Pag.1

1
Synthetic indicators/1
2
Synthetic indicators/2
3
PCT per million in 30 best performing regions,
1998-2000
4
PCT per million in 30 best performing regions,
2002-2004
5
OECD Regions PCT per million population
variability, 1998-2000
6
OECD Regions PCT per million population
variability, 2002-2004
7
Spatial distribution of innovation/1

• The degree of disparities in the regional
  distribution of innovative activities has
  increased across OECD countries for three out of
  four indexes. CV decreases mainly because the
  average value has changed
• This phenomenon has not been homogeneous across
  macroareas (in particular it decreases in the
  United States)
• We would like to perform the same analysis across
  sectors to assess potential differences

8
Spatial distribution of innovative activity/1
9
Spatial dependence of innovative activity/2

• Presence of strong and positive spatial
  autocorrelation among contiguous areas. Spatial
  dependence extends until the 3th order of
  contiguity
• The extent of such a dependence is stable along
  time
• Spatial dependence is also detected when
  distances are used instead of contiguities
• This process has favoured the formation of
  clusters of innovative regions(we need sector
  data in order to see if such a process is
  differentiated across sectors and how much)
• Let us see these clusters

10
Moran scatterplot map, 2002-2004
11
Moran scatterplot map Europe, 2002-2004
12
Moran LISA map, 2002-2004
13
Moran LISA map Europe, 2002-2004
14
Convergence in innnovative efforts?National level
15
Convergence in innnovative efforts?Regional level
16
Summary of main novelties

• We focus on OECD regions.
• We have a set of homogeneous indicators for all
  the countries.
• We are going to estimate KPF at both the regional
  level (and later potentially at the industry
  level)
• We are going to use specific econometric
  techniques to analyse the nature and the spatial
  scope of knowledge creation and diffusion.

17
The determinants of innovative activity at the
local level knowledge production function

• I local patents (per capita) in region j
• RD quota of RD on GDP (j)
• HK tertiary education (j)
• DENS population density (j)
• NAT national dummies
• DU, DR, DCAP dummies for urban, rural, capital
  regions
• DGDP dummy for above and below average GDP per
  capita

• Note
• Variables in log
• Time lags are considered

18
Estimation strategy

• OLS to assess significance of coefficients and
  the presence of spatial dependence
• Discriminate between spatial lag model or spatial
  error model and re-estimate with ML

19
Econometric results
20
Some robustness checks

• Interactive dummies
• DGDPHK and DGDPRD
• Spatial Lag of RD
• KPF with distance matrix (only for EU and North
  America)
• KPF including Japan and Korea (estimation of some
  variables)
• KPF with PCT per worker (instead of per capita)

21
KPF estimation with interactive dummies
22
KPF estimation with spatial lag of RD
23
KPF estimation with distance matrix
24
KPF estimation with Japan and Korea
25
KPF estimation with PCT per worker
26
Final remarks

• Clusters of regional innovative systems have
  formed across OECD countries
• Main determinants of knowledge creation are at
  work both at the local and at the external level
• Human capital has larger effects than RD
• Such determinants are within national innovation
  systems

27
Final remarks and questions

• Clusters of regional innovative systems have
  formed across OECD countries
• Main determinants of knowledge creation are at
  work both at the local and at the external level
• Are they different with respect to industrial
  specialisation?
• Are they within national innovation systems?
• Are they getting stronger or bigger?

28
The research agenda forwhat we have done so far

• There are still some missing values in the
  database (Korea and Switzerland, for example)
• No detail about RD
• Public vs private (possible for some countries)
• Not all spatial externalities are appropriately
  measured
• Citations can be used to measure spillovers both
  within and across regions
• No measure of other local public knowledge
• University and research centers?

29
The research agenda main options

• To deepen and to improve the analysis of the
  general KPF in order to assess the presence of
  differences across macroregions
• To replicate the descriptive analysis at a more
  disaggregated territorial level (that is TL3)the
  replication of the econometric analysis is
  problematic since most data for explanatory
  variables are lacking
• To focus on industrial disaggregation and to
  replicate the analysis for all sectors or for a
  set of them (some high tech). This can be done
  both for the descriptive and the econometric
  analysis. The database has to be built at the
  regional level

30
The determinants of innovative activity at the
local industry level

• Note
• Variables in log
• Time lags are considered

• I local industry patents (per capita) in sector
  i and region j
• IST technological specialisation index based on
  location quotient (ij)
• DIV diversity index based on herfhindhal (ij)
• GDP GDP per capita (j)
• DENS population density (j)
• EDU tertiary education
• RD quota of RD on GDP (j)
• NAT national dummies
• Other controls for macroareas, urban and rural
  regions, citations

31
For your interests

• Oecd patent database includes also data on
  citations regionalised for TL2 regions
• If you are interested in this topic and getting
  hold on the data you can contact me
• stefanousai_at_unica.it