Content uploaded by Fergal A O'Connor
Author content
All content in this area was uploaded by Fergal A O'Connor on Dec 04, 2017
Content may be subject to copyright.
Electronic copy available at: http://ssrn.com/abstract=2411168
Innovation Capacities in Advanced Economies: Relative Performance of Small
Open Economies
Eleanor Doyle and Fergal O’Connor
Abstract
This paper offers an empirical examination of the determinants of a nation’s ability to
produce commercially viable innovations, measured as patents, across a sample of twenty
three advanced economies. The approach employed is based on estimating National
Innovation Capacity that focuses on the long-run ability of economies to produce and/or
commercialize innovative technologies, in the spirit of Furman, Porter and Stern, (2002).
Motivated by differences in the rate of innovation between economies with different
economic structures we examine the Small Open Economies chosen from this country sample
to assess whether there is a significant difference between the determinants of Innovative
Capacity in Small Open Economies and the other developed economies. A number of
alternative specifications are estimated.
We find that advanced Small Open Economies and larger economies do not differ
substantially in their determinants of patenting activities and, notwithstanding the limitations
of patents as measures of innovative activity, we conclude that policy choice and variation
plays a key role in determining the productivity of R&D, when measured as patenting
activity.
Electronic copy available at: http://ssrn.com/abstract=2411168
1
1 Introduction
Innovative capacity lies at the heart of factors affecting future competitiveness particularly
for advanced modern economies since, under a Solow-type growth framework, such
economies are likely to have exhausted their ability to generate increased output from further
investments in capital. According to Furman et al. (2002: 899) an economy’s innovative
capacity represents
the ability ... to produce and commercialize a flow of innovative technology over the
long term.
Many studies (e.g. Gans and Stern (2003) and Gans and Hayes (2008)) have followed such an
approach and have found evidence to support the contention that the intensity to which
countries innovate varies based on a set of variables relating to:
each nation’s Common Innovative Infrastructure (CII);
its Cluster Specific Environment (CSE); and
the Quality of Linkages between both its CIE and CSE.
This approach facilitates the identification of a set of economic factors that drive patenting
activity/intensity and also allows for a policy-centred focus on how best to consider the long-
term choices that impinge on innovation capacity. This policy-centred focus applies as easily
to business development policy, on the one hand, as to business strategy on the other, given
the microeconomic basis of the cluster concept.
Empirically, the variation in the ability of countries to produce new-to-world technologies,
that are patented, is striking. Some countries consistently outperform others by a large
margin. For example, Canada, the US, Finland, Switzerland and Japan produce well over 100
patents per year (per million of population in 2008), while most advanced economies average
approximately 60 patents per million and still others such as Spain, Portugal, New Zealand
and Italy all may be considered to ‘underperform’ with less than 25 patents per million.
2
Source: IMF WEO, USPTO
This variation in patent outcomes is not explained by larger economies performing better, or
smaller ‘nimbler’ economies generating better results. This is depicted by Figure 1 above
where a country’s patent output is ranked by their 2008 GDP from left to right with little or
no correlation, positive or negative, between a country’s GDP and patent output per million
of population.
As Furman et al. (2002) point out there is a strong patenting bias in those countries which
have a history of patents production such the US and Switzerland (due to path dependency
and the importance of the history of resource commitments). However, other ‘new’
innovative countries’ rate of growth in patents per million has been nothing short of
phenomenal: Singapore, for example, has an average annual patent growth rate of 30%
between 1981 and 2008, going from just over 1 patent per million in 1981 to 84 in 2008.
Such performance begs analysis and raises the question for us in this paper as to whether
smaller economies generally are supported or hindered by their relatively low scale, or low
critical mass in economic terms, in achieving innovative success. The varying rates of
increase in patent production is treated in detail in Furman and Hayes (2004).
As with any economic definition of success, innovative success requires elaboration and
explanation. In the context of this study the selected measure of innovative ‘success’ is
represented by patent output, which is far from problematic and will be detailed further in
Section 5.1.
The issues considered in this paper focuses, firstly, on whether the mix of drivers of
Innovative Capacity vary across advanced economies when categorised by their SOE status.
Thus, this paper addresses possible heterogeneities that may exist in relation to different
economy structures. We examine the extent to which a set of factors drive a nation’s
Innovative Capacity as previously found in the literature and question whether or not the mix
of policy choices, in terms of the areas mentioned above, for an SOE are significantly
0
50
100
150
200
250
300
Patents per Million POP (2008),
Ranked by 2008 GDP ($) L to R
3
different from other economies. That is, do SOEs perform differently in terms of their
innovative output (patenting activity) compared to their advanced economy peers when using
the same policy mix?
Secondly we examine whether an SOE’s innovative capacity is optimised by an emphasis on
certain on different, if the same basic mix of factors is found to be effective in the first
instance.
This question addresses a gap in the literature and, therefore, it is necessary to assess the
relative performance of SOEs. While some literature on innovative capacity examines
specific SOEs, such as New Zealand in Marsh (2000), it tends to concentrate on an individual
industry without adopting a broader international perspective which is the chosen perspective
offered here, grounded in the National Innovative Capacity approach.
We set out the background to the National Innovative Capacity approach in Section 2
outlining its constituent parts, and potentially relevant measures. Our model for estimation is
presented in Section 3, with Data described in Section 4. Empirical results for various
specifications will be examined in Section 5.
2 National Innovative Capacity Framework
The National Innovative Capacity framework integrates three perspectives on the sources of
innovation i.e.:
ideas-driven growth theory as outlined in Romer (1990);
microeconomics-based models of national competitive advantage and industrial clusters
developed by Porter (1990); and
research on national innovation systems as proposed in Nelson (1993).
Both the characteristics of the direct producers of patents are relevant in this context, as are
the outcomes generated by previous investments, policies and supports for innovation-based
activities.
Our view is that Innovative Capacity should be viewed differently to science and technology
advances, as we are interested in economically viable applications. The discovery of a new
technology (or significant facts/information) is considered to be independent of its benefit to
an economy unless it can be harnessed domestically through having the structures and
resources available to exploit its value before the knowledge becomes diffused and may be
exploited elsewhere. With limited data availability and suitability for identification of
economically viable applications of scientific advances, however, we limit ourselves to a
proxy in the form of patents.
The National Innovative Capacity Framework is illustrated diagrammatically in Figure. 1.
4
Fig. 1: National Innovative Capacity Framework
Source: Furman et al. (2002: 906)
2.1 Common Innovative Infrastructure
This element of the framework relates to features of an economy’s innovation infrastructure
that confer no particular advantage on any sector or cluster yet provide support for innovation
activities generally across the economy. Furman et al. (2002) appeal to endogenous growth
theory to identify the two main determinants in their model of the quality of the Common
Innovative Infrastructure: the aggregate level of technological sophistication or its
accumulated stock of knowledge - denoted by A in Fig.1 - and the range of the talent pool of
workers appropriate for the generation of new knowledge in an economy (denoted by HA in
Fig. 1). In addition to these to determinants they add other universal factors that aid
innovation such as Higher Education Graduates, Property Rights Protection, the availability
of R&D Tax Credits, all of which are denoted by XINF in Fig. 1.
Gans and Stern (2003) offer a broader list of potentially relevant variables that may be
included under the heading of XINF, given below:
Investment in basic research
Tax policies affecting corporate R&D and investment spending
Supply of risk capital
Aggregate level of education in the population
Pool of talent in science and technology
5
Information and communication infrastructure
Protection of intellectual property
Openness to international trade and investment
Overall sophistication of demand
2.2 Cluster Specific Environment
This aspect of the innovative capacity of an economy makes reference to microeconomic
theory, specifically the fact that while wider policy-related issues facilitate innovation it is
ultimately firms that create new technologies. This firm-level impact on national innovative
capacity depends upon the microeconomic environment present within and across a nation’s
clusters (following the definition by Porter, 1990).
A variety of cluster-specific circumstances, investments, and policies impact on the extent to
which a country’s industrial clusters compete on the basis of technological innovation.
Innovation in particular pairs of clusters may also be complementary to one another, both due
to knowledge spillovers and other interrelationships (represented by lines connecting selected
‘Diamonds’ in Fig. 1).
This is a particularly difficult feature to include when estimating an econometric model as
there are few national or international statistics pertaining directly to the extent of cluster
activity that are available for the period of the analysis conducted here (for more on issues in
the challenges of applying a cluster approach see Doyle and Fanning, 2007). Instead a
number of proxies are identified and estimated for our purposes in this paper.
2.3 Quality of Linkages
The quality of the two previous factors is reinforced by the way in which they are linked, as
depicted in Fig. 1. For instance even firms within a well developed cluster will not be able to
produce economically viable new-to-world technologies unless they have access to a pool of
scientists and engineers and access to basic research and, in some cases, perhaps access to
advice from local universities.
3 Modelling Innovative Capacity
The basis of the model specified by Furman et al. (2002) uses the findings of prior research
into the geographic impact of knowledge spillovers, the differences in access to human
capital and ways that regional differences are driven by public policies. Ideas driven
endogenous growth models form the base of the model that is extended to incorporate
additional and more nuanced factors previously not used from industrial organisation, the
composition of funding (public versus private), public policies and the degree of
technological specialisation. For example, while Public R&D spending adds to the innovative
6
process by reinforcing the Common Innovation Infrastructure, Private R&D spending and the
Specialisation of a country’s technological outputs also reflects the nation’s cluster
innovation environment.
To estimate the relationship between the production of international patents and observable
contributors to national innovative capacity, we adopt the ideas production function of
endogenous growth theory as a baseline (Equation 1):
Aj,t = δHAj,tAφt (1)
where Aj,t is the flow of new-to-the-world technologies from country j in year t, HAj,t the total
level of capital and labor resources or inputs devoted to the ideas sector of the economy, and
Aφj ,t is the total stock of knowledge held by an economy at a given point in time relevant to
drive future ideas production.
As the national innovative capacity framework suggests that a broader set of influences
determine innovative performance a production function for new-to-the-world technologies is
extended from Equation 1 generating the formulation of Equation 2:
Aj,t = δj,t(XINFj,t , YCLUSj,t ,ZLINKj,t )HAj,t Aφj,t (2)
The additional variable XINF refers to the level of general resource commitments and policy
choices that constitute the common innovation infrastructure, YCLUS refers to the particular
environment for innovation in a countries’ industrial clusters, and ZLINK captures the strength
of linkages between the common infrastructure and the nation’s various clusters. Under
Equation (2), we assume that the various elements of national innovative capacity are
complementary in the sense that the marginal boost to ideas production from increasing one
factor is increasing in the level of all of the other factors.
The parameters associated with Equation 2 are estimated using a panel dataset of 23 OECD
countries plus Singapore over 13 years. These estimates can therefore depend on cross-
sectional variation, time-series variation, or both. While comparisons across countries can
easily lead to problems of unobserved heterogeneity, cross-sectional variation provides the
direct inter-country comparisons that can reveal the importance of specific determinants of
national innovative capacity. Time-series variation may be subject to its own sources of
endogeneity (e.g. shifts in a country’s fundamentals may reflect idiosyncratic circumstances
in its environment), yet time-series variation provides insight into how a country’s choices
manifest themselves in terms of observed innovative output.
Recognizing the issues surrounding panel estimations our analysis explicitly compares how
estimates vary depending on the source of identification. In most of our analysis, we include
either year dummies in order to account for the evolving differences across years in the
7
overall levels of innovative output and a dummy on the US to account for differences in the
definition of the dependant variable for that economy (explained further below).
The analysis is organized around a log–log specification, except for qualitative variables and
variables expressed as a percentage. The estimates, thus, have a natural interpretation in terms
of elasticities, are less sensitive to outliers, and are consistent with the majority of prior work
in this area including Jones (1998), Furman et al. (2002), Gans and Stern (2003), Gans and
Hayes (2008).
Finally, with regard to the sources of error, we assume that the observed difference from the
predicted value given by Eq. (2) (i.e. the disturbance) arises from an idiosyncratic
country/year-specific technology shock unrelated to the fundamental determinants of national
innovative capacity. Integrating these choices and letting L denote the natural logarithm, our
main specification takes the following form of Equation 3:
L Aj,t = + +δINFLXINF j,t + δCLUSL Y CLUS j,t
(3)
+δLINKL ZLINK j,t + λ LHA j,t +φL Aφj,t + εj,t
Conditional on a given level of R&D inputs (HA), variation in the production of innovation
(Aφ) reflects R&D productivity differences across countries or over time. For example, a
positive coefficient on elements of δINF, δCLUS or δLINK suggests that the productivity of R&D
investment is increasing in the quality of the common innovation infrastructure, the
innovation environment in the nation’s industrial clusters, and the quality of linkages. As Aj,t,
measured by the level of international patenting, is only observed with delay, our empirical
work imposes a 3-year lag between the measures of innovative capacity and the observed
realization of innovative output. This follows Furman et al. (2002), but differs from Gans
and Stern (2003) and Gans and Hayes (2008), who impose no lag and a two year lag
respectively: including alternative lag structures does not significantly alter our results.
4 Possible Reasons for SOE Heterogeneity
A number of ideas may contribute to SOE’s having a different set of factors that add to its
innovative capacity; or that some factors may be of greater importance in maximising its
potential for ideas production.
4.1 Scale Effects
The idea that in larger conglomerations of people new ideas and innovations will be more
readily available is an old one. William Petty (1682) commenting on the reconstruction of
London after the Great Fire of 1666 wrote:
8
“it is more likely that one ingenious curious man may rather be found amongst
4 millions than among 400 persons.”
This Scale Effect has also been incorporated in neo-classical growth theories such as Romer
(1990) which forms a part of the basis for the National Innovation Capacity framework used
in this paper. In Romer’s (1990) model, the growth rate of an economy is proportional to the
total amount of research undertaken in it. And an increase in the population of an economy
will generally lead to an increase in the R&D workforce, thereby increasing the growth rate
of per capita income.
So that a small open economy’s ability to produce new to world technologies may be affected
due to the lower probability of it’s having the “ingenious curious” persons that William Petty
talks about as well as lacking the critical mass of researchers to maximise growth as in
Romer’s theory.
In studies testing scale effects in economic growth however there is no clear evidence that
larger economies grow faster. Jones (1995b) studied time series evidence and concluded
against scale effects in economic growth. However on a more concentrated level Backus
(1992) found scale effects were evident in the manufacturing output in the variety of models
used.
4.2 Knowledge Spillovers
Studies such as Faehn (2008) emphasise the importance of knowledge spillovers for SOE’s.
Cohen and Levinthal (1989) point out that R&D has “two faces” in its interaction with an
economy. Not only does it produce new innovations it also allows for the easier absorption
and understanding of new technologies, both domestically and internationally.
Due to the lack of scale in SOE’s discussed in the earlier section the importance to SOE’s of
absorbing all knowledge internationally in order to be able to act at the technological fromtier
in producing new to world technologies means that a different policy emphasis may be
required. Faehn (2008) discusses how national policy can enhance the exploitation of the
international knowledge stock and find that subsiding R&D is important for domestic
innovation as it is effective in generating these knowledge spillovers from abroad. Faehn
(2008) also finds that eports play an important role for SOE’s in encouaraging knowledge
spillovers.
9
5 Data
5.1 Innovative Output
This analysis requires an observable country-specific indicator of the level of commercially
valuable innovative output in a given year. We follow previous research and employ as the
dependant variable the number of “international patents” (PATENTS), defined as “the
number of patents granted to inventors from a particular country other than United States by
the USPTO in a given year. For the United States, PATENTS is equal to the number of
patents granted to corporate or government establishments (this excludes individual
inventors)” (Furman et al. 2002: 909).
Following Eaton and Kortum (1996), Kortum and Lerner (1999), Griliches (1984) and
Furman et al. (2002) we recognise a number of difficulties in relation to using patents as a
measure of innovation at a national level, such as:
Not all inventions are patentable,
Not all inventions of economic value are patented,
Not all patented innovations have the same quality or value to an economy,
There are varying degrees of willingness to patent across countries and sectors.
However a large number of previous studies used patents based on the assumption that they
are “the only observable manifestation of inventive activity with a well-grounded claim for
universality” (Trajtenberg, 1990:183). We temper this assumption by interpreting our
findings carefully, noting that our dependant variable is an imperfect proxy relating to
innovations that are economically viable applications and that the true rate of innovation is
unobservable.
This belief is based on the expense that is required for a non-american investor to register a
patent with the USPTO acting as a barrier unless there is a strong belief that the innovation
will produce a sufficient return. A patent registered by an America will be either from a firm
or the government, again reducing the number of patents registered lacking economic value.
Any asymmetry this may cause between US and non-US patents does not affect our results as
we include a US dummy variable in all regressions, in keeping with the previous literature.
5.2 Defining Small Open Economies
There are 30 economies in the OECD and this study also includes Singapore as a strong
example of an SOE for which patenting activity has become particularly strong. However,
Portugal, Turkey and Iceland, Greece were not used as observations as required data was
available for the Specialisation variable (detailed below). In addition to this Luxembourg was
excluded as its small size added very little to the data and as it would be given the same
10
weighting would bias the results. Mexico and Poland were also excluded due to the extremely
low level of patenting when shown relative to population as in Fig 2.
In the literature there appears to be no one method applied to define an SOE. The conceptual
definition of an SOE is an economy that participates in international trade, but is small
enough compared to its trading partners that its policies do not alter world prices, interest
rates, or incomes. However this paper must divide the 23 countries estimated into SOEs and
others and in order to do this data was collected on:
1) Import/Export Openness of the economy, calculated as exports plus imports divided
by GDP taken from the Penn World Tables.
2) Size of the economy, calculated as the relative GDP weighting of each in our overall
sample. The GDPs of the 31 countries were aggregated and the proportion each
accounted for was calculated.
For the purposes of this paper an SOE is defined as one whose GDP makes up less than 2%
of the 31 countries aggregate GDP and its exports plus imports over GDP is equal to or
greater than 70%, which is within half a standard deviation of the mean of 100%.
Fig. 2: Patents per Million of Population: Large Countries and SOEs
0
50
100
150
200
250
300
350
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
Patents per Million
Patents Per Million Pop
- Large
Canada
Germany
France
Italy
Japan
Korea
Spain
11
Table 1 provides a full list of the sampled countries and their status as an SOE or ‘other’:
Turkey was not an SOE under these criteria, and Portugal and Iceland were SOEs. This
process was somewhat ad hoc and some countries were border-line. Canada, for instance, is
studied as an SOE by Appelbaum and Kohli (1979), but its GDP was nearly 3% of the
aggregate, while its international trade openness was 87% of GDP.
Table 1: Selection of Small Open Economies and Others
SOE
Size:% of
Openness
Size: % of
Openness
Aggregate
GDP
Large
aggregate
GDP
Australia
1.36%
46%
USA
37.81%
26%
Austria
0.74%
101%
UK
5.70%
58%
Belgium
0.90%
169%
Japan
17.98%
20%
Czech Republic
0.22%
147%
Germany
7.34%
67%
Denmark
0.62%
80%
France
5.14%
56%
Finland
0.47%
76%
Italy
4.24%
56%
Hungary
0.18%
127%
Canada
2.79%
87%
Ireland
0.37%
176%
Spain
2.24%
62%
Netherlands
1.49%
130%
South Korea
2.06%
87%
New Zealand
0.20%
72%
Norway
0.65%
77%
Singapore
0.36%
342%
Sweden
0.95%
89%
Switzerland
0.96%
88%
Source: Authors’ calculations based on data from Penn World Tables.
0
50
100
150
200
250
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
Patents per Million Pop
Patents Per Million Population
- SOE's Australia
Austria
Belgium
Denmark
Finland
Hungary
Ireland
Netherlands
12
5.3 Independent Variables
Following the previous literature this paper uses proxies for measures of Common Innovative
Infrastructure, Cluster Specific Environment and the Quality of Linkages in order to estimate
the determinants of National Innovative Capacity. These are detailed in Table 2 which
includes variable definitions and sources. Table 3 details variable means and standard
deviations.
Table 2: Variable Descriptions and Sources
Dependent Variable
Full Variable Name
Definition
Source
Patents j,t
International Patents
Granted by Year of
Application
Non-US countries:
patents granted by the
USPTO.
US: patents granted by
the USPTO to
corporations or
government.
USPTO Patent
Database
Independent Variables
QUALITY OF THE COMMON INNOVATION INFRASTRUCTURE
R&D PPL j,t
Aggregate Personnel
Employed in R&D
Full time equivalent
R&D personnel in all
sectors
OECD Science &
Technology
Indicators
R&D $ j,t
Aggregate
Expenditure on R&D
Total R&D
expenditures in
Mill of US$ (base 2000)
OECD Science &
Technology
Indicators
Property Rights
Protection j,t
Legal Structure and
Security of Property
Rights
Average survey
response by executives
on a 1-10 scale
regarding relative
strength
of Legal Structure and
Security of Property
Rights
Economic Freedom of
the World Index
ED SHARE j,t
Share of GDP Spent
on Secondary and
Tertiary Education
Public spending on
secondary + tertiary
educ. as share of GDP
World Bank: Edstats
OPENNESS j,t
Freedom to Trade
Internationally
Average survey
response by executives
on a 1-10 scale
regarding relative
strength of freedom to
trade internationally
Economic Freedom of
the World Index
13
GDP/POP j,t
GDP Per Capita
Gross Domestic
Product per capita,
constant price, chain
series, US$
IMF: World Economic
Outlook
GDP j,t
GDP
Gross Domestic
Product constant price,
chain series,
US$, Billions.
IMF: World Economic
Outlook
CLUSTER-SPECIFIC INNOVATION ENVIRONMENT
PRIVATE R&D j,t
Percentage of R&D
Funded by Private
Industry
R&D expenditures
funded by industry
divided by total
R&D expenditures
OECD Science &
Technology
Indicators
SPECIALISATION j,t
E-G concentration
index, excluding the
US
Relative concentration
of innovative output in
chemical, electrical
and mechanical USPTO
patent classes
Computation from
USPTO data using
formulae from Furman
et al. (2002) detailed
below
QUALITY OF LINKAGES
UNIV R&D j,t
Percentage of R&D
Performed by
Universities
R&D expenditures
of universities divided
by total national R&D
expenditures
OECD Science &
Technology
Indicators
Table 3: Variable means and standard deviations
Variable
N
Mean
Standard Deviation
Patents j,t
293
4,607
10,081
QUALITY OF THE COMMON INNOVATION INFRASTRUCTURE
R&D PPL j,t
293
193,276
300,649
R&D $ j,t
293
24,687
49,779
Property Rights
Protection j,t
293
8.21
1.21
ED SHARE j,t
293
3.4
0.94
OPENNESS j,t
293
8.1
0.63
GDP/POP j,t
293
25,619
9,480
GDP j,t
293
1,120
2,059
CLUSTER-SPECIFIC INNOVATION ENVIRONMENT
PRIVATE R&D j,t
293
61
11
SPECIALISATION j,t
293
0.53
0.64
QUALITY OF LINKAGES
UNIV R&D j,t
293
22
6
14
5.4 Specialisation
While most variables for our analysis were readily available, SPECIALISATION was
estimated based on a methodology developed by Ellison and Glaeser (1997). Since
individual clusters will tend to be associated with technologies from specific technological
areas, this is a measure of the degree of technological focus by a country
(SPECIALISATION) acts as a proxy for the intensity of innovation-based competition in a
nation’s clusters. SPECIALISATION is a “relative” concentration index based on the degree
to which a given country’s USPTO-granted patents are concentrated across three broad
technology classes (chemical, electronics, and mechanical) which cover all patents. While the
measure of specialization is too general to identify specific clusters and the role of the mix of
clusters in shaping R&D productivity, SPECIALIZATION is designed as a noisy but
unbiased measure capturing an important consequence of cluster dynamics, the relative
specialization of national economies in specific technologies fields.
Specifically, traditional measures of specialization, such as the Herfindahl Index ignore two
issues important for cross-country comparisons: technology classes differ in terms of their
average share across all countries and some countries have only a small number of patents
overall. While the Ellison and Glaeser index was developed and applied for measuring the
specialization of industries across geographic regions, Furman et al (2002) applied in several
other contexts, including the measurement of the degree of specialisation of research output
following previous authors such as Lim (2000). In the present context, the Ellison and
Glaeser formula adjusts the country observed shares for each technology class to account for
the average share for that technology group across the sample; and for the total number of
patents in each “country–year” observation, as shown:
jtji
ti
ititji
tji
tji
tji PATENTS
x
xs
Patents
Patents
ionSpecalisat
,,
2
,
2
,,,
1,,
,,
,, 1
1
)(
where: Patents i,j,t = Patents of country i in year t across each technology class j,
s i,j,t = share of class j patents in total country patents in year t,
xi,t = average share of patent class j over all i in any t
Figure 2 below offers a sample of results of the Specialisation measure.
15
Source: Authors’ calculations based on data from UPSTO data.
6 Empirical Results
This section outlines the results from our empirical analyses enabling a dissection of the
drivers of national innovative capacity across our sample of 23 countries and with specific
focus on the results for the sample of SOEs, as defined earlier. We first test all models
presented in Section 6.2 for parameter stability. We then included slope and level dummies
into our regressions to assess if there are specific differences in the way that SOE’s and other
economy types produce new to world technologies.
The panel regression method used is the Random Effects Method. As Baltagi (2005: 28)
explained:
The random effects model is an appropriate specification if we are drawing N
individuals randomly from a large population
As this study focuses on a sample of SOEs and non-SOE economies, this is an appropriate
method. In addition, each regression is tested using the Breusch and Pagan Lagrangian
Multiplier test of Fixed versus Random Effects to assess the appropriateness of the method:
all regressions are found to be suitable for Random Effects estimation.
6.1.1 Chow Tests for Parameter Stability: Methodology
All models estimated in Section 6.2 were tested for parameter stability using the Chow test.
This test assesses if there is a difference in the structure of the relationship between the
dependant variable and the independent variables, when estimated for the SOE’s or larger
economies.
This is done by estimating each model 3 times: Equ.1 with all country observations, Equ.2
with just SOE countries and Equ.3 for the non-SOE countries. Equ.1 assumes that the
intercept as well as the slope coefficients remains the same for both economy types; that is,
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Fig. 2: Specalisation Index
GERMANY
FRANCE
CANADA
BELGIUM
AUSTRIA
Australia
16
there is no structural difference in how the two economy types produce patents. Equ’s 2 and 3
assume there is a structural difference.
To carry out the chow test we run all three regressions to find the residual sum of squares
(RSS). From Equ.1 we find the restricted RSS (RSSr) as we are forcing the coefficients to
have the same value for both economy types. We now assume the other two regressions to be
independent and add their RSS’s to get the RSSur. If there is no structural change, then the
RSSR and RSSUR should not be statistically different. Therefore, if we form the following
ratio:
)]2(,[
32 21
~
)2/( /)( knnk
ur
urrF
knnRSS kRSSRSS
F
Where:
k = No. of parameters estimated
n2 + n3 = Number of observations in regression 2 and 3 respectively
We do not reject the null hypothesis of parameter stability (i.e., of no structural change) if the
computed F value in an application does not exceed the critical F value found in the F tables
at a given level of significance.
6.1.2 Chow Tests for Parameter Stability: Results
All models estimated and tested for parameter stability showed that there was no change in
the structure of the relationship between the dependant variable and the independent
variables, when estimated for SOE’s or larger economies. This means that based on these
findings SOEs innovation is driven by the same set of factors as other economy structures.
Section 6.2 will investigate where there are specific factors that a have statistically
significantly different effect on innovative output. Full results of regressions using a full
sample of countries and the SOE sample are shown in Appendix 2.
6.2 Dummy Regressions: Results
Results for regressions including an SOE slope dummy for each of the variables national
innovative capacity are shown in Table 3, 4 and 5 below, along with an intercept dummy.
Regression coefficients from each single regression line are shown in two columns with the
independent variables in the right hand column and the equivalent dummy in the left hand
column.
Regressions are grouped into three categories. Ideas Production Functions (Table 1),
Common Innovative Infrastructure and National Innovative Capacity (Tables 3 and 4).
Tables 3 uses GDP, Population and GDP per capita as proxies for a countries knowledge
stock while in Table 4 the stock of patents built up by the country is used as a proxy.
17
Table 3: Determinants of New to World Technologies (GDP/POP as
Knowledge Stock)
Ideas Production Functions
Equ. 1.1
Equ. 1.3
Independent
Variables
Equivalent
Dummy
Independent
Variables
Equivalent
Dummy
Constant
-2.6044
-1.6688
1.2125
-6.1050
(0.036)
(0.289)
(0.444)
(0.002)
L GDP
-0.0026
-0.0885
(0.983)
(0.588)
L GDP PER CAPITA
0.03111
0.01014
(0.811)
(0.544)
L POP
1.6762
-2.120
(0.000)
(0.000)
L R&D PPL
0.2696
0.6972
0.4919
0.3668
(0.002)
(0.000)
(0.000)
(0.004)
R2
0.6673
0.5954
Table 4: Determinants of New to World Technologies (GDP/POP as Knowledge Stock)
Common Innovative Infrastructure
National Innovative
Capacity
Equ. 1.4
Equ. 1.5
Equ. 1.6
Independent
Variables
Equivalent
Dummy
Independent
Variables
Equivalent
Dummy
Independent
Variables
Equivalent
Dummy
Constant
-6.3061
-1.1061
-6.0268
-3.9321
-8.2346
-1.7041
(0.000)
(.0556)
(0.000)
(0.088)
(0.000)
(0.459)
L GDP
0.0674
0.2552
(0.600)
(0.120)
L GDP PER
0.0504
0.2608
0.2682
0.0168
CAPITA
(0.697)
(0.102)
(0.072)
(0.924)
L POP
0.4300
-0.6694
(0.146)
(0.046)
L R&D PPL
-0.05881
0.9280
-0.0779
0.1631
0.0151
0.0523
(0.584)
(0.686)
(0.461)
(0.437)
(0.887)
(0.798)
L R&D $
1.0101
-0.1767
1.1898
-0.3597
1.2692
-0.4024
(0.000)
(0.407)
(0.000)
(0.069)
(0.000)
(0.055)
ED SHARE
0.0050
0.0782
-0.0189
0.1050
-0.0016
0.0701
(0.938)
(0.305)
(0.796)
(0.165)
(0.979)
(0.338)
IP
0.1790
0.0378
0.1904
0.0375
0.1648
0.0679
(0.000)
(0.629)
(0.000)
(0.628)
(0.001)
(0.372)
Openness
0.1379
0.1890
0.1246
0.2028
0.2401
0.0793
(0.069)
(0.043)
(0.102)
(0.030)
(0.003)
(0.415)
Private R&D
-0.0293
0.0283
(0.000)
(0.003)
Specialisation
0.3730
-0.7489
(0.289)
(.0.077)
University
-0.0381
0.0510
R&D
(0.000)
(0.000)
R2
0.9380
0.9487
0.9517
18
Table 5: Determinants of New to World Technologies (Patent Stock
as Knowledge Stock)
Common Innovative
Infrastructure
National Innovative
Capacity
Equ. 3.1
Equ. 3.2
Independent
Variables
Equivalent
Dummy
Independent
Variables
Equivalent
Dummy
Constant
-5.7483
0.8037
-4.0317
-0.7616
(0.000)
(0.667)
(0.018)
(0.722)
L Patent
0.3586
-0.1770
0.4138
-0.2075
Stock
(0.001)
(0.089)
(0.000)
(0.098)
L R&D PPL
-0.048
-0.2499
0.0461
-0.2405
(0.613)
(0.297)
(0.657)
(0.220)
L R&D $
0.6445
0.2499
0.5803
0.2832
(0.001)
(0.297)
(0.015)
(0.310)
ED SHARE
-0.0484
0.0921
-0.0689
0.0864
(0.434)
(0.208)
(0.261)
(0.233)
IP
0.1092
0.1089
0.1055
0.0770
(0.031)
(0.155)
(0.038)
(0.320)
Openness
0.0878
0.1376
0.1465
0.0800
(0.207)
(0.113)
(0.042)
(0.363)
Private R&D
-0.0210
0.0189
(0.017)
(0.058)
Specialisation
0.0859
-0.7476
(.808)
(0.084)
University
-0.0333
0.0345
R&D
(0.000)
(0.005)
L GDP 1993
0.3553
-0.1901
0.1861
-0.0865
(0.038)
(0.367)
(0.056)
(0.716)
R2
0.949
0.953
As pointed out above, a country’s Patent Stock has been shown to be a major factor in
determining its current and future patent output. This analysis agrees also finds that it plays a
statistically significant role with a 10% non-SOE’s sample. The SOE dummy variable is
significant and shows that patent stock has roughly half the effect in an SOE.
This may point to the fact that Patent stock not only captures the accumulated knowledge
stock of the country but also the fact that a country with a large stock of patents may well
have a more fully developed innovative infrastructure. The sample of SOE’s tends to contain
the lesser developed of the sample countries with notable exceptions.
The level of economic development was proxied by GDP and GDP per Capita. When these
factors were examined without reference to patent stock it was found that results were neither
statistically or economically significant in general but had a positive relationship with patent
output. This could be explicable as it was essentially measuring whether changes in the level
of economic development resulted in changes in patent output. But as the countries in this
19
sample are all well developed it is possible that improvements in an already advanced
economy had only a marginal effect.
When Patent Stock was included, rather than test for changes in the level of development
from year to year only the level of development in 1993, the beginning of the sample, was
used to give a different and static level of development for each country. For the whole
sample of advanced economies this was significant and large with a 10% difference in a
country’s level of development in 1993 resulting in a 3.5% increase in patenting. For SOE’s
it is a less important determinant being both economically and statistically significant.
20
Appendix 1: Countries sampled in this paper and Furman et al. (2002), with time scale.
Furman et al. (2002),
Sample countries (1973–1995)
Doyle et al. (2009), Sample
Countries (1993–2005)
Australia
Australia
Austria
Austria
Canada
Belgium
Denmark
Canada
Finland
Czech Republic
France
Denmark
Germany
Finland
Italy
France
Japan
Germany
Netherlands
Hungry
New Zealand
Ireland
Norway
Italy
Spain
Japan
Sweden
Netherlands
Switzerland
New Zealand
UK
Norway
United States
Singapore
South Korea
Spain
Sweden
Switzerland
UK
United States
21
References
Appelbaum, E. and Kohli, U. (1979). Canada-United States Trade: Tests for the Small-Open-
Economy Hypothesis. The Canadian Journal of Economics, 12 (1): 1-14.
Baltagi, B.H. (2005) 3rd Edition. Econometric Analysis of Panel Data. John Wiley & Sons,
England.
Doyle, E. and Fanning, C. (2007) The Role for Clusters in Irish Economic Development
Policy, Chapter 16 in Aylward, C., O’Toole, R. (eds.), Perspectives on Irish Productivity: A
Selection of Essays by Irish and International Economists, Dublin: Forfas, pp. 267-281.
Ellison, G. and Glaeser, E. (1997). Geographic concentration in US manufacturing industries:
a Dartboard approach. Journal of Political Economy 105 (5):889–927.
Furman, J., Porter, M. and Stern, S. (2002). The Determinants of National Innovative
Capacity. Research Policy, 31: 899-933.
Furman, J., Hayes, R. (2004). Cathing up or standing still? National Innovative capacity
amoung ‘follower’ countries, 1978 - 1999. Research Policy, 33: 1329 - 1354.
Gans, J. and Stern, S. (2003). Assessing Australia’s Innovative Capacity in the 21st Century,
IPRIA.
Gans, J. and Stern, S. (2008). Assessing Australia’s Innovative Capacity: 2007 Update,
IPRIA.
Griliches, Z. (Ed.) (1984). R&D, Patents and Productivity. Chicago University Press,
Chicago, IL.
Jones, C. (1995a). R&D based models of economic growth. Journal of Political Economy,
103:739–784.
Jones CI (1995b) Time series tests of endogenous growth models. Quarterly Journal of
Economics 110: 495±525
Lim, K. (2000). Basic research, applied research, and innovation in the semiconductor and
pharmaceutical industries. Unpublished doctoral dissertation, MIT Sloan School of
Management.
Marsh, D. (2000). Fostering innovation in a small open economy: The case of the New
Zealand biotechnology sector. (Department of Economics Working Paper Series, Number
1/00). Hamilton, New Zealand: University of Waikato.
Pavitt, K. (1982). R&D, patenting, and innovative activities: a statistical exploration.
Research Policy, 11 (1):33–51.
22
Petty W (1682) Another essay on political arithmetick, concerning the growth of the city
of London. London
Porter, M.E.. (1990). The Competitive Advantage of Nations. Free Press, New York.
Solow, R.M. (1956). A contribution to the theory of economic growth. Quarterly Journal of
Economics, 70:65–94.
Trajtenberg, M. (1990). Patents as Indicators of Innovation, Economic Analysis of Product
Innovation. Harvard University Press, Cambridge, MA.
