-level innovation and learning jointly shape theglobal productivity distribution and Ricardian comparativeadvantageNational innovation systems determine R&D efficiency (Nelson1993)Steady state technology gaps larger in more innovation-dependentindustries with lower advantage of backwardness and greaterlocalization of knowledge spillovers
2 How quantitatively important are technology gaps in explainingcross-country wage and income inequality?
Use R&D and bilateral trade data to estimate model and quantifywage and income variation due to technology gapsTechnology gaps account for around one-quarter to one-third ofnominal wage variation within OECD
Thomas Sampson (LSE) Technology Gaps May 2019
Related literature
Endogenous comparative advantageGrossman & Helpman 1991, Redding 1999, Somale 2016
Global productivity distributionParente & Prescott 1994, Eaton & Kortum 1999, Klenow &Rodrıguez-Clare 2005, Gancia, Muller & Zilibotti 2013, Buera &Oberfield 2016
Technology adoptionCaselli & Coleman 2001, Comin, Hobijn & Rovito 2008, Comin &Mestieri 2018
Idea flows and learningLucas & Moll 2014, Perla & Tonetti 2014, Sampson 2016
Incumbent innovationKlette & Kortum 2004, Atkeson & Burstein 2010, Akcigit & Kerr2016, Konig, Lorenz & Zilibotti 2016
Thomas Sampson (LSE) Technology Gaps May 2019
Environment
s = 1, . . . ,S countries, j = 1, . . . , J industries
Continuous time t , competitive markets
Labor is the only factor of production, constant population Ls
Firms differ along two dimensions: time invariant R&D capability ψand endogenous productivity θ
Firm that employs lP production workers has output
y = θ(
lP)β, 0 < β < 1
Thomas Sampson (LSE) Technology Gaps May 2019
R&D
Firm can upgrade productivity through R&D or adoption. R&Dtechnology
θ
θ= ψBs
(θ
χRjs
)−γj (lR)α− δ
Bs country-level R&D efficiency. Captures absolute advantage ininnovation due to national innovation system
χRjs R&D knowledge level in country s. Captures knowledge
spillovers
γj > 0⇒ growth decreasing in current productivity relative toknowledge level⇒ advantage of backwardness (Gerschenkron1962)
lR R&D employment, δ knowledge depreciation rate, α ∈ (0,1)returns to scale in technology investment
Thomas Sampson (LSE) Technology Gaps May 2019
Adoption
Adoption technology
θ
θ= BA
(θ
χAjs
)−γj (lA)α− δ
where lA denotes adoption employment
Differences from R&D technology
Firm’s R&D capability ψ does not enter the adoption technologyAdoption efficiency BA constant across countriesKnowledge level available for adoption χA
js = ηχRjs where η > 1
Thomas Sampson (LSE) Technology Gaps May 2019
Knowledge spillovers
Knowledge has two components1 Domestic knowledge depends on domestic productivity frontier θmax
js(excluding firm’s own productivity)
2 Global knowledge depends on global knowledge capital χj
χRjs =
(θmax
js
) κj1+κj χ
11+κjj
Localization of knowledge spillovers increasing in κj > 0
R&D spillovers generate growth in global knowledge capital
χj
χj=
S∑s=1
Mjs
∫ψλjs(ψ)lRjs (ψ)dG(ψ)
where Mjs denotes mass of firms and{λjs(ψ)
}determines the
strength of R&D spillovers
Thomas Sampson (LSE) Technology Gaps May 2019
Knowledge spillovers
Knowledge has two components1 Domestic knowledge depends on domestic productivity frontier θmax
js(excluding firm’s own productivity)
2 Global knowledge depends on global knowledge capital χj
χRjs =
(θmax
js
) κj1+κj χ
11+κjj
Localization of knowledge spillovers increasing in κj > 0
R&D spillovers generate growth in global knowledge capital
χj
χj=
S∑s=1
Mjs
∫ψλjs(ψ)lRjs (ψ)dG(ψ)
where Mjs denotes mass of firms and{λjs(ψ)
}determines the
strength of R&D spillovers
Thomas Sampson (LSE) Technology Gaps May 2019
Closing the model
Representative consumer in each country with unit intertemporalelasticity of substitution and discount rate ρ
Final consumption good is Cobb-Douglas aggregate acrossindustries with expenditure shares µj
Within industries output is differentiated by country of origin withArmington elasticity σ
Iceberg trade costs τjss to export from s to s
Free entry with initial (ψ, θ) drawn from joint distribution amongincumbent firms
Firms face exogenous exit shock at rate ζ
Details
Thomas Sampson (LSE) Technology Gaps May 2019
Balanced growth path
Firm opts for R&D rather than adoption if capability ψ exceeds
ψ∗js = ηγjBA
Bs
Threshold ψ∗ decreasing in R&D efficiency Bs, but increasing inadvantage of backwardness γj
Technology gaps (i.e. relative productivity levels) are stationaryand balance dispersion and concentration forces
Dispersion: R&D capability, R&D efficiency, localization ofknowledge spilloversConcentration: Advantage of backwardness, global knowledgespillovers
Productivity distribution in industry j shifts outwards at rate gj in allcountries and all firms in steady state with productivity growth rategj Details Firm steady state
Thomas Sampson (LSE) Technology Gaps May 2019
Balanced growth path
Firm opts for R&D rather than adoption if capability ψ exceeds
ψ∗js = ηγjBA
Bs
Threshold ψ∗ decreasing in R&D efficiency Bs, but increasing inadvantage of backwardness γj
Technology gaps (i.e. relative productivity levels) are stationaryand balance dispersion and concentration forces
Dispersion: R&D capability, R&D efficiency, localization ofknowledge spilloversConcentration: Advantage of backwardness, global knowledgespillovers
Productivity distribution in industry j shifts outwards at rate gj in allcountries and all firms in steady state with productivity growth rategj Details Firm steady state
Thomas Sampson (LSE) Technology Gaps May 2019
Equilibrium technology gaps
Average technology gap between countriesEθ∗ 1
1−β
js
Eθ∗ 1
1−β
j s
1−β
=
Bs
Bs
(Ψjs
Ψj s
) γj (1−β)
1+κj−α
1+κjγj
where Ψjs is average effective capability accounting for adoptionbeing equivalent to R&D with capability ψ∗js. Ψjs decreasing in Bs
Ψjs =
∫ ψmax
ψ∗js
ψ1
γj (1−β)−α dG(ψ) +(ψ∗js
) 1γj (1−β)−α G(ψ∗js)
Technology gap increasing in Bs and Ψjs, but Ψjs decreasing in Bs
Define innovation-dependence of industry j as the elasticity ofaverage technology gap to R&D efficiency
Thomas Sampson (LSE) Technology Gaps May 2019
Comparative advantage
Balanced growth path exports from s to s in industry j given by
log EXjss = υ1j + υ2
s − (σ − 1)(log ws + log τjss
)+ (σ − 1)
1 + κj
γj
[log Bs +
(γj(1− β)
1 + κj− α
)log Ψjs
]where υ1
j and υ2s are endogenous industry and destination terms
PropositionOn a balanced growth path, countries with higher R&D efficiency havea comparative advantage in more innovation-dependent industrieswhere the advantage of backwardness is low and the localization ofknowledge spillovers is high
Thomas Sampson (LSE) Technology Gaps May 2019
International income inequality
Income per capita is increasing in R&D efficiency and under freetrade
∂ log ws
∂ log Bs=σ − 1σ
J∑j=1
νjsInnovation-dependencejs
where∑J
j=1 νjs = 1
PropositionOn a balanced growth path with free trade, the elasticities of wages,income per capita and consumption per capita to R&D efficiency areweighted averages of industry innovation-dependence levels
Trade costs complicate analysis, but leave intuition unchangedDetails
Thomas Sampson (LSE) Technology Gaps May 2019
Quantifying technology gaps
Estimate model to obtain
Country R&D efficiency levels from within-industry variation in R&DintensityIndustry innovation-dependence levels from bilateral trade patterns
Calibrate model and quantify differences in wages and incomesdue to observed variation in R&D efficiency
Estimate under first order approximation that is valid when fewfirms perform R&D. 9.9% UK goods firms invested in R&D in2008-09
Approximation implies innovation-dependence given by
IDj =(1− β)κj
γj(1− β)− α
Thomas Sampson (LSE) Technology Gaps May 2019
R&D efficiency
Industry-level R&D intensity
RDjs =α(δ + gj)
ρ+ ζ + γj(δ + gj)
k[γj(1− β)− α
]k[γj(1− β)− α
]− 1
η−kγj
(Bs
BA
)k
Estimate bs = k log Bs from relative R&D intensity acrosscountries
bs =1
Ns
2014∑t=2010
20∑j=1
log(
RDjst
RDj st
),
RDjs is ratio of R&D expenditure to value-added from OECD datafor 25 countries and 20 manufacturing industries
Estimates
Thomas Sampson (LSE) Technology Gaps May 2019
Estimating innovation-dependence
Estimate IDjk from bilateral trade data
log(
EXjss
EXj ss
)− (σ − 1) log
(ws
ws
)= −(σ − 1)
IDj
kbs
− (σ − 1) log τjss + εjss
Parameterize trade costs as function of distance, border, commonlanguage, free trade agreement and exporter-industry fixed effect
Set σ − 1 = 6.53 from Costinot, Donaldson & Komunjer 2012Include controls for other determinants of
Comparative advantage: interaction of industry dummy variableswith physical capital abundance, human capital, rule of law,financial developmentProductivity: institutional quality, business environment
Use pooled trade data 2010-14 for 22 ISIC 2 digit goods industriesEstimates Model validation
Thomas Sampson (LSE) Technology Gaps May 2019
Calibration
Calibrate model for 25 OECD countries in 2012
22 goods industries plus non-tradable services. Assumeinnovation-dependence is zero in services
Solving for relative wages and incomes does not require fullcalibration, e.g. do not need to calibrate R&D spillovers λjs(·)
Calibrate: R&D intensity, population, industry expenditure shares,firm exit rate, profit share, trade elasticity, discount rate Details
Compare equilibria with and without R&D efficiency differences toobtain nominal wage and real income variation due to R&Defficiency
Thomas Sampson (LSE) Technology Gaps May 2019
Nominal wages
AUS
AUT BEL
CAN
CHL
CZE
DEUDNK
ESP
FIN
FRA
GBR
HUN
IRLITA
JPN
KOR
MEX
NLD NOR
POL
PRT
SVN
TUR
USA
-.8-.6
-.4-.2
0.2
Cal
ibra
ted
log
wag
e
-2 -1.5 -1 -.5 0 .5Log nominal wage
Elasticity of calibrated wages to observed wages equals 0.30
Ratio of standard deviations of calibrated and observed log wagesequals 0.36 Robustness
Thomas Sampson (LSE) Technology Gaps May 2019
Real incomes
AUS
AUTBEL
CAN
CHL
CZE
DEUDNK
ESP
FIN
FRA
GBR
HUN
IRLITA
JPN
KOR
MEX
NLD NOR
POL
PRT
SVN
TUR
USA-.3
-.2-.1
0.1
Cal
ibra
ted
log
GD
P p
er c
apita
-1 -.5 0 .5Log GDP per capita
Elasticity of calibrated to observed GDP per capita is 0.14 and standarddeviation ratio is 0.19
Thomas Sampson (LSE) Technology Gaps May 2019
Conclusions
Develop theory of how innovation and adoption determinetechnology gaps and income levels
Industries with lower advantage of backwardness and higherknowledge localization are more innovation-dependent
Estimate R&D efficiency and innovation-dependence from R&Dand trade data and use to calibrate model
Technology gaps account for around one-quarter to one-third ofnominal wage variation within the OECD
Thomas Sampson (LSE) Technology Gaps May 2019
Intertemporal preferences
Representative consumer in each country has intertemporalpreferences
U(t) =
∫ ∞t
e−ρ(τ−t) log c(τ)dτ
ρ > 0 discount rate
Budget constraint
a(t) = ιs(t)a(t) + ws(t)− zs(t)c(t)
a per capita assetsιs interest ratews wagezs price of final consumption
Thomas Sampson (LSE) Technology Gaps May 2019
Demand
Final consumption good is Cobb-Douglas aggregate acrossindustries with expenditure shares µj
Within industries output is differentiated by country of origin withArmington elasticity σ
Iceberg trade costs τjss from s to s
Export value from s to s in industry j given by
EXjss = τ−σjss
(pjs
Pj s
)1−σµjzscsLs
where pjs is price of industry j output produced in country s andPj s is industry price index
Thomas Sampson (LSE) Technology Gaps May 2019
Entry
Free entry determines mass of firms Mjs
Entrant must hire f E workers to create unit flow of new firms
Entrants draw initial (ψ, θ) from joint distribution Hjs(ψ, θ) ofcapability and productivity among incumbent firms
Firms face exogenous exit shock at rate ζ
Parameter restriction1
1− β> γj >
α
1− β+
κjγj
1 + κj
⇒ Firms’ intertemporal optimization problems concave
Set global consumption expenditure as the numeraireS∑
s=1
zscsLs = 1
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Balanced growth path properties
Let gj be the growth rate of global knowledge capital χj .Necessary conditions for a balanced growth path
Productivity distribution Hjs(θ) shifts outwards at rate gj in allcountriesKnowledge levels χR
js , χAjs grow at rate gj in all countries
Output prices decline at rate gj , nominal wages time invariantInterest rate ιs = ρ constant across countries and over time
Output and consumption per capita grow at rate q =∑J
j=1 µjgj in allcountries
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Firm steady state properties
1 Steady state relative productivity and R&D investment increasingin capability ψ if perform R&D
2 R&D intensity independent of firm size, decreasing in γj ,increasing in gj
ws lR∗js (ψ)
pjsy∗js(ψ)=
α(δ + gj)
ρ+ ζ + γj(δ + gj)
3 Adoption equivalent to R&D with capability ψ∗js4 Within-country firm size inequality increasing in α and β,
decreasing in γj
θ∗js(ψ′)
θ∗js(ψ)=
(ψ′
ψ
) 1−βγj (1−β)−α
, ψ′ ≥ ψ ≥ ψ∗js(ψ′
ψ∗js
) 1−βγj (1−β)−α
, ψ′ ≥ ψ∗js ≥ ψ
1, ψ∗js ≥ ψ′ ≥ ψ
Back
Thomas Sampson (LSE) Technology Gaps May 2019
R&D problem
Firm chooses R&D investment to maximize expected presentdiscounted profits net of R&D costs
maxφ,lR
∫ ∞t
e−(ρ+ζ)(t−t)ws
1− ββ
(βpjsχ
Rjs
ws
) 11−β
φ− lR
dt
subject to the R&D technology where φ ≡(θ/χR
js
) 11−β denotes
productivity relative to the R&D knowledge level
Firm’s optimization problem is a discounted infinite-horizon optimalcontrol problem with state variable φ and control variable lR
Solving the optimal control problem implies each firm has aunique steady state which is saddle-path stable
Thomas Sampson (LSE) Technology Gaps May 2019
Firm steady state & transition dynamics
lR
φ
φ = 0
lR = 0
lR∗js (ψ)
φR∗js (ψ)
Stable arm
Thomas Sampson (LSE) Technology Gaps May 2019
Steady state
Steady state relative productivity
φR∗js (ψ) =
αβ β1−β (Bsψ)
1α
(pjsχ
Rjs
ws
) 11−β (δ + gj)
α−1α
ρ+ ζ + γj(δ + gj)
α
γj (1−β)−α
Steady state R&D employment
lR∗js (ψ) =
(δ + gj)(φR∗
js (ψ))γj (1−β)
ψBs
1α
Thomas Sampson (LSE) Technology Gaps May 2019
Firm steady state under adoption
Adoption problem isomorphic to R&D problem
Adoption problem has unique steady state with
φA∗js =
αβ β1−β
(Bsψ
∗js
) 1α
(pjsχ
Rjs
ws
) 11−β (δ + gj)
α−1α
ρ+ ζ + γj(δ + gj)
α
γj (1−β)−α
where φ ≡(θ/χR
js
) 11−β
Adoption investment intensity equals R&D intensity of firms thatperform R&D
ws lA∗js
pjsy∗js=
α(δ + gj)
ρ+ ζ + γj(δ + gj)
Back
Thomas Sampson (LSE) Technology Gaps May 2019
General equilibrium
Balanced growth path wages ws, assets as and growth rates gjsatisfy
Ls =J∑
j=1
µj
ρ+ ζ
(ζ + βρ+
αρ(δ + gj
)ρ+ ζ + γj
(δ + gj
))Zjs
asLs =J∑
j=1
µj
ρ+ ζ
(1− β −
α(δ + gj
)ρ+ ζ + γj
(δ + gj
))wsZjs
Thomas Sampson (LSE) Technology Gaps May 2019
General equilibrium
gj =S∑
s=1
µjα(δ + gj
)ρ+ ζ + γj
(δ + gj
) Zjs
Ψjs
∫ ψmax
ψ∗js
λjs(ψ)ψ1
γj (1−β)−α dG(ψ)
where:
Zjs ≡S∑
s=1
τ1−σjss (ρas + ws) Lsw−σs
BsΨ
γj (1−β)
1+κj−α
js
(σ−1)(1+κj )
γj
∑Ss=1 τ
1−σj ss w1−σ
s
BsΨ
γj (1−β)
1+κj−α
j s
(σ−1)(1+κj )
γj
Back
Thomas Sampson (LSE) Technology Gaps May 2019
R&D efficiency
AUS
AUTBEL
CAN
CHL
CZE
DEU
DNK
ESP
FIN
FRA
GBR
HUN
IRLITA
JPN
KOR
MEX
NLD NOR
POL
PRT
SVN
TUR
USA
-3-2
-10
Log
R&D
effi
cien
cy
9.5 10 10.5 11Log GDP per capita
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Innovation-dependence
0103
1012
13
1415
16
17
18
19
20
21
22
2324
25
26
27
28
2930
3133
.2.3
.4.5
.6In
nova
tion-
depe
nden
ce
-5 -4 -3 -2 -1Log Share of firms that perform R&D
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Model validation
Undertake two validation exercises
1 Firm-level R&D investment Details
2 Out-of-sample test Details
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Firm-level R&D
Model implies
1FiRDj
= − 1αk log η
log ShRDjs −log(BA/Bs
)α log η
+ρ+ ζ
α(δ + gj)
FiRDj is R&D intensity conditional on performing R&DShRDjs is share of firms that perform R&D
Compute FiRDj and ShRDjs from UK Annual Business Survey andBusiness Expenditure on R&D survey 2008-09
Thomas Sampson (LSE) Technology Gaps May 2019
Validation test
0103
0508
1012
13
14
15
16
17
18
19
20
21
22
23
2425
26
27 28
29
303133
050
100
150
1 / (
R&D
inte
nsity
of f
irms
that
per
form
R&D
)
1 2 3 4 5Negative Log Share of firms that perform R&D
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Out-of-sample test
Use innovation-dependence estimates to perform out-of-sampletest of model’s comparative advantage predictions
Sample: 9 European countries with Eurostat R&D intensity data
Estimate
log(
EXjss
EXj ss
)−(σ−1) log
(ws
ws
)= ξCompAdvj s +Controlsjss +εjss,
where CompAdvj s = −(σ − 1)IDjk log bs and other controls are
same as baseline specification
Model predicts ξ = 1
Estimate ξ = 0.74 with standard error of 0.06
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Parameter values
Structural estimates: R&D efficiency, innovation-dependence,trade costs
UK firm-level surveys: R&D intensity conditional on performingR&D
OECD data: population, industry expenditure shares, exit rate
β = 0.85 implies 15% profit share (Gabler & Poschke 2013, Barkai2017)
σ − 1 = 6.53, ρ = 0.04
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Robustness checks
Baseline GDP per capita
Homogeneous innovation‐dependence
Low trade elasticity
Moderate trade
elasticity
High trade elasticity
Industry trade
elasticities(a) (b) (c) (d) (e) (f) (g)
Elasticity 0.300 0.279 0.300 0.305 0.310 0.283 0.238Standard deviation ratio 0.363 0.331 0.353 0.366 0.369 0.345 0.304
Elasticity 0.144 0.125 0.156 0.123 0.135 0.141 0.141Standard deviation ratio 0.195 0.167 0.208 0.167 0.182 0.192 0.197
Average 0.316 0.277 0.295 0.340 0.324 0.306 0.379Standard deviation 0.113 0.123 0.348 0.184 0.091 0.258Correlation with baseline 0.924 0.848 0.945 0.974 0.548
(iv) Trade elasticity 6.53 6.53 6.53 2.5 4.5 8.5 IndustryRow (i) gives the elasticity of calibrated log wages to observed log nominal wages, and the ratio of the standard deviation of calibrated log wages to the standard deviation of observed log nominal wages. Row (ii) gives the same statistics for real income per capita. Row (iii) reports summary statistics on the innovation‐dependence estimates for goods industries used in the calibration. Correlation with baseline is the correlation with the baseline estimates used in column (a). Row (iv) shows the trade elasticity used in the calibration. Column (a) uses estimates from Table 1, column (c) to calibrate the model. For column (b) innovation‐dependence is estimated including the interaction of industry dummy variables with the importer's log GDP per capita as an additional control. In column (c) innovation‐dependence is restricted to be the same in all goods industries. Industry‐specific trade elasticities used for column (g) from Caliendo and Parro (2015).
(i) Nominal wage
(ii) Real income per capita
(iii) Innovation‐dependence
Table 2: Calibration results
Back
Thomas Sampson (LSE) Technology Gaps May 2019
Income & comparative advantage
-1.5
-1-.5
0.5
120
00 in
com
e el
astic
ity
-2 -1 0 1 21962 income elasticity
Estimate sectoral income elasticity βj of exports for 2 digit SITC sectors from
EXjs = αs + δj + βj GDPPCs + εjs
Thomas Sampson (LSE) Technology Gaps May 2019
Industrial development & comparative advantage
05
1019
62-2
000
expo
rt gr
owth
-2 -1 0 1 21962 income elasticity
China
-6-4
-20
24
1962
-200
0 ex
port
grow
th
-2 -1 0 1 21962 income elasticity
Japan
-50
510
1519
62-2
000
expo
rt gr
owth
-2 -1 0 1 21962 income elasticity
Korea-5
05
1019
62-2
000
expo
rt gr
owth
-2 -1 0 1 21962 income elasticity
Singapore-5
05
1019
62-2
000
expo
rt gr
owth
-2 -1 0 1 21962 income elasticity
Taiwan
-50
510
1962
-200
0 ex
port
grow
th
-2 -1 0 1 21962 income elasticity
Thailand
Thomas Sampson (LSE) Technology Gaps May 2019
Income elasticities & industry characteristics
1516
17
1819
20
2122
23
24
2526
27
2829 30
31
3233
34
35
36
-1-.5
0.5
Inco
me
elas
ticity
-7 -6 -5 -4 -3R&D intensity
1516
17
1819
20
2122
23
24
2526
27
2829 30
31
3233
34
35
36
-1-.5
0.5
Inco
me
elas
ticity
10 11 12 13Capital intensity
1516
17
1819
20
2122
23
24
2526
27
2829 30
31
3233
34
35
36
-1-.5
0.5
Inco
me
elas
ticity
.2 .3 .4 .5 .6Skill intensity
Thomas Sampson (LSE) Technology Gaps May 2019