Does Demographic Aging Contribute to the Innovation Divide

Transcription

Does Demographic Aging Contribute to the Innovation Divide
Does Demographic Aging Contribute to the
Innovation Divide Across German Labor Markets?
Melanie Arntz
ZEW Mannheim and University of Heidelberg
Terry Gregory
ZEW Mannheim
8th European workshop on ”Labour Markets and Demographic Change” ,
Vienna 12/9/2013
Regional projections of average population age, 2010-2030
Figure : 2010
Figure : 2030
Source: Federal Institute for Research on Building, Urban Affairs and Spatial Development (BBSR)
Aging may affect innovative capabilities . . .
. . . at the individual level
I
declining cognitive abilities after some peak in mid-ages
(Levin and Stephan 1991; Rouvinen 2002)
I
a turnaround point between 50 and 55 for inventors in
longitudinal studies (Mariani and Romanelli 2007; Schettino
et al. 2008)
. . . at the firm level
I
a mostly hump-shaped link between age and
productivity/innovations (e.g. Schneider 2008, Verworn and
Hipp 2009)
I
positive effect of age diversity (Grund and Westergard-Nielsen
2008) at least for creative tasks (Backes-Gellner and Veen
2012)
⇒ firm performance may be more than the sum of individual
performances
. . . at the macro level
I
performance may diverge from sum of individual and firm
performances due to knowledge externalities
I
I
I
I
I
younger workers: ability to generate and recombine new
knowledge
older workers: ability to use skills, knowledge and experience
different skill endowments may complement each other and
create positive spillovers
interactions may take place inside and outside the workplace
(e.g. social interaction)
innovations are considered to be a regional outcome
⇒ Total effect of aging should be analysed at the macro/regional
level
Macro-level evidence on age-performance link
Study and country
Performance indicator
Age indicators
Age-performance
pattern
OLS
FE
IV
Growth rate of
real GDP per capita
Growth rate of
real GDP per capita
Total factor productivity
Population age shares
(15+ years)
Population age shares
(15+ years)
Working-age population
age shares (10-70 years)
T
T
T
T
T
T
T
T
T
Growth rate of
real GDP per capita
New firm formation rate
Population age shares
(15-75 years)
Population age shares
(20-64 years)
Workforce age shares
T
Country-level studies:
Lindh/ Malmberg 1999
(OECD)
Prskawetz et al. 2007a
(EU-25)
Feyrer 2008
(worldwide)
Regional-level studies:
Brunow/ Hirte 2006
(Germany)
Bönte et al. 2009
(West Germany)
Frosch 2008
(Europe)
Tang/ MacLeod 2006
(Canada)
Invention rates
Real GDP per worker
T
two age groups: younger than 55,
and older than 55 years
= hump-shaped, ↑ = positive effect, ↓= negative effect
T
T
T
↓
↑
Objective and contribution of this work
1. estimate knowledge production function and age-innovation
profile for German regions
2. conduct a cross-sectional analysis and contrast this approach
to a panel estimation that accounts for unobserved
heterogeneity
3. address potential endogeneity by instrumenting the regional
workforce age structure by lagged population age data
4. test for interaction effects between age groups
5. examine to what extent demographic trends contribute to
East-West innovation divide
Data on regional in- and output measures (1994-2008)
Innovation output
I
number of published patents from European Patent Office
(EPO)
I
focus on patent inventor (fractional accounting approach)
I
calculate number of regional patents
I
innovation count data provide a fairly reliable measure of
innovative activity on the regional level (Acs et al. 2002, RP)
R&D input measures
I
private R&D expenditures (German Stifterverband)
I
public R&D expenditures, i.e regular and external funding
(DeStatis)
Data on regional workforce (1994-2008)
Use Sample of Integrated Labour Market Biographies (SIAB) to
calculate
I
age structure
I
skill composition
I
share of creative professionals (Florida 2002)
I
workforce size
at the level of 141 labour market regions (Kosfeld and Werner
2012)
Summary statistics for labour market regions, 1994-2008
Variable
(1)
All
regions
(2)
Lowest
innovative
regions
(3)
Highest
innovative
regions
(4)
Differential
between
(3) and (2)
Data
Source*
number of patents per worker
private RaD expenditures (in 1000 Euro)
public RaD expenditures (in 1000 Euro)
average workforce age
workforce age dispersion
share of workers aged < 36 (in %)
share of workers aged 36-49 (in %)
share of workers aged 50+ (in %)
share of creative professionals (in %)
share of high-skilled workers (in %)
share of medium-skilled workers (in %)
share of low-skilled workers (in %)
workforce size (in 1000)
population density (population per 100 km2)
33.87
243.92
135.19
40.28
10.35
18.08
59.99
21.92
5.12
5.72
82.73
11.55
4.10
443.11
5.20
19.97
38.04
40.91
10.18
16.08
60.71
23.21
3.69
6.01
88.01
5.97
2.61
226.30
80.81
624.78
217.87
40.17
10.47
18.54
59.46
22.01
6.26
6.59
78.08
15.33
4.96
630.83
75.61
604.81
179.83
-0.74
0.29
2.46
-1.25
-1.21
2.56
0.58
-9.94
9.36
2.35
404.53
EPO
GST
DeStatis
SIAB
SIAB
SIAB
SIAB
SIAB
SIAB
SIAB
SIAB
SIAB
SIAB
DeStatis
* EPO: European Patent Office, SIAB: Sample of Integrated Labour Market Biographies released by
German Federal Employment Agency, DeStatis: Regional database released by Federal Statistical Office,
GST: German Stifterverband (Innovation Agency for the German science system)
The demographic and the innovation divide
Average workforce age
Number of patents per worker
Kiel
Kiel
Lübeck
Bremershaven
Hamburg
Lübeck
Bremershaven
Mecklenburgi
Schwerin
Bremen
Hamburg
Hannover
Wolfsburg
Hannover
Berlin
Osnabrück
Wolfsburg
Berlin
Osnabrück
Potsdam-Mitt
Potsdam-Mitt
Magdeburg
Münster
Mecklenburgi
Schwerin
Bremen
Magdeburg
Münster
Bielefeld
Bielefeld
Cottbus
Dortmund
Cottbus
Dortmund
Göttingen
Düsseldorf
Düsseldorf
Kassel
Kassel
Dresden
Erfurt
Dresden
Erfurt
Jena
Aachen
Jena
Aachen
Chemnitz
Köln
Chemnitz
Köln
Fulda
Bonn
Göttingen
Fulda
Bonn
Koblenz
Koblenz
Frankfurt am
Frankfurt am
Bayreuth
Bayreuth
Mainz
Mainz
Darmstadt
Darmstadt
Würzburg
Heidelberg
Saarbrücken
Würzburg
Heidelberg
Saarbrücken
Nürnberg
Nürnberg
Regensburg
Regensburg
Karlsruhe
Karlsruhe
Stuttgart
Stuttgart
Ingolstadt
Ulm
Augsburg
Ingolstadt
Augsburg
Ulm
Patents per
Average age
100 worker
München
Freiburg
Konstanz
Kempten
München
(41.0,41.6]
(40.5,41.0]
(40.1,40.5]
(39.7,40.1]
[37.7,39.7]
Freiburg
Konstanz
Kempten
Source: Calculations based on Sample of Integrated Labour Market Biographies (SIAB)
ρage,patents = −0.34
(53,218]
(35,53]
(18,35]
(9,18]
[1,9]
Related trends in aging and innovation intensities
∆ age between 1995 and 2008
∆ patents between 1995 and 2008
Kiel
Kiel
Lübeck
Bremershaven
Hamburg
Lübeck
Bremershaven
Mecklenburgi
Schwerin
Bremen
Hamburg
Emsland
Emsland
Hannover
Wolfsburg
Hannover
Berlin
Osnabrück
Wolfsburg
Berlin
Osnabrück
Potsdam-Mitt
Potsdam-Mitt
Magdeburg
Münster
Borken
Mecklenburgi
Schwerin
Bremen
Magdeburg
Borken
Bielefeld
Münster
Bielefeld
Cottbus
Dortmund
Cottbus
Dortmund
Göttingen
Düsseldorf
Wuppertal
Kassel
Düsseldorf
Wuppertal
Dresden
Oberhavel
Kassel
Dresden
Oberhavel
Erfurt
Erfurt
Jena
Aachen
Altenkirchen
Chemnitz
Köln
Fulda
Bonn
Jena
Aachen
Chemnitz
Köln
Göttingen
Altenkirchen
Fulda
Bonn
Koblenz
Koblenz
Frankfurt am
Frankfurt am
Bayreuth
Mainz
Trier
Bayreuth
Aschaffenbur
Darmstadt
Würzburg
Trier
Mainz
Aschaffenbur
Darmstadt
Würzburg
Erlangen
Erlangen
Heidelberg
Saarbrücken
Heidelberg
Saarbrücken
Nürnberg
Nürnberg
Regensburg
Regensburg
Karlsruhe
Karlsruhe
Stuttgart
Stuttgart
Ingolstadt
Ingolstadt
Passau
Ulm
Augsburg
Passau
Change in
Augsburg
Ulm
Change in Patents
average age
per worker
München
Freiburg
Konstanz
Kempten
München
(3.2,4.3]
(2.9,3.2]
(2.5,2.9]
(2.2,2.5]
[1.0,2.2]
Freiburg
Konstanz
Kempten
Source: Calculations based on Sample of Integrated Labour Market Biographies (SIAB)
ρ∆age,∆patents = −0.32
(28.3,103.6]
(15.3,28.3]
(10.2,15.3]
(5.5,10.2]
[-2.4,5.5]
number of patents per worker
40
80 100 120
0
20
60
Scatterplot between workforce age and innovation
(cross-section, 1995-2008)
38.5
39
39.5
40
40.5
average workforce age
95% CI
East German regions
41
41.5
Fitted values
West German regions
Estimating the regional knowledge production function
ln Pit
=
α + β ln RDit + γ1 MAGEit + γ2 MAGEit2 + γ3 PROFit
+δXit + ci + τ + it
Pit - innovation output in region i and (3yr-)period t
RDit - public and private R&D investments
MAGEit - mean workforce age
PROFit - share of creative professionals
Xit - controls including dummies for industries (16 categories), population
density (plus squared term), size of local workforce (log)
ci - time-constant unobserved effect
τ - time fixed effects
Choice of instruments
I
I
age may be endogeneous due to age-selective migration and
hiring
use lagged population demographics as instruments
I
I
should address migration and reverse causality if lags
sufficiently long
should address potentially age-selective hiring since
instrumenting the supply of workers
Estimation results for Germany (cross-section, 1994-2008)
Dependent variable: ln Pit
(1)
OLS
(2)
OLS
(3)
OLS
(4)
IV GMM
0.57***
(11.51)
0.05**
(2.37)
0.48***
(9.17)
0.04**
(2.07)
0.41***
(8.74)
0.05**
(2.29)
0.35***
(5.33)
0.08**
(2.42)
18.20***
(3.08)
-0.23***
(-3.09)
0.16*
(1.90)
20.95***
(3.75)
-0.26***
(-3.78)
0.37
(1.43)
20.75***
(4.20)
-0.26***
(-4.23)
0.10
(0.35)
70.20***
(2.68)
-0.88***
(-2.69)
0.09
(0.28)
-0.58***
(-3.59)
-360.29***
(-3.05)
-0.19
(-1.14)
0.40***
(4.46)
-0.28
(-1.13)
-412.45***
(-3.71)
-0.07
(-0.34)
0.34***
(3.77)
0.08
(0.30)
-416.72***
(-4.27)
0.17
(0.58)
0.46***
(4.12)
0.09
(0.28)
-1398.89***
(-2.67)
no
141
0.910
229.4
no
141
0.926
232.8
yes
141
0.946
131.7
yes
141
0.894
89.4
0.812
0.666
R&D inputs
private RaD exp. (log, in 100 tsd Euro)
public RaD exp. (log, in 100 tsd Euro)
Human capital inputs
average workforce age
average workforce age (squared)
num. of creative professionals (in log)
Regional indicators
dummy for East Germany
population density (log, in tsd)
workforce size (log, in tsd)
constant
With industry dummies?
N
R2
F
hansen (j statistic)
Hansen (p-value)
Note: t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01. Standard errors are
clustered by region.
Regional workforce age-innovation profile (cross-section)
80
West
Germany
East
Germany
number of patents
40
60
37 patents
20
1.2 years
37.8
38.3
38.8
39.3
39.8
40.3
average workforce age
40.8
41.3
41.8
→ turnaround point at 39.5 years
→ 37/144*100= 26 percent of East-West innovation divide can be explained
by differences in the average workforce age (1.2 years)
Estimation results for Germany (panel, 1994-2008)
Dependent variable: ln Pit
(1)
POLS
(2)
POLS
(3)
FE
(4)
FE IV GMM
0.33***
(9.48)
0.06**
(2.29)
0.32***
(9.32)
0.06***
(2.66)
0.07***
(3.34)
-0.07**
(-2.60)
0.05**
(2.10)
-0.09***
(-3.43)
1.02
(0.96)
-0.01
(-0.96)
0.21
(0.87)
1.37
(1.35)
-0.02
(-1.48)
0.28
(1.17)
1.11
(1.31)
-0.01
(-1.38)
0.17
(0.89)
9.53***
(5.29)
-0.12***
(-5.46)
-0.14
(-0.63)
0.20**
(2.30)
0.01
(0.04)
2.04***
(2.95)
-0.95*
(-1.80)
0.15***
(3.67)
-18.07
(-1.09)
-0.10
(-0.12)
-2.34***
(-4.03)
0.37***
(4.44)
-28.76
(-1.42)
0.28***
(3.12)
-0.03
(-0.13)
0.15***
(3.43)
-33.88*
(-1.74)
no
705
0.909
109.6
yes
705
0.913
127.8
yes
705
0.685
43.7
yes
705
0.575
33.6
1.840
0.399
R&D inputs
private RaD exp. (log, in 100 tsd Euro)
public RaD exp. (log, in 100 tsd Euro)
Human capital inputs
average workforce age
average workforce age (squared)
num. of creative professionals (in log)
Regional indicators
population density (log, in tsd)
workforce size (log, in tsd)
time trend
constant
With industry dummies?
N
R2
F
hansen (j statistic)
Hansen (p-value)
Note: t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01. Standard errors are
clustered by region.
Regional workforce age-innovation profile (panel)
East
Germany
29 patents
1.2 years
0
20
number of patents
40
60
80
West
Germany
36.5 37 37.5 38 38.5 39 39.5 40 40.5 41 41.5 42 42.5 43
average workforce age
→ turnaround point at 38.6 years
→ 29/144*100= 20 percent of East-West innovation divide can be explained
by differences in the average workforce age (1.2 years)
Alternative age specifications (cross-section)
age group shares
(1)
(2)
(3)
age group size
(4)
R&D inputs (coef. for R&D exp. suppressed)
share of workers aged < 36 (log)
share of workers aged 50+ (log)
-3.18
(-1.49)
-6.42***
(-3.05)
-111.07**
(-2.48)
-91.27
(-1.25)
num. of workers aged < 36 (log)
num. of workers aged 36-49 (log)
num. of workers aged 50+ (log)
num. of creative professionals (log)
0.61
(1.40)
-0.02
(-0.06)
-4.13
(-1.44)
5.74**
(2.16)
-5.55***
(-2.92)
3.90*
(1.66)
-11.86
(-1.18)
-10.72
(-0.22)
14.78
(0.31)
1.12
(0.95)
Interaction terms
young x old
35.29**
-1.70
(2.25)
(-0.21)
0.38
3.02
(0.37)
(0.39)
midage x old
-2.19
-0.91
(-0.39)
(-0.64)
Regional indicators (coef. for pop. density and E-W dummy suppressed)
young x midage
workforce size (log, in tsd)
constant
N
R2
F
hansen (j statistic)
hansen (p-value)
-0.49
(-1.14)
13.67
(1.52)
5.89
(0.30)
317.00**
(2.36)
-16.32**
(-2.37)
11.04
(0.23)
141
0.902
71.9
5.730
0.220
141
0.912
86.2
3.942
0.414
141
0.815
33.7
0.759
0.384
141
0.903
42.3
5.242
0.022
Note: t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01. Standard errors are
clustered by region. All specifications are estimated with FE IV GMM and include industry
dummies.
Conclusion
I
evidence in favor of hump-shaped age-innovation profile
I
aggregate impact of demographic aging may be negative,
given the German age structure
I
East-West age divide largely explains East-West innovation
divide
I
positive interaction effect between youngest and oldest
workers hints at knowledge spillovers
Future work
I
use flexible production function to allow for age
complementarities (translog)
I
estimate substitution elasticities between age group shares
I
simulate regional innovation outcomes based on demographic
projections
Thank you for your attention!
0
number of patents per worker
80
40
60
100
20
120
Scatterplot II
10
10.2
10.4
10.6
age dispersion of workforce
95% CI
East German regions
10.8
Fitted values
West German regions
The regional production of knowledge
Regional knowledge production function by Griliches (1979)
Pi
Pi
= αRDiβ × HKiγ
- innovation output in region i
RDi - public and private R&D investments (Jaffe 1989, AER)
HKi - human capital base proxied by
- share of high-skilled (Feldman 1999, EINT, Malecki 1997)
- share of creative professionals (Florida 2002)
- workforce age structure
α
- constant shift-factor
Definition of the creative sector
Definition: group of technological employees characterized as improving technology in
the line of business they pursue (Wedemeier, 2012, Murghy et al. 1991)
Occupational title
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Mechanical and vehicle engineers.
Electrical engineers.
Architects and construction engineers.
Surveyors, mining, metallurgists and related engineers.
Miscellaneous engineers.
Chemists, physicists, chemical/physical engineers.
mathematicians, and civil engineering technicians.
Mechanical engineering technicians.
Electrical engineers technicians.
Surveyors, chemical, physical, mining, metallurgists, and miscellaneous engineering technicians.
Miscellaneous technicians.
Biological/mathematical/physical-technical assistant, chemical and related laboratory technician workers.
Draft persons.
Computer related professions.
Statisticians, humanists, natural scientists, and pastors.
Firm-level evidence on age-innovation profile
Study and country
Performance indicator
Age indicators
Age-performance
pattern
FE
IV
OLS
Firm-level studies:
Backes-Gellner & Veen 2012
Productivity
Germany
Crépon et al. 2003
Productivity/ Output
France
Daveri & Maliranta 2007
Productivity in electronics
Finland
Frosch, Göbel & Zwick 2011
New products or services
Germany
Göbel & Zwick 2009
Productivity
Germany
Grund & Westergard-Nielsen 2008
Value added per employee
Denmark
Ilmakunnas & Maliranta 2007
Productivity growth in ICT
Finland
Malmberg et al. 2008
Value added per employee
Schweden
Meyer 2009
Adoption of new technologies
Germany
Schneider 2008
New products or services
Germany
Verworn & Hipp 2009
New products or services
Germany
T
= hump-shaped, → = positive effect,
Workforce mean age,
age diversity
Workforce age shares
T
T
Workforce mean
tenure and tenure 2
Workforce age shares
→
T
↓
→
Workforce age shares
T
Workforce mean age,
standard deviation
Workforce age shares
T
Workforce Age shares
T
Workforce age shares
↓
Workforce mean age
and mean age 2
Share of older
workers in workforce
T
↓= negative effect
→
→
→
T
→
T
→
→
T
→
→
T
→
Instruments in first stages, 1994-2008
Dependent variable:
lagged pop. share, 15-18 (in log)
lagged pop. share, 18-20 (in log)
lagged pop. share, 30-35 (in log)
lagged pop. share, 40-45 (in log)
(1)
(2)
cross-section
mwfage
mwfagesq
455.33*** 36404.38***
(2.72)
(2.72)
-597.04*** -47691.64***
(-2.81)
(-2.80)
-16.12**
-1268.84**
(-2.24)
(-2.21)
43.25***
3452.50***
(3.85)
(3.88)
lagged pop. group size, 10-15 (in log)
lagged pop. group size, 15-18 (in log)
lagged pop. group size, 20-25 (in log)
lagged pop. group size, 35-40 (in log)
constant
N
R2
F
F-Test of excluded insturments
45.41***
(12.93)
2031.31***
(7.32)
141
0.800
19.2
16.04
141
0.804
19.8
16.31
(3)
(4)
panel
mwfage mwfagesq
-1.17***
(-3.33)
0.31**
(2.43)
-1.14***
(-5.24)
2.16***
(6.94)
-84.72***
(-3.08)
28.68***
(2.86)
-89.10***
(-5.09)
162.40***
(6.62)
705
0.947
392.6
22.24
705
0.949
410.8
22.01
Estimation results for West-Ger (cross-section, 1994-2008)
Dependent variable: ln Pit
(1)
OLS
(2)
OLS
(3)
OLS
(4)
IV GMM
0.53***
(9.11)
0.04
(1.47)
0.46***
(7.67)
0.03
(1.25)
0.40***
(8.16)
0.03
(1.21)
0.40***
(10.02)
0.02
(1.15)
16.00**
(2.40)
-0.20**
(-2.40)
0.23**
(2.48)
19.66***
(3.24)
-0.25***
(-3.26)
0.45
(1.55)
19.72***
(3.27)
-0.25***
(-3.28)
-0.08
(-0.25)
27.73**
(1.98)
-0.35**
(-2.00)
-0.12
(-0.43)
-317.23**
(-2.39)
0.41***
(4.29)
-0.31
(-1.12)
-387.32***
(-3.21)
0.33***
(2.99)
0.33
(1.09)
-395.00***
(-3.31)
0.38***
(3.83)
0.37
(1.41)
-551.60**
(-1.98)
no
108
0.886
161.7
no
108
0.910
162.9
yes
108
0.939
111.7
yes
108
0.936
100.0
3.011
0.390
R&D inputs
private RaD exp. (log, in 100 tsd Euro)
public RaD exp. (log, in 100 tsd Euro)
Human capital inputs
average workforce age
average workforce age (squared)
num. of creative professionals (in log)
Regional indicators
population density (log, in tsd)
workforce size (log, in tsd)
constant
With industry dummies?
N
R2
F
hansen (j statistic)
Hansen (p-value)
Note: t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01. Standard errors are
clustered by region.
Estimation results for West Germany (panel, 1994-2008)
Dependent variable: ln Pit
(1)
POLS
(2)
POLS
(3)
FE
(4)
FE IV GMM
0.32***
(8.03)
0.03
(1.16)
0.32***
(7.87)
0.04
(1.34)
0.08***
(3.18)
-0.06**
(-2.05)
0.07***
(2.74)
-0.09***
(-3.63)
2.50**
(2.26)
-0.03**
(-2.29)
0.05
(0.21)
3.06***
(2.97)
-0.04***
(-3.10)
0.12
(0.48)
1.77*
(1.86)
-0.02**
(-1.99)
0.04
(0.18)
7.70***
(4.42)
-0.10***
(-4.69)
-0.07
(-0.32)
0.19**
(2.01)
0.24
(0.92)
-0.05
(-0.06)
-0.32
(-0.46)
0.17***
(3.49)
-34.59*
(-1.97)
-3.48***
(-3.03)
-1.38*
(-1.81)
0.54***
(5.15)
-53.28**
(-2.47)
0.26**
(2.47)
0.18
(0.70)
0.09**
(2.00)
-64.09***
(-3.17)
no
540
0.904
96.9
yes
540
0.906
108.0
yes
540
0.650
40.6
yes
540
0.511
21.4
1.430
0.489
R&D inputs
private RaD exp. (log, in 100 tsd Euro)
public RaD exp. (log, in 100 tsd Euro)
Human capital inputs
average workforce age
average workforce age (squared)
num. of creative professionals (in log)
Regional indicators
population density (log, in tsd)
workforce size (log, in tsd)
time trend
constant
With industry dummies?
N
R2
F
hansen (j statistic)
Hansen (p-value)
Note: t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01. Standard errors are
clustered by region.
50
60
70
80
90
differential in patents per worker
differential in workforce age group shares
-4
-2
0
2
4
100
Innovation gap and age shifts between most and least
innovative regions
1995
1998
35<age<50
age<=36
2001
year
2004
2007
age>=50
patents per worker
.4
50
60
70
80
90
differential in patents per worker
differential in average workforce age *(-1)
.8
1
1.2
.6
100
Innovation gap and age shifts between most and least
innovative regions, by age group
1995
1998
2001
year
average workforce age
2004
2007
patents per worker
.1
Kernel densities of average regional innovation
0
.02
density
.04
.06
.08
East 1994-1996
West 1994-1996
East 2006-2008
West 2006-2008
0
100
200
average regional number of patents
300
.8
Kernel densities of average regional innovation
0
.2
density
.4
.6
East 1994-1996
West 1994-1996
East 2006-2008
West 2006-2008
36
38
40
42
average regional workforce age
44