SlideShare une entreprise Scribd logo
1  sur  71
Télécharger pour lire hors ligne
Linking National Systems of Innovation and Economic
 Growth under the Knowledge Economy Framework
     With an overview of the Colombian case-study

                                   by


                     Andrés Barreneche García
         B.Sc., University of the Andes (Colombia), 2008




       A Dissertation Submitted in Partial Ful llment of the
                   Requirements for the Degree of


               M             A          E




              Main Supervisor: Prof. Dr. Yoichi Koike
            Second Supervisor: Prof. Dr. Yongjin Park




                       Ritsumeikan University
                       Graduate School of Economics
                       July 2010
ii
Linking National Systems of Innovation and Economic
 Growth under the Knowledge Economy Framework
     With an overview of the Colombian case-study

                                           by


                            Andrés Barreneche García∗
                  B.Sc., University of the Andes (Colombia), 2008



                      Main Supervisor: Prof. Dr. Yoichi Koike
                     Second Supervisor: Prof. Dr. Yongjin Park




                                       Abstract

   This dissertation applies the Knowledge Economy (KE) Framework developed by
   the W         B    as a means to assess the e ect of National System of Inno-
   vation (NSI) performance on economic growth. The KE approach integrates the
   NSI concept as one of the four factors deemed to enhance economic output in
   terms of knowledge creation, di usion and adaptation; these are: the Economic
   Regime, the Innovation System, Education and Information and Communica-
   tion Technologies. This framework is employed for an empirical study about the
   connection between KE variables and economic growth with a sample of 75 coun-
   tries (developed and developing) in the [1998, 2007] period. This work concludes
   that higher NSI performance, as a function of foreign technology transfer (man-
   ufactures imports and FDI) and knowledge appropriation (R&D expenditure and
   high-technology exports) variables, is conducive to superior increments of GDP.
   Furthermore, this dissertation advances the discussion of the Colombian case-
   study and diagnoses that the country has failed to harness opportunities of foreign
   technology transfer as a consequence of the disjunction between NSI actors.




∗ E-mail:   barreneche@gmail.com

                                           iii
“We have the good fortune to live in democracies, in which individuals can ght for their
perception of what a better world might be like. We as academics have the good fortune to be
         further protected by our academic freedom. With freedom comes responsibility: the
responsibility to use that freedom to do what we can to ensure that the world of the future be
     one in which there is not only greater economic prosperity, but also more social justice.”
                                     Joseph Stiglitz. Nobel Prize Lecture, December 8th, 2001.




                                             iv
Acknowledgements

I would like to thank:

Professor Yoichi Koike, for his guidance, support, helpful comments, patience, and
     for giving me the opportunity to cultivate my ideas while keeping me on track.

Ritsumeikan University: professors, sta members and colleagues, for contribut-
     ing towards a gratifying academic experience in Japan.

The Government of Japan (MEXT), for funding me with a scholarship.

Alejandro Hoyos Suárez, for his friendship and for providing me with thought-
     ful advice in the writing of this dissertation and through my studies in the
     master’s program.

My family: my parents Juan José Barreneche Silva and Maria Cristina García de
     Barreneche and my brother Alejandro Barreneche García, for the uncondi-
     tional love they have provided me in spite of the thousands of kilometers that
     have separated us.

Sebastián Perez Saaibi, for his encouragement and unquestionable companionship
     as a fellow Colombian expatriate.




                                         v
Contents

Abstract                                                                                 iii

Acknowledgements                                                                          v

Table of Contents                                                                        vi

List of Tables                                                                          viii

List of Figures                                                                          ix

1 Introduction                                                                            1

2 Background                                                                              3
   2.1 Endogenous Growth and Innovation . . . . . . . . . . . . . . . . . .               3
   2.2 The ‘National System of Innovation’ Approach . . . . . . . . . . . .               5
        2.2.1     Empirical Studies of NSI . . . . . . . . . . . . . . . . . . . . .      8
   2.3 Leveraging on the Knowledge Economy Framework . . . . . . . . .                    9
   2.4 Research Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . .         12

3 Data and Methodology for Analysis                                                      14
   3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .    14
        3.1.1     Rationale for Variable Selection . . . . . . . . . . . . . . . . .     15
   3.2 The Construction of KE Pillar Indices . . . . . . . . . . . . . . . . . .         19
   3.3 Speci cation of Models . . . . . . . . . . . . . . . . . . . . . . . . . .        22

4 Results and Evaluation                                                                 25
   4.1 KE Pillar Indices vs GDP per Capita . . . . . . . . . . . . . . . . . . .         25
   4.2 Econometric Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .        29

5 An Overview of the Colombian National System of Innovation                             35


                                            vi
5.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .      35
   5.2 Current Status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .    38
   5.3 Prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .   40
   5.4 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .       43

6 Conclusions                                                                          47

A Constructed KE Pillar Indices                                                        50

B Statistical Tests                                                                    54
   B.1 Endogeneity: Durbin-Wu Hausman Test . . . . . . . . . . . . . . . .             54
   B.2 Heteroskedasticity: White Test . . . . . . . . . . . . . . . . . . . . . .      55

C Estimations with GDP per Capita as the Explained Variable                            56
   C.1 Speci cation of Models . . . . . . . . . . . . . . . . . . . . . . . . . .      56
   C.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .   57

Bibliography                                                                           58




                                           vii
List of Tables

 Table 3.1 Summary Statistics for the Selected KE Variables . . . . . . . .       16
 Table 3.2 The Constitution of the KE Pillar Indices . . . . . . . . . . . . .    21
 Table 3.3 Summary Statistics for the Regression Variables . . . . . . . . .      22
 Table 3.4 Econometric Analysis: Model De nitions . . . . . . . . . . . .         23

 Table 4.1 OLS Regression Results for GDP Growth. . . . . . . . . . . . .         30

 Table 5.1 Detailed Innovation System Indicators for Colombia . . . . . .         43
 Table 5.2 Retrospective Estimates for Colombia’s GDP Growth (Model 4)            45

 Table A.1 KE Pillar Indices for [1998, 2002] . . . . . . . . . . . . . . . . .   50
 Table A.2 KE Pillar Indices for [2003, 2007] . . . . . . . . . . . . . . . . .   52

 Table B.1 GDP Growth OLS Regressions with added Proxy Residuals . .              54
 Table B.2 White Test Summary Results . . . . . . . . . . . . . . . . . . . .     55

 Table C.1 Model De nitions for GDP per Capita . . . . . . . . . . . . . .        56
 Table C.2 OLS Regression Results for GDP per Capita . . . . . . . . . . .        57




                                       viii
List of Figures

 Figure 4.1 GDP per Capita and Innovation System . . . . . . . . . . . . .       26
 Figure 4.2 GDP per Capita and Economic Incentives . . . . . . . . . . . .       26
 Figure 4.3 GDP per Capita and Governance . . . . . . . . . . . . . . . . .      27
 Figure 4.4 GDP per Capita and Education . . . . . . . . . . . . . . . . . .     27
 Figure 4.5 GDP per Capita and ICT . . . . . . . . . . . . . . . . . . . . . .   28
 Figure 4.6 Marginal E ect of the Innovation System Scores (Model 4) . .         31
 Figure 4.7 Regression’s Residuals vs Innovation System Scores (Model 4)         32




                                       ix
Chapter 1

Introduction

              has become understood to emerge through the interactions of a vari-
I   ety of agents such as rms, universities and governmental bodies. These actors
are considered to have particular roles in processes where knowledge is created,
adapted, di used and incorporated into a speci c good or service. The synergies
taking place in a given country have been notably identi ed and studied by re-
searchers using the National System of Innovation (NSI) concept. This comprehen-
sion has provided tools for science, technology and innovation (STI) policy design.
As follows, these instruments have been widely adopted by public administrators
from a diversity of countries, ranging from OECD founding members such as France
and Finland, to developing countries like Korea and Brazil. The development of an
NSI theory has been, however, hampered by the inherent di culties for empirical
analysis. In contrast to other elds such as nancial economics, the interactions
involved in innovation are di cult to parametrize and thus analyze quantitatively.
As a consequence, the impact of policies derived from the NSI concept has yet to be
fully understood.
    This dissertation leverages the Knowledge Economy (KE) Framework developed
by the W       B      as a means to assess the e ect of NSI performance on eco-
nomic growth. The framework centers on the following idea: the manner in which
applicable knowledge is produced and owing within society is crucial for increas-
ing economic output. The KE approach integrates the NSI concept as one of the
four components, referred as KE Pillars, deemed to enhance growth in terms of
knowledge creation, di usion and adaptation; these are: the Economic Regime, the
Innovation System, Education and Information and Communication Technologies.
This framework allows an empirical study about the connection between KE vari-

                                        1
ables and economic growth for a sample of 75 countries (developed and developing)
in the [1998, 2007] period. This analysis yields a signi cantly positive impact com-
ing from the level of NSI performance, as a function of foreign technology transfer
(manufactures imports and FDI) and knowledge appropriation (R&D expenditure
and high-technology exports) variables, on increments of GDP. Furthermore, this
dissertation advances the discussion of the Colombian case-study and diagnoses
that the country has failed to harness opportunities of foreign technology trans-
fer as a consequence of the disjunction between NSI actors. The organization of
document is described below.

Chapter 2 elaborates on the main problem of concern: to establish a quantitative
     link between the NSI concept and economic growth. A hypothesized solution
     is proposed as an application of the KE Framework. The main concepts and the
     previous studies, in which this dissertation intends to build upon, are revised.

Chapter 3 procures to describe the data and methodology selected to approach the
     hypothesis. The W             B   data for representative variables of the KE
     Framework is presented. It then explains how this information is prepared for
     an empirical study.

Chapter 4 is where the the empirical analysis takes place. The gathered evidence
     is fully described and evaluated in detail.

Chapter 5 reviews the Colombian case-study considering the current approach and
     the gathered evidence. It seeks to place the exposed link between the NSI
     approach and economic growth at the service of policy-making in the context
     of this particular country.

Chapter 6 synthesizes the claims, the process for their validation and the ndings
     of this dissertation. It also mentions the recognized research opportunities for
     further development of the NSI concept and its potential applications.




                                          2
Chapter 2

Background

        chapter surveys the theoretical preliminaries and previous works required
T     to de ne the research hypothesis which aims to connect the NSI approach
with economic growth. On this account, three main subjects are covered. First, a
concise background of economic theory is presented regarding endogenous growth,
in order to understand the role of innovation in the development of markets. The
concept of a NSI is the second topic of discussion. The trajectory of this approach is
described through the exposure of representative studies, in which this dissertation
is based on. Lastly, the KE Framework is introduced in order to lay the grounds for
the quantitative analysis that will be undertaken in the forthcoming chapters.


2.1    Endogenous Growth and Innovation
The Solow model is signi cant for economic growth theory, not only because of its
revealed ndings, but also due to the shortcomings that can be derived from it. This
model has brought a general understanding on how savings, population growth and
technological progress a ect the level of an economy’s output and its growth over
time [Mankiw, 2006]. However, the standard implementation of this model cannot
explain the di erences between per capita production in high-income countries and
that of the least developed countries, or why the average growth of GDP per capita
is much higher in the present time than 200 years ago [Jones and Manuelli, 2005].
Total Factor Productivity has come as a way to recognize these di erences, although
the reasons behind them are still a matter of debate under this approach. These
limitations of the Solow model have been the main inspiration for the subsequent


                                          3
development of Endogenous Growth Theory (EGT).
    According to [Jones and Manuelli, 2005], EGT models have focused on the pro-
duction and dissemination of knowledge, whether by inclusion under the assump-
tions or directly in the system’s speci cation. Studies of this branch have arrived to
a common conclusion: asymmetries in development rely on the di erences between
social institutions across time and countries (e.g. countries with inadequate protec-
tion of property rights will grow at a slower pace). In the case of the basic Solow
model, the role of knowledge (technology) was considered along with labor and
capital, but left unexplained and assumed exogenous. There are two main criticisms
for considering knowledge as an externality, which endogenous growth centered
in resolving: the mentioned di culty to explain observed long run di erences in
economic growth and the fact that changes of productivity are a result of conscious
decisions made by socio-economic agents.
    Innovation1 is a process in which new products and services are developed
through R&D activities that originate in market competition. This permanently on-
going process results in technological progress which is, in turn, the centerpiece of
endogenous growth theory. Most economists accept technological change and inno-
vation as the principal constituents of economic growth [Aghion and Howitt, 2005].
In other words, growth can di cultly be sustainable in the absence of steady tech-
nological improvements [Barro and Sala-i-Martin, 2003]. This idea is supported by
the fact that innovation has been integrated to many contemporary growth models.
Paradoxically, innovation has only received scholarly attention until recent years.
Furthermore, economic studies of innovation have been centered in microeconomics
[Fagerberg, 2006]. However, a particular approach regarding the procurement of in-
novation under the macroeconomic perspective has been developed based on the
concept of a NSI.




    1
     This work implicitly uses the following de nition of innovation. It is a product or service
which ful lls the following conditions: i. it contains a technological novelty either by new devel-
opment, combination or application of one or more technologies; ii. it addresses a speci c mar-
ket need; and iii. it generates pro ts, meaning the investment involved yielded positive returns
[Escorsa, 1998].

                                                4
2.2     The ‘National System of Innovation’ Approach
In the eld of economics, the NSI concept gained attention with the arrival of the-
ories which highlighted the role of technology, such as the EGT. The NSI approach
has received researchers’ interest due to its focus on the endogenous building of ca-
pabilities for development and also because it provides a speci c role for government
policy towards the technological catch-up process [Gancia and Zilibotti, 2005].
   The connection between the NSI approach and economic growth is still rather
unexplored. This is mainly explained by two reasons. First of all, the measure-
ment of innovation systems for practical purposes has been a matter of debate
[Holbrook, 2006]. This issue is important because this dissertation aims to focus
in quantitative rather than qualitative analysis; the measurement problem, along
with its present resolution, will be discussed and illustrated with empirical studies
later on. Secondly, the NSI approach has been found insu cient to explain growth
by itself, which is why this concept has been mainly used to study industrial de-
velopment dynamics qualitatively [Lundvall, 2007]. For this limitation, it will be
described below how the KE Framework provides a robust platform in order to link
the NSI concept to economic growth.
   The concept of a NSI began to be developed more than 20 years ago. Since
then, it has been employed to analyze industrially advanced countries such as Nor-
way, Sweden and Australia; to explain successful development case-studies in Asia:
Japan, South Korea, Taiwan, China and India; and likewise in Latin America: Brazil
and Chile [Feinson, 2003]. Among this diversity of studies, the de nition of a NSI
presented below is commonly seen. It corresponds to the rst introduction of the
term in print by Christopher Freeman2 , in his book about the favorable outcome of
Japanese technological and economic policy in the 1960s and 1970s [Edquist, 1997].

       De nition of National System of Innovation:
          “The network of institutions in the public and private sectors whose
       activities and interactions initiate, import, modify and di use new tech-
       nologies.”
                                                             Source: [Freeman, 1987].


   2
    According to Freeman, the rst person he heard use the expression of ‘National System of
Innovation’ was Bengt-Åke Lundvall.

                                            5
Although being increasingly popular within scholars and policymakers, the NSI
approach has experienced complications since its birth. Generally, theories come
within a speci c eld of science i.e. addressed and built by academics and scien-
tists from a particular discipline which, to a certain extent, share similar interests,
methodologies, terminology, etc. This has certainly not been the case for the NSI
approach. Attempts to de ne, describe and explain this concept have ascended from
a variety of elds, such as engineering, economics and management. Furthermore,
the study of NSI is not only decentralized in terms of discipline, but also geograph-
ically: pertaining books and papers are being published worldwide.
   However, the exibility of the term can be seen as an advantage, as it allows
scholars to adapt the analytical tool for studying di erent contexts. The rst study
by Freeman mentioned above, which began popularizing the concept, was con-
ducted to understand the successful development experience of Japan. Ever since,
the NSI approach has been widely used to understand how knowledge is produced
and applied in industrially advanced countries and how developing economies catch-
up in this process. Some exemplary studies are depicted below.
   [Freeman, 1995] reports how, since the 1970s, empirical evidence has begun to
be gathered regarding R&D investment and innovation, particularly in in Japan, the
United States and Europe. The data permitted to demonstrate how the success of
innovation depends on R&D expenditure. Furthermore, not only the links between
 rms were found to have a critical importance, but also the external relationships
with other types of institutions such as universities. In the 1950s and 1960s, the
Japanese success was simplistically endorsed to product imitation and the importa-
tion of foreign technology. However, when Japan’s exports started outperforming
those of the United States, this explanation turned inappropriate. This overcom-
ing is associated with relatively higher levels of industrial R&D spending in Japan.
Nevertheless, this factor does not o er a su cient explanation. The Soviet Union
and other Eastern European countries proved that dedicating R&D resources with-
out any elaboration did not guarantee innovation, di usion and productivity gains.
This elaboration refers to the linkages within the innovation system i.e. the impact
of technical innovations in society depended on how suitable these are for domestic
business, combined with the e orts devoted by rms to adopt them.
   A more recent NSI case-study is Korea. [Feinson, 2003] states that this country’s
experience displays the bene ts of dynamic, responsive science policies towards the

                                          6
technological catch-up process. By articulating the NSI agents, the Korean govern-
ment was able to drive the transition from a subsistence farming economy to one
in which technology is acquired, di used throughout the nation and employed in
favor of innovation. Korea’s rst stage was to promote technology in ows. For this
purpose, the traditional path of promoting FDI and licensing was not followed. Al-
ternatively, policy at this stage concentrated on establishing turnkey businesses.
The steel, paper, chemical and cement industries were all founded in the form
of turnkey factories, which were domestically expanded afterwards. Rather than
fostering licensing, policymakers opted for the promotion of import capital goods
which embody technology; the importation of this kind of goods may have been the
most productive method of technology transfer. At that time, Korea probably relied
more on this channel than any other newly industrialized country. The second stage
was to assimilate the imported technology into its domestic production lines. This
was addressed by public funding and a series of incentives towards R&D, including
tax breaks and exemption from military services for key personnel. The third stage
consisted in an outward orientation in the form of export promotion policies. These
included the liberalization of access to imported intermediate products, the facili-
tation of banking loans for working capital destined to export-related investments
and the elimination of restrictions to foreign capital.
   The NSI approach has also been used to evaluate STI policies in countries of the
Latin American region. For example, [Holm-Nielsen and Agapitova, 2002] studied
how Chile has increased its competitiveness due to a favorable macroeconomic en-
vironment for STI. However, in this country, research institutions have remained
rather disjointed from the productive sector, which wastes potential for continu-
ous product innovation and hinders increments of living standards. The analysis
suggests that Chile needs to make its NSI more e ective in two main ways: by
strengthening the venture capital market and introducing more measures for pro-
moting networking and cooperation between science and industry. This achieve-
ments are deemed to increase the returns to R&D investments.
   The unconstrained aspect of the de nition for NSI has allowed a wide variety
of analyses. However, this exibility brings complications because just as a NSI
diverges between countries and regions, so does the approach from their analysts
[Lundvall, 2007]. In this ambivalence and lack of consensus relies the criticisms
of the NSI concept. In natural sciences, agreeing in strict de nitions is seemed

                                           7
as crucial to allow scienti c progress. Modern economics is characterized for the
emulation of this rigidity. This might be the reason why the NSI approach has
slowly penetrated economic theory. The components, attributes and relationships
that compose a NSI are tremendously di cult to quantify because the level of anal-
ysis corresponds to that of an entire nation. To overcome for this limitation, the
quanti cation of National Systems of Innovation has been focused not in its com-
ponents but in the overall performance of the system [Carlsson et al., 2002]. Data
recording has recently begun in this respect, across a wide range of countries. As
follows, concrete indicators regarding manufactures imports and exports, FDI, R&D
expenditure can be used to measure how a given NSI is performing. This quantita-
tive approach is undertaken in the KE Framework and employed here, as explained
in the forthcoming Section 2.3. However, it is pertinent to revise before how previ-
ous works have pursued to identify statistical evidence for a relationship between
NSI performance and economic growth. These studies re ect various important
lessons which are recurrently taken into account in this dissertation.


2.2.1 Empirical Studies of NSI
[Freeman, 2002] discusses the relevance of innovation systems for economic growth
over the last two centuries. Based on this timespan, the study uses gathered indi-
cators which describe NSI performance and aims to identify to which extent their
variations resulted in faster or slower rates of growth. While labor and capital pro-
ductivity are employed for the rst century, the second is analyzed using more spe-
cialized indicators such as manufactures exports, information and communication
technologies expenditure and R&D personnel, among others. The analysis is rather
limited by data availability. However, it argues that these indicators show a clear
pattern of how, after the industrial revolution, the NSI approach can be employed
quantitatively to understand the divergence in paths of economic growth.
   [Rodríguez-Pose and Crescenzi, 2006] analyzes the link between R&D invest-
ment and patents with economic growth. The study focuses on the connection be-
tween the e ciency of innovation systems and the geographical di usion of knowl-
edge spillovers. With an econometric analysis, that includes only European coun-
tries, the work highlights two main results. First, the interaction between research
and socio-economic institutions determined the potential for maximizing the capac-


                                          8
ity to innovate. Secondly, proximity had an important role in allowing the di usion
of economically productive knowledge and its impact in overall growth.
   [Krammer, 2008] executes a cross-country analysis for Eastern Europe to explore
what enables countries to innovate more than others at a national level. As a proxy
for innovation, Krammer uses the number of international patents granted by the
US patent o ce. His results suggest R&D commitments and the ‘innovative tradi-
tion’ were key for increasing the knowledge stock. Openness and the protection of
intellectual property rights determined higher international patenting, while struc-
tural industrial distortions had a negative in uence in the quantity of patents.
   Lastly, [Fagerberg and Srholec, 2007] uses a broader set of data for studying the
role of innovation within a set of capabilities for development. The work bases its
methodology on a factor analysis that compromises 25 indicators and 115 countries
from 1992 until 2004. By this method, four types of capabilities were identi ed: the
development of the innovation system, the quality of governance, the character of
the political system and the degree of openness. Of these, the innovation system
and governance were found to be of noteworthy pertinence for economic growth.
As it will be shown later in Chapter 3, this dissertation will particularly build upon
this study. In contrast, however, di erent indicators are selected, arranged and
statistically analyzed here, using the KE Framework. This method of analysis points
to similar conclusions to those of [Fagerberg and Srholec, 2007].
   In synthesis, the trajectory of the NSI approach has been more qualitative than
quantitative. This approach aims to explain the dynamics of industrialization, tech-
nological catch-up and development. Empirical analysis has been made di cult due
to the problems for measuring innovation systems. However, researchers have pro-
posed performance as a plausible scale of reference, which has allowed some studies
to emerge. Still, there is a need for more quantitative research to validate the NSI
concept. This dissertation adds to these e orts by turning to the KE Framework,
described in the following section.


2.3    Leveraging on the Knowledge Economy Framework
This section describes how the KE Framework integrates the NSI approach, comple-
menting it in order to allow a robust analysis of its role towards economic growth.


                                          9
For this purpose, this section includes relevant excerpts from the book Building
Knowledge Economies: Advanced Strategies for Development [World Bank, 2009a];
in this publication, the Wold Bank compiled its work on the KE Framework.
   The W        B     has addressed the KE Framework through its Knowledge for
Development (K4D) Program. This program contributes to the framework by pro-
ducing publications and distributing those of third-party specialists in this eld of
study. Its aim is to promote the framework’s awareness among policymakers world-
wide. Through the KE literature and the e orts for data compilation from the K4D
Program, the W        B      has constructed the KE Framework to analyze, study
and devise policy recommendations for knowledge-driven economic growth.

     The Knowledge Economy Framework...
         “...describes how an economy relies on knowledge as the key engine
     for growth. It is an economy in which knowledge is acquired, created,
     disseminated and applied to enhance economic development.”
                                                     Source: [World Bank, 2009a].

   The KE Framework, as depicted in its name, highlights the increasingly protag-
onist role knowledge has in an economy. Countries worldwide, both industrially
advanced and developing, have been recognizing know-how and expertise as criti-
cal as other economic resources. Industrial production requires appropriate policies
that re ect the current interconnected and globalized economic context.
   According to [World Bank, 2009a], the KE Framework rests on four pillars: the
Economic Regime, the Innovation System, Education and Information and Com-
munication Technologies. These have been previously supported as foremost for
economic development by an ample literature and empirical works. Their de nition
and pertinence are described below.

Economic Regime. It is de ned as the set of economic and institutional incentives
     designed to promote an environment that permits knowledge creation, assim-
     ilation and di usion. This pillar covers a broad set of macroeconomic issues
     and policies, such as trade, nance and banking and governance. Due to this
     broadness, this pillar is sometimes divided into two sub-pillars, Economic In-




                                        10
centives and Governance, to facilitate analysis3 . The former is related to how
        resources can mobilize within an economy, while the later deals with how
        the political circumstances and its stability provide an appropriate business
        climate.
        A favorable Economic Regime is required to obtain better policy results from
        the other, more functional pillars. Industrially advanced countries generally
        have solid institutions based on democracy and free markets. Governments
        promote the development of their institutional regimes by improving labor
        and nancial markets, and by strengthening governance (e.g. increasing the
        enforcement of contracts and controlling corruption).

Innovation System. It consists of rms, research centers, universities, think tanks
        and other institutions within a given country, that import or produce knowl-
        edge and adapt technologies to the local context4 . STI activities require pub-
        lic support in an ample range of ways such as the funding of basic research
        and the facilitation of knowledge di usion. The latter is of particular impor-
        tance for developing countries, where knowledge and technology, the inputs
        for innovation, arrive from abroad in the form of FDI and manufactures im-
        ports, among other channels. Indigenous knowledge capabilities should also
        receive attention. The importance of this pillar relies on the empowerment
        for achieving desired social and economic outcomes through the application
        of knowledge.

Education. This pillar is related to the human skills required for the acquisition
        and exercise of knowledge. The preparation of the labor force includes the
        primary, secondary and tertiary levels of education, vocational training and
        continuous learning. The focus on a given level of education depends on the
        country’s stage in economic development. A member from the Least Devel-
        oped Countries group should give more attention to primary education, as
        literacy and arithmetic skills are required before more advanced competences
        are gained. As the country’s economy grows, the relevancy of continuous
    3
      This sub-pillar distinction will be taken into account in this dissertation. In Chapter 4 the re-
sults will show that it is important to analyze both Economic Incentives and Governance separately
than as a whole.
    4
      W       B     ’s nomenclature omits the word ‘national.’ However, by comparing the de ni-
tions of ‘Innovation System’ and NSI, it can be a rmed that both terms are concurrent.

                                                  11
learning increases, as this type of education is necessary for innovation re-
       sulting from the constant adaptation of knowledge. Education creates jobs,
       reduces poverty levels and increases empowerment. It is a fundamental pillar
       for the KE.

Information and Communication Technologies (ICT). ICT encompass the types
       of technologies that enable the di usion of knowledge. ICT, bearing tele-
       phone, television, radio and Internet networks, are critical for the economies
       of today, based on globalization and information. These reduce transaction
       costs signi cantly by providing accessibility to knowledge. A strong ICT pillar
       allows rapid and reliable exchange of information within a country and across
       its borders. Recent advances are a ecting how knowledge is acquired, created,
       shared and applied, which has positively impacted manufacturing, trade, gov-
       ernance and education activities, among others. Regarding this pillar, policies
       consider telecommunication legislation, along with the investment required
       for building and capitalizing ICT through the socio-economic dimension.


2.4     Research Hypothesis
The problem for the measurement of innovation systems and the lack of a robust
framework, mentioned in Section 2.2, must be solved for studying the connection
between NSI theory and economic growth. These issues are both addressed by the
KE Framework i.e. by adopting the accountability of performance, as the achieve-
ments of a given NSI are analyzed using indicators such as R&D expenditure and
high-technology exports. The KE Framework also integrates NSI theory with the
other three5 pillars mentioned above, which allows to study economic development
from a knowledge-based perspective.
   With the concepts that have been revised up to this point, this dissertation’s
research hypothesis is structured below.




   5
     Four (a total of ve KE Pillars), if Governance and Economic Incentives are considered as
separate pillars, as it would be the case later on in this dissertation.

                                             12
Research hypothesis:
          “A positive connection between NSI performance indicators and eco-
      nomic growth can be quantitatively found under the KE Framework.
      This connection reveals challenges and opportunities for a developing
      economy such as Colombia, along with STI policy recommendations.”

   The relevancy of this hypothesis lies in three main aspects. First of all, innova-
tion has been considered to be a central issue for EGT and in Economics in general.
Exploring the dynamics of technological progress using new metrics represents an
attractive contribution to the eld. Second, it would contribute to the e orts for the
empirical veri cation of the NSI approach reviewed above, by the means of a new
methodology. Thirdly, the validation of the hypothesis statement would favor the
position of the NSI concept as one of the centerpieces in the development process.
Understanding the role of National Systems of Innovation would nurture policy-
makers in the areas of STI. The subsequent pages aim to address this hypothesis. In
particular, the next chapter describes the gathering of data and its analysis, taking
into account the literature revised until this point.




                                           13
Chapter 3

Data and Methodology for Analysis

               , it was discussed how the KE provides the required framework for
P     understanding the role of a given NSI in its economy. Upon this background
and seeking to contribute towards a deeper apprehension of the NSI concept and its
validity, a research hypothesis was de ned. Aiming to link NSI performance with
economic growth, this chapter states and thoroughly describes the employed data,
and then declares the methodology for the respective statistical analysis.
   With this purpose in mind, the subject of the KE data is discussed at rst. The
source and the process of selection and recollection are concisely portrayed. After,
the issue of a dataset with an excessive number of variables is exposed. This is
resolved via a Principal-Component Factor Analysis, which groups the variables
of a given pillar in the construction of an associated index. With the constructed
indices, the OLS regression models are stated for the hypothesis’s testing.


3.1     Data
To investigate any e ect of NSI performance in economic growth, data was gathered
seeking to satisfy the following two criteria: a diversity that covers all the features
of the KE and the availability of observations for signi cant amount of time.
   Through the K4D Program, mentioned in Section 2.3, the W            B      classi es
a variety of pertinent statistics from the World Development Indicators (WDI) under
the four main KE Pillars [World Bank, 2010b]. The program’s dataset makes refer-
ence to more than 100 WDI variables. Data recollection began with a depuration
of this source, aiming to comply with the two criteria stated above. In particular,


                                          14
the process took into account the fact that several of the referred KE variables have
started being recorded, across a signi cant amount of countries, only until recently.
These variables were discarded, for the limited observations would not allow a sig-
ni cant timespan for analysis.
       The selection process yielded a total of 19 variables for 75 countries in the the
period [1998, 2007]. To balance the dataset and compensate for missing gures, the
timespan was divided into two 5-year intervals. These are: [1998, 2002] and [2003,
2007]. Observations were de ned as the average values of the variables for each of
these two periods. The summary statistics of these variables are displayed in Table
3.1.


3.1.1 Rationale for Variable Selection
Although most of the variables are already classi ed under the KE Framework by
the K4D Program, it is necessary to discuss why each is signi cant for the pillar it
represents. Starting with the Economic Regime Pillar, there is market capitalization
of listed companies, domestic credit provided by the banking sector and domestic
credit to the private sector, all measured in % of GDP. The rst variable is de ned as
the sum of the product between share price and the number of shares outstanding,
for all companies listed in the country’s stock exchange. The second variable refers
to the totality of credits conceded to various sectors on a gross basis, excluding those
provided to the central government. The third includes nancial resources provided
to the private sector (e.g. loans, non-equity securities and trade credits). These three
variables have been employed in studies concerning nancial market development
and economic growth; although the importance of the former in the latter has been
a matter of debate, several studies have evidenced on a signi cantly positive e ect
[Levine, 1997];[Levine and Zervos, 1998].
       There are six variables for the Governance Pillar. Their de nitions are presented
as stipulated in [Kaufmann et al., 2009]. Voice and Accountability captures the per-
ceptions to which citizens from a given country are able to participate in elections,
along with freedom of expression, freedom of association, and free media. Political
Stability re ects perceptions of the probability that the government will be destabi-
lized or overthrown by unconstitutional or violent ways, including political violence
and terrorism. Government E ectiveness captures the perception of the quality of


                                            15
Table 3.1: Summary Statistics for the Selected KE Variables
                                                                             Obs [1998,2002]
                                 Obs   Mean    Std. Dev.   Min      Max      Obs [2003,2007]   ∗ 100


Economic Incentives (Values in % of GDP)
Market capitalization of
                                 225   53.21     60.05      0.07    434.31           49.78
listed companies
Domestic credit provided by
                                 362   57.66     53.63     -57.35   304.29           50.00
the banking sector
Domestic credit to the
                                 362   46.6      44.58      0.72    220.73           50.00
private sector


Governance (Indices)
Voice and Accountability         396   -0.03       1       -2.19     1.66            49.24
Political Stability              389   -0.06     0.98      -2.78     1.64            48.07
Government E ectiveness          397   -0.02       1       -2.15     2.26            49.12
Regulatory Quality               391   -0.04       1       -2.46     1.96            49.10
Rule of Law                      394   -0.05     0.99      -2.33     2.07            48.73
Control of Corruption            391   -0.02       1       -1.79     2.49            49.10

Innovation System
Manufactures        imports
                                 336   66.67     11.04     21.16    90.89            50.60
(% of merchandise imports)
High-technology     exports
                                 328   9.76      12.59       0      73.09            50.91
(% of manufactures exports)
Foreign direct investment,
                                 347   4.96      5.58      -6.58    39.35            49.86
net in ows (% of GDP)
Research and development
                                 200   0.87      0.92       0.01     4.47            52.50
expenditure (% of GDP)

Education
Public spending on educa-
                                 305   4.69      2.09       0.6     15.57            52.79
tion, total (% of GDP)
School             enrollment,
                                 355   72.54     31.06      5.93    156.48           49.86
secondary (% gross)
School enrollment, tertiary
                                 318   26.67     23.32      0.14    91.35            50.31
(% gross)

ICT (Values per 100 people)
Personal computers               366    11.9      17.3      0.01    84.69            49.45
Mobile phones and landlines      398   51.03     49.19      0.17    186.37           50.25
Internet users                   393   14.67     18.98        0     81.21            49.62

                                               Source: calculations based on [World Bank, 2009b].




                                               16
public services, the civil service and the extent of its independence from political
pressures, along with the credibility towards the government’s formulation and im-
plementation of policies. The Regulatory Quality indicator perceives the ability of
the government to devise and carry out robust policies and regulations that allow
and foster the development of the private sector. Rule of Law captures the impres-
sion on how socio-economic agents have con dence in and abide to the rules of
society i.e. speci cally, the quality of contract enforcement, property rights, the po-
lice and the courts, as well as the protection from crime and violence. Lastly, the
Control of Corruption captures perceptions of the extent to which public power is
safeguarded from rent-seekers, considering all levels of corruption, and the seclu-
sion of the State from private interests.
   Over the last decade, governance has been a central topic of growth promotion
policies, especially in developing countries. According to [Gray, 2007], the most
prevalent approach in governance policy-making is known as the ‘good governance’
agenda, which contains the six variables previously mentioned. Representatives of
this agenda highlight its importance not only in the satisfaction of citizens’ aspira-
tions regarding public institutions, but also as a means to foster economic growth
and as a sustainable mechanism to reduce poverty. While the link between institu-
tions and growth was a central matter of classical economics, the notion of ‘good
governance’ had its grounds laid only until the 1970s and 1980s. The creation of
quantitative measurements has been key to structure a consensus of the positive
relationship between governance and economic growth.
   Regarding the (National) Innovation System Pillar, manufactures imports (%
of merchandise imports, foreign direct investment (net in ows, % of GDP), high-
technology exports (% of merchandise exports) and research and development ex-
penditure (% of GDP) were selected as representative variables under the NSI per-
formance approach. The case-studies included in Section 2.2 show that the rst
two variables are pertinent channels of foreign technology transfer for the catch-
up process. Manufactures imports incorporate foreign technology and represent
intermediate capital goods necessary for producing value added exports. FDI is rel-
evant as a source of capital for export promotion albeit does not necessarily ow
into sectors intensive in technology, which is why public policy is sometimes em-
ployed to foster investments that imply technology transfer. The cited case-studies
also show that other two variables of the Innovation System Pillar measure the in-

                                            17
digenous appropriation of knowledge i.e. these are related to how the mentioned
technology transfer channels are being utilized in the economy for producing inno-
vations and towards the promotion of technological capabilities. High-technology
exports account for this explicitly, as it refers to domestic production which em-
ploys indigenously developed or adapted technology. Regarding R&D expenditure,
the referenced authors of NSI studies recognize it as decisive in the adaptation of
foreign technology to the local context. By understanding the signi cance of these
four indicators, this dissertation seeks to elaborate on previous empirical research
and explore the link between NSI performance and economic growth.
   Before proceeding with the remaining KE Pillars, it is important to acknowledge
that, similar to other quantitative studies, the variables selected here emphasize
formal modes of learning and innovation based in science and technology activi-
ties [Lundvall, 2007]. This emphasis is re ected here in the selection of indicators
of R&D and capital-embedded industrial goods. However, innovation strategies
based on experience and the “doing, using and interacting” learning mode are rather
overlooked. This is explained by the lack of standardized variables to represent
experience-based innovations.
   In the case of the Education Pillar, three representative variables were selected.
Public spending on education (% of GDP) adds up expenditure on education of
public authorities at all levels, along with the subsidies to private education at the
primary, secondary, and tertiary levels. Secondary school enrollment (% gross) and
tertiary school enrollment (% gross) are the ratio of total enrollment, regardless of
age, to the population of the age group that o cially corresponds to the level of
education. According to [World Bank, 2009a], secondary education completes the
provision of basic education that began at the primary level and lays the founda-
tion for future learning. It yields both individual and social returns and provides an
important amount of human capital required for countries’ economic growth. The
role of tertiary education is crucial. Universities and research institutions have to
address the call for creating a pool of experts capable of acquiring science and tech-
nology and adapting it to the domestic context. Regarding the link between educa-
tion and economic growth, [Teles and Andrade, 2004] states that while the evidence
has been asymmetrical it mainly points towards a positive causal relationship. The
same study identi ed a positive relation between public spending on education and
economic growth. The reported signi cance of the relationship, however, varied

                                         18
depending on the composition of governmental spending between basic and higher
education i.e. it lost its signi cance when the latter was not promoted.
   For the remaining Pillar, ICT, the variables measured per 100 people are: per-
sonal computers, mobile phones and landlines and Internet users. As reported
by [Batchelor et al., 2005], previous studies agree on how ICT can help develop-
ing countries address a wide range of socioeconomic activities: the use of ITC en-
hances the production of goods and the provision of services and thus increases
productivity. There is less agreement, however, on how much a priority it should
be to promote the increase of ICT infrastructure. These technologies are increas-
ingly being seen as means to other development requirements rather than as an end
themselves. Policies associated with this KE Pillar have been focused on alleviating
the wide disparities in access; the poor is the part of society most out of reach from
ICT.


3.2    The Construction of KE Pillar Indices
Even after depuration, the dataset from Table 3.1 is still composed of too many
variables for an econometric analysis. [Fagerberg and Srholec, 2007] faced a sim-
ilar problem in its attempt to explore the relationship between NSI and economic
growth. To face a dataset with numerous variables, the work employed a Principal-
Component Factor (PCF) Analysis. As described in the study, this process is based
on the idea that variables from the same category are likely to be signi cantly cor-
related and thus can be reduced into a smaller number of indicators, which re ect
the variance dimension of the data. The PCF Analysis assigns speci c “loadings”
which weigh in the calculation of the factor score for each country. Countries re-
ceive scores for each of its KE Pillars by adding the product of the pillar variables’
values and the corresponding coe cients, which are derived from the loadings.
   As it was mentioned in Chapter 2, [Fagerberg and Srholec, 2007] identi ed four
factors from a set of variables: the innovation system, governance, the political sys-
tem, and openness. In contrast, this dissertation uses a di erent set of representative
variables for the KE Pillars. Using the K4D Program’s classi cation, individual PCF
Analyses were carried out for each of the pillars. This ensured that the result-
ing indicators kept the structure suggested by the KE Framework. Consequently,


                                          19
the indicators for each of the pillars contain information only from their respective
variables. The Economic Regime Pillar is divided into two sub-pillars, as suggested
by the W            B     : Economic Incentives and Governance [World Bank, 2009a].
   The PCF Analysis executed here successfully identi ed an underlying structure
for each of the KE Pillars and generated proxy indices. For every sampled country,
an associated index (value) was calculated. The resulting set of KE Pillar Indices
can be viewed in Appendix A.
   The PCF Analysis maximizes the amount of overall variance (accounting by all
the variables as a group) that is to be captured by the index. The correlations in
Table 3.2 indicate the “relevance” each variable has in its corresponding index. For
example, in the case of Education, both secondary and tertiary school enrollment
have a higher weigh in the calculation of the associated index values, compared with
public spending variable. The constructed Education Index is more correlated with
the rst two variables because the two enrollment indicators are more correlated
with each other in comparison to the scal indicator1 .
   The variance explained by the Innovation System Index is 40.18%. While this
value is rather low, it is still signi cant on what is considered best practices for PCF
Analysis, as stated in [Costello and Osborn, 2005]. Furthermore, correlations in this
index are all above the 32% recommended borderline. The values show that, in this
constructed index, the more relevant variable is high-technology exports, followed
by manufactures imports, R&D expenditure and, lastly, FDI in ows.
   It is critical to recognize that this particular Innovation System Index is designed
to be functional only in the context of the KE and thus should not be employed in-
dependently. There are other innovation indices more suitable for a comparative
analysis or ranking purposes; a prominent one is the National Innovative Capac-
ity (NIC) Index used in the Global Competitiveness Report [Porter and Stern, 2002].
Stand-alone innovation indices are, however, not suitable for this dissertation as
these consider a series of factors that are better classi ed in other KE Pillars e.g.
in the case of the NIC index: venture capital availability (Economic Incentives),
the quality of institutions (Governance), human capital (Education), and the social
penetration of information and communications infrastructure (ICT). The KE ap-
proach allows the separation of these factors and thus an independent analysis of

   1
       For more details on PCF Analysis, please refer to [Smith, 2002].

                                                  20
Table 3.2: The Constitution of the KE Pillar Indices

        Economic Incentives Index                            Education Index
        Variance Explained: 81.12%        Correlation        Variance Explained: 63.38%      Correlation
        Market capitalization of listed                      Public spending on education,
                                             0.79                                               0.45
        companies                                            total
        Domestic credit provided by
                                             0.94            School enrollment, secondary       0.93
        the banking sector
        Domestic credit to the private
                                             0.97            School enrollment, tertiary        0.91
        sector

        Governance Index                                     ICT Index
        Variance Explained: 87.84%        Correlation        Variance Explained: 91.09%      Correlation
        Voice and Accountability             0.89            Personal computers                 0.95
        Political Stability                  0.86            Mobile phones and landlines        0.94
        Government E ectiveness              0.97            Internet users                     0.97
        Regulatory Quality                   0.95
        Rule of Law                          0.98
        Control of Corruption                0.96

        Innovation System Index
        Variance Explained: 40.18%        Correlation
        Manufactures imports                 0.69
        High-technology exports              0.76
        Foreign direct investment,
                                             0.35
        net in ows
        Research and development
                                             0.65
        expenditure


NSI performance.
   To compare the NIC Index with the Innovation System Index, developed in this
dissertation, the correlation was calculated between the 2001 value of the former2
and the average for the years [1998, 2002] of the latter3 . A correlation of 65% sug-
gests that the native Innovation System Index captures a considerable portion of the
NIC dataset, while some of the remaining percentage is likely to be balanced with
the information included in the other pillars. As mentioned earlier when elaborat-
ing on its variables, the Innovation System Pillar Index should be interpreted as an
attempt to represent the general impression of how a given NSI performs through
its development. The indicator intends to re ect the several case-studies presented
in Section 2.2.

   2
       Based on data from [Porter and Stern, 2002].
   3
       Using the values of Appendix A.

                                                        21
3.3    Speci cation of Models
The KE Pillar Indices, which contain the information of the variables in Table 3.1,
can now be used as proxy variables in an econometric analysis for testing the hy-
pothesis devised in Chapter 2. The summary statistics of the variables to be included
in the upcoming regressions are presented in the following Table 3.3.

            Table 3.3: Summary Statistics for the Regression Variables

                                           Obs    Mean   Std. Dev.   Min     Max


           Explained Variable
           Annual GDP Growth (%)           134    1.27     0.67      -1.90   2.57

           Proxy Variables
           (KE Pillar Indices)
           Economic Incentives             134     0        1        -1.25   2.85
           Governance                      134     0        1        -1.88   1.65
           Innovation System               134     0        1        -2.05   2.26
           Education                       134     0        1        -2.95   2.35
           ICT                             134     0        1        -1.27   2.48

           Control Variables
           GDP per Capita
                                           134    8.60     1.35      5.53    10.61
           (constant 2000 US $)

           Dummies (Binary Variables)
           [1998, 2002] Observation        134    0.52     0.50       0       1
           Sub-Saharan Africa              134    0.07     0.26       0       1
           Latin America & the Caribbean   134    0.15     0.36       0       1
           East Asia & Paci c              134    0.06     0.24       0       1
           Middle East & North Africa      134    0.05     0.22       0       1
           South Asia                      134    0.01     0.12       0       1
           Europe & Central Asia           134    0.17     0.38       0       1


   The logarithm of GDP growth is selected as the explained variable and the ve
KE Pillar Indices are included as explanatory proxy variables. Furthermore, eight
control variables are included in the analysis. Six of them are regional dummies
(binary variables), which group developing countries according to W                  B   ’s
geographic classi cation. These are: Sub-Saharan Africa, Latin America & the
Caribbean, East Asia & Paci c, Middle East & North Africa, South Asia and Eu-
rope & Central Asia [World Bank, 2010a]. The null case of the regional dummy

                                             22
variables corresponds to high-income countries. The mean of the regional dum-
mies represent their share in the dataset (e.g. 15% of the considered countries are
from Latin America & the Caribbean). Adding all the means result in the propor-
tion of developing countries in the [1998, 2007] sample: 51%. Thus, the remaining
49% of observations conform the sampled high-income countries. Another control
variable is the logarithm of GDP per capita, to consider conditional convergence4 .
Lastly, there is one more dummy that indicates if the data-point corresponds to a
[1998, 2003] observation, to consider time e ects i.e. temporal variations that are not
captured by the KE Pillars and the other control variables.
    The econometric analysis is undertaken in a set of four models. These models are
represented in Table 3.4. All of these have the logarithm of GDP growth lngdpgi as
the explained variable and have the KE Pillar Indices as proxy variables, a constant
and an error term, represented as Pki , C and                     i,   respectively. The basic model
includes only the ve proxies, while the subsequent models incorporate the control
variables progressively.

                    Table 3.4: Econometric Analysis: Model De nitions

               5                                                         6
 lngdpgi =          βki Pki   +   β6i f irsti   +   β7i lngdpli    +          β(k+7)i Rki   +   β14 C   +   i
              k=1                                                       k=1

  Model 1                                                                                                 
  Model 2                                                                                                
  Model 3                                                                                               
  Model 4                                                                                              


    The second model adds the f irsti dummy variable, which equals to one when
the observation corresponds to the [1998, 2002] period and zero otherwise, thus tak-
ing into account time e ects. The third augments the analysis with the variable
lngdpli (logarithm of GDP per capita) to consider the in uence of conditional con-
vergence. Finally, the fourth model adds the six regional dummy variables, written
as Rki , to check for the geographical particularities that might a ect growth and are
not captured by the other variables.
    4
     The theory of conditional convergence states that, given certain conditions, poorer countries
grow faster than their richer counterparts, until all economies reach the same level of GDP per
capita. Developing countries have the potential to increase their economic output levels at a faster
rate, due to the fact that the e ects of diminishing returns are not as consolidated as in higher
income countries.

                                                     23
To recapitulate, this chapter described how data was extracted using W
B      guidelines and databases, in order to robustly represent the KE Pillars of
75 countries worldwide for a ten-year period of analysis: [1998, 2007]. This time
interval was divided into two consecutive quinquennial periods: [1998, 2002] and
[2003, 2007]. For each of these, averages of available data were calculated. The
resulting variables were then employed for the construction of the associated KE
Pillar Indices using PCF Analysis. These indices are to be used as explanatory proxy
variables for GDP growth, along with eight control variables. The coe cients from
the models stipulated above are to be estimated through regressions. In the next
chapter, these results are displayed and analyzed in detail.




                                         24
Chapter 4

Results and Evaluation

          on the data and the methodology for its analysis, both presented in the
B     last chapter, this part of the dissertation seeks to exhibit the relevant outcomes
in the validation of the hypothesis: the existence of a statistical link between NSI
performance and economic growth under the KE Framework. This chapter is di-
vided into two parts. First, scatter plots are included for each of the constructed
KE Pillar Indices and GDP per capita. These give a rst view on how each index
is relevant within the economic activity. The second part focuses on giving out
information from the econometric analysis of the previously de ned models.


4.1    KE Pillar Indices vs GDP per Capita
Before revising the econometric analysis, it is appropriate to get an initial sense of
the roles the Innovation System Index and the other KE Pillar Indices have on eco-
nomic growth. For this purpose, the present section will use scatter plots. These
show a graph with the scores obtained by the 75 sampled countries on each pillar
(see Appendix A) on the x-axis and their respective log of GDP per capita on the
y-axis. To check for di erences between the two time periods, the dots are clas-
si ed accordingly. A tted line is included to illustrate the general trend of the
relationship between the index’s scores and the production levels.
   Regarding the main Pillar Index of concern, the Innovation System, Figure 4.1
displays a positive relationship with GDP per capita, with a correlation equivalent
to 0.6373. Countries that scored a better NSI performance i.e. a better acquisition,
production and absorption of applicable knowledge, at the same time experienced


                                          25
Figure 4.1: GDP per Capita and Innovation System




higher levels of income. The tted lines suggest that the relationship strengthened
between the rst period and the second.
   For the Economic Incentives Index, Figure 4.2 shows an apparently similar as-
sociation. The correlation is slightly stronger compared to the Innovation System
index: 0.6926. The gure depicts that countries which scored high in this index,
with an environment suitable for a better allocation of resources represented by
superior levels of domestic investment, had greater GDP per capita levels between
[1998, 2007]. The slope decreased from the former quinquennial period to the latter,
although not by much.


               Figure 4.2: GDP per Capita and Economic Incentives




                                         26
Figure 4.3: GDP per Capita and Governance




   The case of the Governance Index, shown in Figure 4.3, exhibits the largest cor-
relation among all pillars: 0.8714. Compared to the rest graphs, the Governance
 tted value lines are steepest. It shows that countries with better institutions si-
multaneously experienced superior positions of GDP per capita. This relationship
appears to have slightly decreased over the time of analysis.
   Figure 4.4 shows the scatter plot for the Education Index. This index has a cor-
relation of 0.7546 with GDP per capita. The slopes of the tted value lines remained
practically the same. The data from the analyzed time interval supports the idea
that richer nations have a more skilled labor force, which e ciently generates and


                    Figure 4.4: GDP per Capita and Education




                                         27
applies knowledge.
   Lastly, Figure 4.5 is the respective graph for the ICT Index. As the other KE Pil-
lars, it also displays a strong positive correlation with GDP per capita: 0.8392. The
slopes of the tted value lines, however, presented the most enunciated decrease
between [1998, 2002] and [2003, 2007]. This change might have a ected the econo-
metric analysis, as mentioned later on. Still, it can be said that better infrastructure
for the communication and di usion of information and knowledge is strongly re-
lated to higher income per capita levels.

                         Figure 4.5: GDP per Capita and ICT




   Although ones in greater measure than others, all the KE Pillar Indices display
a positive correlation with economic growth. The fact that the Innovation Systems
Index scores account for the lowest correlation with GDP per capita (albeit still
high), is noteworthy. In a broader perspective, these graphs support the notions of
the authors mentioned in Chapter 2: the NSI concept and the KE framework are
relevant towards economic output. The Innovation System’s performance, along
with the set of Economic Incentives, the level of Governance, Education and the
expansion of ICT are playing signi cant roles in the economies of today. For a
deeper understanding on how these economies grow in relation to the KE Pillars,
the following section presents and discusses the results of the econometric analysis
based on the models proposed in Chapter 3.




                                            28
4.2     Econometric Analysis
The KE Pillar Indices constructed in Section 3.2 are now employed for regressions
with economic growth as the explained variable, using the models de ned in Section
3.3. Table 4.1 contains the results from the regressions against the logarithm of GDP
growth. The table points out the estimated coe cients for each of the models. All
the estimations are based on the Ordinary Least Squares (OLS) method. However,
Model 1 and Model 3 were estimated using the ‘Huber-White Sandwich Estimator’
of variance in order to calculate robust standard errors, as evidence of heteroskedas-
ticity was found for these models1 . Columns (1) through (4) correspond to standard
OLS regressions, while (5) and (6) use a stepwise estimation to identify the speci ca-
tion with the best statistical properties2 . Column (5) begins the estimation process
with Model 3 and excludes the quartile of countries with lowest GDP per capita,
while (6) starts with Model 4 without any variation to the sample. It is necessary
to acknowledge that the stepwise regressions are included speci cally to provide
secondary evidence for the relationship between the Innovation System Index and
GDP growth. The resulting estimations (5) and (6), unlike (1) through (4), are not
intended to be representative for the KE Framework as the stepwise regressions
discarded some of the KE Pillar Indices.
    The possibility of endogeneity3 was addressed by the means of the Durbin-Wu-
Hausman Test, which is conformed by two steps. In the rst one, each potentially
endogenous proxy variable was regressed on all exogenous variables (the other
proxies), along with the variables that were used in the construction of the regressed
index. The resulting residuals are, in the second step, added to the new regression
of the original model [Wooldridge, 2002]. If any residual coe cient was to come
as signi cant in one of these latter regressions, endogeneity of the corresponding
proxy variable needs to be accepted and the associated model should be estimated
by two-stage least squares in order to achieve consistent results. The resulting co-

    1
       The results of White’s heteroskedasticity tests are included in Appendix B.2.
    2
       The stepwise estimation seeks to dismiss variables that do not provide explanatory power
to the model given a particular signi cance level, in this case 10%. The process begins with the
full model and checks whether the calculated p-value of a variable falls farther from the selected
frontier. It then excludes the most statistically meaningless variable and starts over. At each step
the procedure also inspects if a variable that was discarded earlier has become signi cant.
     3
       Endogeneity occurs when there is a causality loop between the explained and the explanatory
variables.

                                                29
Table 4.1: OLS Regression Results for GDP Growth.

                                              C
  V                            Standard Regressions                Stepwise Regressions
                       (1)         (2)          (3)          (4)      (5)          (6)
                    Model 1    Model 2†     Model 3    Model 4†    Model 3‡    Model 4†
  Economic          -0.213**    -0.194**     -0.150*      -0.125   -0.151**    -0.220***
  Incentives        (0.0852)    (0.0860)    (0.0781)    (0.0942)   (0.0700)     (0.0737)
                      -0.14       0.17       0.271**     0.280**    0.219*
  Governance
                     (0.119)     (0.127)     (0.127)     (0.141)    (0.113)
  Innovation          0.04       0.119*     0.139**    0.243***    0.178**     0.201**
    System          (0.0747)    (0.0683)    (0.0686)    (0.0847)   (0.0749)    (0.0876)
                      -0.02       0.01         0.05         0.01
  Education
                    (0.0100)    (0.0802)    (0.0911)    (0.0959)
      ICT             0.190     -0.306**      -0.20      -0.329*
                     (0.119)     (0.148)     (0.147)     (0.174)
  [1998, 2002]                 -0.771***   -0.717***   -0.819***   -0.500***   -0.544***
  Observation                    (0.165)     (0.143)     (0.173)    (0.108)     (0.105)
    log(GDP                                 -0.221**     -0.207*   -0.361***
   per capita)                              (0.0933)     (0.119)    (0.112)
  Sub Saharan                                              -0.08               0.440***
     Africa                                              (0.252)                (0.148)
 Latin America                                             -0.27
 the Caribbean                                          (0.212)
   East Asia                                               -0.65
     Paci c                                             (0.491)
  Middle East                                               0.21               0.552***
 North Africa                                           (0.219)                (0.162)
      South                                                 0.42               1.100***
       Asia                                              (0.309)                (0.236)
   Europe                                                  0.19               0.467**
  Central Asia                                           (0.209)                (0.181)
                    1.279***   1.684***    3.554***     3.518***   4.703***    1.404***
   Constant
                    (0.0568)   (0.0734)     (0.793)     (1.0950)    (0.999)    (0.0793)
 Observations          134        134         134           134       100         134
  R-squared           0.09       0.25         0.29          0.37      0.30       0.31
                           Note: Standard errors in parentheses
                          *** p  0.01, ** p  0.05, * p  0.1
                †
                  Using the Huber-White Sandwich Estimator of variance.
                      ‡
                        Excluding the quartile of poorest countries.




                                           30
e cients of this test are included in Appendix B. In this occasion, no evidence of
endogeneity was found.
   Straightforwardly, the most notable result is the recurrent signi cance of the
Innovation System proxy variable through the regressions, suggesting a positive
e ect from this particular Pillar Index upon GDP growth. Out of all the KE Pillars,
it displays the most signi cant evidence: in columns (2), (3), (4) (5) and (6) with
p-values lower than 10%, 5%, 1%, 5% and 5%, respectively. These results indicate
that countries with better NSI performance experienced greater GDP growth in the
period of analysis.
   This positive relationship is depicted in the following Figure 4.6. The graph
shows the marginal e ect of the Innovation System Index regressor on GDP growth,
after taking into account the associations between the other variables included in
the regression (column (4); Model 4). The slope of the tted line corresponds to the
Innovation System Index’s β calculated in the regression. As a support for the va-
lidity of this model’s particular speci cation, another plot is included as Figure 4.7.
The regression’s residuals do not display any apparent pattern with the Innovation
System Index, supporting the absence of endogeneity discussed earlier.

       Figure 4.6: Marginal E ect of the Innovation System Scores (Model 4)




   To analyze the relationship between Innovation System’s performance and GDP
growth, it is necessary to recall Table 3.2 (p. 21), which shows that the index is more
correlated to high-technology exports, manufactures imports and RD expenditure
(in that order) than FDI in ows. Although this structure should be considered as

                                          31
Figure 4.7: Regression’s Residuals vs Innovation System Scores (Model 4)




it provides insights about its constitution, the Innovation System Index, due to its
nature (calculated by PCF Analysis), must be appreciated as a whole. Determining
which of the indicators that belong to this index is more critical for growth falls out
of the scope of this approach.
   The Governance Pillar Index returned signi cant coe cients less consistently
as with the IS index: in columns (3), (4) and (5) with respective p-values lower
than 5%, 5% and 1%. Still, the unchanging positive sign supports the theory; ‘good
governance’ has been relevant for higher growth.
   The outcome of the coe cients for the ICT and Economic Incentives indicators
are more paradoxical, being both signi cantly negative in some iterations: (2) and
(4) for the former and the latter in all but (4). This is explained, to an extent, due
to the e ect of conditional convergence. The regression table portrays how intro-
ducing the variable log (GDP per capita) in (3) reduces the signi cance of both ICT
and Economic Incentives indicators, the latter losing all explanatory power. Coher-
ently, these two pillars are strongly correlated with GDP per capita, 0.84 and 0.69
respectively.
   Particularly for the ICT Index, it is worth to revisit Figure 4.5. The scatter plot
shows how the slope of the tted line became atter over time. This adjustment
is most likely explained by the characteristic of the ICT variables employed here.
‘Personal computers,’ ‘Internet access’ and ‘mobile phones and landlines’ are all
technologies that mature and continue to fall in price, thus become more accessible


                                          32
to poorer countries. As follows, a more dynamic ICT Index that contains this e ect
might be more suitable for the present approach.
    Education did not yield a signi cant result. There are two main possible expla-
nations for this. First, the calculation of the variance in ation factors4 for each of
the explanatory variables suggested the presence of multicollinearity. It does not
appear to be so severe, as the signs of the coe cients in Table 4.1 seldom changed.
However, it might have deterred the signi cance levels. The second issue is that
the Education Pillar Index variables represent current investment and enrollment,
which do not re ect so strongly in present growth, but rather have a more important
e ect in future increments of GDP.
    Regarding the control variables, the signi cance of the f irst dummy variable’s
negative coe cient indicates a strong and generalized trend of greater growth for
the years [2003, 2007] in comparison to [1998, 2002]. Also, although in some iter-
ations more so than others, there were meaningful negative coe cients for lngdpl,
which support the theory of conditional convergence. Finally, the regional dum-
mies had a more secondary role on the model. In column (4), although not yielding
signi cant coe cients, they eliminated the explanatory power of the Economic In-
centives proxy variable and reduced that of lngdpl. Thus, the standard regression of
Model 4 suggests that a portion of the negative e ect from the Economic Incentives
Pillar and the ‘conditional convergence’ observed in the previous columns is related
to regional particularities.
    Interestingly, the stepwise regression in column (6) traded lngdpl for the regional
dummy variables; this trade-o , however, did not produce a big change the signi -
cance and values of the other coe cients. For Latin America  the Caribbean and
East Asia  Paci c no signi cant coe cients were produced, this suggests that the
particularities of these regions are expressed by the proxy variables of the Inno-
vation System and the Economic Incentives pillars, the latter containing the e ect
of conditional convergence. Excluding the poorest 25% of countries in column (5)
reduced the explanatory power of the Governance Index, suggesting its importance

    4        ˆ
        vif (Bi ) =     1         2
                         where Ri corresponds to the R-squared of the OLS regression in which the
                      1−Ri ,
                          2

                                         ˆ
explanatory variable associated with Bi becomes the explained variable, as a function of all the
other explanatory variables of the original model. A large R-squared suggests a high goodness of
 t and so, in this case, multicollinearity in the original model. As the R-squared increases, so does
vif . The “rule of thumb” states that if the vif for a particular explanatory variable exceeds ve,
multicollinearity is present.

                                                 33
in the excluded countries.
   To check for the consistency of the constructed KE Pillar Indices, the economet-
ric analysis is replicated with GDP per capita as the explained variable. The set of
regressions is included in Appendix C. First of all, it is necessary to highlight the
well documented likelihood for the results with respect to GDP per capita to su er
from endogeneity. With this setup, it is di cult to know whether the KE Pillar In-
dices a ect the levels of GDP per capita, or if the relationship is opposite e.g. richer
countries can a ord better public education. In the previous regressions no evidence
of endogeneity was found i.e. the tests did not show that fast growth rendered better
performance of the pillars. Even though a causal relationship cannot be identi ed
in the regressions with GDP per capita as the dependent variable, these illustrate a
positive connection between all KE Pillars.
   In this chapter, the coe cients of models derived from the KE Framework were
estimated using the Pillar Indices constructed in Chapter 3. The Innovation System
Index, as a function of inward FDI, manufactured goods imports, high-technology
exports and RD expenditure, was found to have promoted economic growth dur-
ing [1998, 2007] in the 75 sampled countries. This relationship reached con dence
levels lower than 1% when taking into account time e ects, conditional conver-
gence and regional particularities. Furthermore, under this approach the ‘good gov-
ernance’ index displayed a similar positive e ect, albeit less prominently. These
results are concurrent to those of [Fagerberg and Srholec, 2007] which also calcu-
lated innovation system and governance indices, albeit using di erent variables,
complementary factors for economic growth (i.e. the political system and openness)
and other methodological di erences as to this dissertation. For the remaining KE
Pillars, the evidence was inconclusive.
   The constructed NSI performance index is an insightful proxy variable for ex-
plaining how economies have grown between. This evidence is pertinent for an
developing country like Colombia, whose policymakers have traditionally focused
KE e orts towards the Economic Incentives Pillar and neglected the consolidation
of the Innovation System Pillar. The following chapter explores the signi cance of
the exposed link, between NSI performance and GDP growth, for this particular
country.




                                          34
Chapter 5

An Overview of the Colombian National
System of Innovation

            point towards a positive e ect of NSI performance on economic growth.
R     What can can be inferred from this particular link, in favor of policy-making?
This chapter focuses on answering this question. Due to the fact that each NSI re-
lies heavily upon a particular context, a speci c country is chosen as a case-study.
Colombia’s NSI is to be analyzed using the considered theoretical background and
the gathered evidence. First of all, the general circumstances in which the system
began to be recognized as a public institution are described. Secondly, attention
is given to various studies which characterize the present state of Colombia’s NSI.
Thirdly, the current policy framework and instruments are described, in order to
grasp the system’s outlook. Finally, based on the revision of this case-study, recom-
mendations are provided in relation to the observed link between NSI performance
and economic growth.


5.1    Background
Since 1991, Colombia has decidedly shifted into a full market-oriented economic ap-
proach. The Import Substitution Industrialization model, commonly seen through-
out the region, was discarded in the mid-1970s and full edged liberalization was
gradually undertaken. In 1999 the country experienced a recession, caused by the
generalized capital out ows experienced across the developing world at that time,
which was aggravated in Colombia by an internal mortgage crisis. Although the
downturn worsened the country’s poverty gures, it was followed by a recovery

                                         35
stage i.e. the economy experienced accelerated growth between 2002 and 2007; for
this period, the increase of real GDP averaged 5.32%. This was mainly due to
more favorable economic conditions abroad and a series of policies that enhanced
the Colombian business climate, perhaps the most signi cant one being the sus-
tained progress in domestic security. Nevertheless, with the current global economic
downturn, the most recent gures are rather timid. Colombia’s economy grew only
2.53% in 2008 and the following year GDP growth was practically nonexistent.
   To enter the globalized economy, the Colombian State tore down tari barriers
and other protectionist measures. This new approach to international trade, while
rightfully seeking to improve domestic productivity levels, a ected many Colom-
bian rms which could not compete with the foreign companies that entered the
country. Policymakers, aware of this, have appointed export promotion mecha-
nisms, which mainly seek to improve the country’s level of competitiveness. The
current policy approach to competitiveness can be roughly understood by two doc-
uments written in the second half of the 2000s by the Consejo Nacional de Política
Económica y Social (C        ; National Council for Socio-economic Policy), a gov-
ernmental institution which provides the framework for Colombia’s development
policies. The following extracts are representative of their respective documents.

         “...[in Colombia] a series of measures and projects must be estab-
     lished and carried out in order to advance competitiveness in interna-
     tional markets. These measures may go from the construction and the
     improvement of the physical infrastructure or the training of the labor
     force, to the reorganization of institutions or the eliminations of [bureau-
     cratic] procedures. All these projects [...] seek to eliminate the obstacles
     faced by the productive sector during its operation...”
                               Source: translated from [Mincomercio and DNP, 2004].

         “A nation’s competitiveness is de ned as the degree to which a coun-
     try can produce goods and services capable of competing successfully in
     globalized markets and, at the same time, improve the population’s in-
     come conditions and the quality of life. Competitiveness is the result
     of the interaction of multiple factors related to the conditions faced by
     business which a ect their performance e.g. infrastructure, human re-
     sources, science and technology, institutions, the macroeconomic envi-
     ronment, and productivity.”
                               Source: translated from [Mincomercio and DNP, 2006].

                                         36
These two extracts from the policy documents show that, in Colombia, there
has been a bias towards exogenous mechanisms for increasing competitiveness i.e.
by infrastructure improvements (e.g. highways, access to utilities and airports) and
through the re nement of institutions (e.g. a more e cient bureaucracy, decreasing
corruption and a more e ective enforcement of legal contracts). In both documents,
the role of innovation is seldom mentioned and the respective policy guidelines are
practically absent. As follows, the background of Colombia’s approach to compet-
itiveness is, in terms of the KE Framework, focused in improvements of the Eco-
nomic Incentives, Governance, ICT and Education Pillars, leaving the Innovation
System Pillar as secondary at best.
   Contrary to other countries under the export-promotion scheme, Colombian pol-
icy has lagged in addressing the role innovation has as an endogenous mechanisms
that favors competitiveness. A domestically developed innovative product or ser-
vice will (by de nition) outperform its competition, while implicitly contributing
towards technological learning inside the rm and thus enhancing its productivity1 .
The fact that innovation has been belatedly adopted as a State policy has brought
upon several weaknesses in the Colombian NSI, as reported by the several studies
included in the following section.
   In Colombia, the institutional framework for a NSI was introduced in 1995. It
was preceded by the Sistema Nacional de Ciencia y Tecnología (SNCYT; National
Science and Technology System), conceived in 1990. The SNCYT aimed to integrate
a diversity of institutions which shared a common vision, mission and objectives. It
included rms, universities, public research institutions, research centers, and tech-
nological institutions, among other actors. The innovation system was born as a
sub-system of the SNCYT, at a similar time as in other Latin American countries.
Both systems are, in practical terms, considered as one: the Sistema Nacional de
Ciencia, Tecnología e Innovación (SNCTI; National Science, Technology and In-
novation System)2 [Monroy Varela, 2006]. To conserve the terminology employed
in this dissertation, the SNCTI will be referred as the Colombian NSI, henceforth
CNSI. The public organization in charge of devising and executing policies that aim
to structure and strengthen the system is C                  , an institution that seeks to

   1
     For a detailed description of this process see [Kim, 1997], pp. 88-90.
   2
     This is due to the fact that both systems are fundamentally composed by the same actors,
share common concepts, basic strategies and challenges.

                                             37
lead the generation an utilization of knowledge in favor of socio-economic develop-
ment in Colombia. C              de nes the CNSI as follows.

          “The National System of Science, Technology and Innovation is an
      open system composed by policies, strategies, programs, methodologies
      and mechanisms in favor of the management, promotion, nancing, pro-
      tection and di usion of scienti c investigation and technological inno-
      vation, as well as the public, private or mixed organizations which carry
      out or promote the development of scienti c, technology and innovation
      activities.”
                                         Source: translated from [Colciencias, 2010].

    According to [Monroy Varela, 2006];[Forero, 2000], the CNSI has faced di cul-
ties since its conception. The most conspicuous have been: an unstable and weak
budget, inadequate policy formulation (e.g. shortsightedness and vertical decision-
making), a lack of public directives, the social apathy towards the socio-economic
value of scienti c research, a stagnant scienti c community, the phenomenon re-
ferred as “brain drain,” and the disjunction among the system’s actors. The follow-
ing sub-section presents representative studies which diagnose the CNSI’s weak-
nesses.


5.2       Current Status
Several studies have faced the task of diagnosing the CNSI. Three in-depth sur-
veys are worth mentioning, as these reveal important aspects of the system. The
Encuesta de Desarrollo Tecnológico en el establecimiento industrial colombiano
(EDT; Technological Development Survey on the Colombian industrial settlement)
began to be recorded since 1996 through the e orts of Colombia’s Departamento
Nacional de Planeación (National Planning Department) and with the support of
C            . The introduction of an institutional NSI framework matched with
an increased interest in the measurement and tracking of the industry’s techno-
logical development. For this purpose, the rst EDT was conducted in 1996 and
these e orts were continued with iterations of the survey in 2003, 2005 and 2006.
However, the methodology has been constantly amended. While the survey meets
higher standards today and has a considerably larger sample size (6080 rms in 2006
[DANE, 2010] as to 885 in 1996 [DNP, 1997]), its more recent results are hard to

                                         38
be compared with those of 1996 [Vargas Pérez and Malaver Rodríguez, 2004]. One
telling indicator which has remained practically unchanged is the percentage of in-
novative rms3 : 11.3% (1996) and 11.8% (2003)4 .
   [Monroy Varela, 2006] is another survey-based study, which queried about the
articulation of the CNSI. It found that while the knowledge-producing agents of
the system knew about the existence of a CNSI framework, business owners were
mostly unaware of it. This feature suggests the system’s bias towards the supply-
side of knowledge, which is particularly problematic for the development of inno-
vations i.e. indigenous knowledge and technologies are not being incorporated into
Colombian products or services. The survey shows that this bias is re ected on the
disjunction of the system: agents primarily interact with others of the same type.
   The third survey reported in [Malaver Rodríguez and Vargas Pérez, 2006] was
carried out in 2005, aiming to advance the results of the EDT albeit delimited to Bo-
gotá (the capital city district) and its circumvent department of Cundinamarca (one
of Colombia’s thirty-two administrative divisions). It found that less than a third of
the 400 sampled rms in the aforementioned region interacted with other agents of
the CNSI in order to complement their e orts in producing innovations and improve
their technological capabilities. Furthermore, less than 10% of the companies turned
to the public institutions of the CNSI. However, the low share of rms that engaged
the system found few di culties in doing so, meaning that the problem lies in the
absence of associations rather in than the links themselves. Once again the cultural
problem is brought to focus, as 44% of the sampled rms considered unnecessary to
reach the CNSI, while 19% did not innovate because it was not considered pro table
or signi cant to do so.
   [Torres et al., 2007] gathered data through the Premio Innova (Innova Award), a
technological innovation competition for Colombian small and medium enterprises
(SMEs), created in 2004 as a mechanism to promote innovation in technological
start-ups5 . The work analyzes how the participating companies are seeking to in-
novate and typi es the interactions between the CNSI agents. The study is based

   3
       The survey classi es a rm as innovative if, at the corresponding period of inquiry, it
has launched at least one good or service signi cantly improved by international standards
[DANE, 2010].
     4
       As reported by [Vargas Pérez and Malaver Rodríguez, 2004] and [DANE, 2010].
     5
       The award aims to bring attention to Colombian innovations, benchmark technological capa-
bilities among di erent sectors, and to encourage an innovation culture in SMEs.

                                              39
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework
Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework

Contenu connexe

En vedette

Innovation Ecosystems at Internet Festival 2013
Innovation Ecosystems at Internet Festival 2013Innovation Ecosystems at Internet Festival 2013
Innovation Ecosystems at Internet Festival 2013Salvatore Iaconesi
 
Victor Mulas - Tech Innovation Ecosystems
Victor Mulas - Tech Innovation EcosystemsVictor Mulas - Tech Innovation Ecosystems
Victor Mulas - Tech Innovation Ecosystemsinnovationhubs
 
Smart Cities as Innovation Ecosystems sustained by the Future Internet - Land...
Smart Cities as Innovation Ecosystems sustained by the Future Internet - Land...Smart Cities as Innovation Ecosystems sustained by the Future Internet - Land...
Smart Cities as Innovation Ecosystems sustained by the Future Internet - Land...European Network of Living Labs (ENoLL)
 
European smart cities and smart city projects in user empowered innovation ec...
European smart cities and smart city projects in user empowered innovation ec...European smart cities and smart city projects in user empowered innovation ec...
European smart cities and smart city projects in user empowered innovation ec...European Network of Living Labs (ENoLL)
 
Keynote Markku Markkula - From Smart Cities to Pioneering Regional Innovation...
Keynote Markku Markkula - From Smart Cities to Pioneering Regional Innovation...Keynote Markku Markkula - From Smart Cities to Pioneering Regional Innovation...
Keynote Markku Markkula - From Smart Cities to Pioneering Regional Innovation...Mindtrek
 
Smart Cities Day 2 Urban Innovation
Smart Cities Day 2 Urban InnovationSmart Cities Day 2 Urban Innovation
Smart Cities Day 2 Urban Innovation4 All of Us
 

En vedette (6)

Innovation Ecosystems at Internet Festival 2013
Innovation Ecosystems at Internet Festival 2013Innovation Ecosystems at Internet Festival 2013
Innovation Ecosystems at Internet Festival 2013
 
Victor Mulas - Tech Innovation Ecosystems
Victor Mulas - Tech Innovation EcosystemsVictor Mulas - Tech Innovation Ecosystems
Victor Mulas - Tech Innovation Ecosystems
 
Smart Cities as Innovation Ecosystems sustained by the Future Internet - Land...
Smart Cities as Innovation Ecosystems sustained by the Future Internet - Land...Smart Cities as Innovation Ecosystems sustained by the Future Internet - Land...
Smart Cities as Innovation Ecosystems sustained by the Future Internet - Land...
 
European smart cities and smart city projects in user empowered innovation ec...
European smart cities and smart city projects in user empowered innovation ec...European smart cities and smart city projects in user empowered innovation ec...
European smart cities and smart city projects in user empowered innovation ec...
 
Keynote Markku Markkula - From Smart Cities to Pioneering Regional Innovation...
Keynote Markku Markkula - From Smart Cities to Pioneering Regional Innovation...Keynote Markku Markkula - From Smart Cities to Pioneering Regional Innovation...
Keynote Markku Markkula - From Smart Cities to Pioneering Regional Innovation...
 
Smart Cities Day 2 Urban Innovation
Smart Cities Day 2 Urban InnovationSmart Cities Day 2 Urban Innovation
Smart Cities Day 2 Urban Innovation
 

Similaire à Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework

Technical efficiency of Vietnam rice farms a stochastic frontier production a...
Technical efficiency of Vietnam rice farms a stochastic frontier production a...Technical efficiency of Vietnam rice farms a stochastic frontier production a...
Technical efficiency of Vietnam rice farms a stochastic frontier production a...jackjohn45
 
20090712 commodities in the if study undp exeuctive summarywith covers
20090712 commodities in the if study undp exeuctive summarywith covers20090712 commodities in the if study undp exeuctive summarywith covers
20090712 commodities in the if study undp exeuctive summarywith coversLichia Saner-Yiu
 
Digital business ecosystems
Digital business ecosystemsDigital business ecosystems
Digital business ecosystemsYam Montaña
 
Gerard-Daphne-6077589-Economics-Thesis
Gerard-Daphne-6077589-Economics-ThesisGerard-Daphne-6077589-Economics-Thesis
Gerard-Daphne-6077589-Economics-ThesisDaphne Gerard
 
NEWCOMB-BENFORD’S LAW APPLICATIONS TO ELECTORAL PROCESSES, BIOINFORMATICS, AN...
NEWCOMB-BENFORD’S LAW APPLICATIONS TO ELECTORAL PROCESSES, BIOINFORMATICS, AN...NEWCOMB-BENFORD’S LAW APPLICATIONS TO ELECTORAL PROCESSES, BIOINFORMATICS, AN...
NEWCOMB-BENFORD’S LAW APPLICATIONS TO ELECTORAL PROCESSES, BIOINFORMATICS, AN...David Torres
 
Monitoring And Evaluation For World Bank Agricultural Projects
Monitoring And Evaluation For  World Bank Agricultural  ProjectsMonitoring And Evaluation For  World Bank Agricultural  Projects
Monitoring And Evaluation For World Bank Agricultural ProjectsMalik Khalid Mehmood
 
Literature Review for Knowledge Exchange and Enterprise Network (KEEN) Resear...
Literature Review for Knowledge Exchange and Enterprise Network (KEEN) Resear...Literature Review for Knowledge Exchange and Enterprise Network (KEEN) Resear...
Literature Review for Knowledge Exchange and Enterprise Network (KEEN) Resear...University of Wolverhampton
 
Incr global renewable energy mkt share un report - beijing re-report
Incr global renewable energy mkt share   un report - beijing re-reportIncr global renewable energy mkt share   un report - beijing re-report
Incr global renewable energy mkt share un report - beijing re-reportJ Suresh -
 
Research
ResearchResearch
Researchestri
 

Similaire à Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework (20)

Thesis
ThesisThesis
Thesis
 
CASE Network Reports 74 - Assessing the Development Gap
CASE Network Reports 74 - Assessing the Development GapCASE Network Reports 74 - Assessing the Development Gap
CASE Network Reports 74 - Assessing the Development Gap
 
Cnr 74
Cnr 74Cnr 74
Cnr 74
 
Technical efficiency of Vietnam rice farms a stochastic frontier production a...
Technical efficiency of Vietnam rice farms a stochastic frontier production a...Technical efficiency of Vietnam rice farms a stochastic frontier production a...
Technical efficiency of Vietnam rice farms a stochastic frontier production a...
 
CASE Network Report 78 - Scenarios For Health Expenditure in Poland
CASE Network Report 78 - Scenarios For Health Expenditure in PolandCASE Network Report 78 - Scenarios For Health Expenditure in Poland
CASE Network Report 78 - Scenarios For Health Expenditure in Poland
 
Mekar
MekarMekar
Mekar
 
20090712 commodities in the if study undp exeuctive summarywith covers
20090712 commodities in the if study undp exeuctive summarywith covers20090712 commodities in the if study undp exeuctive summarywith covers
20090712 commodities in the if study undp exeuctive summarywith covers
 
Digital business ecosystems
Digital business ecosystemsDigital business ecosystems
Digital business ecosystems
 
tese
tesetese
tese
 
Bma
BmaBma
Bma
 
Gerard-Daphne-6077589-Economics-Thesis
Gerard-Daphne-6077589-Economics-ThesisGerard-Daphne-6077589-Economics-Thesis
Gerard-Daphne-6077589-Economics-Thesis
 
NEWCOMB-BENFORD’S LAW APPLICATIONS TO ELECTORAL PROCESSES, BIOINFORMATICS, AN...
NEWCOMB-BENFORD’S LAW APPLICATIONS TO ELECTORAL PROCESSES, BIOINFORMATICS, AN...NEWCOMB-BENFORD’S LAW APPLICATIONS TO ELECTORAL PROCESSES, BIOINFORMATICS, AN...
NEWCOMB-BENFORD’S LAW APPLICATIONS TO ELECTORAL PROCESSES, BIOINFORMATICS, AN...
 
CASE Network Studies and Analyses 284 - Differences in productivity and its d...
CASE Network Studies and Analyses 284 - Differences in productivity and its d...CASE Network Studies and Analyses 284 - Differences in productivity and its d...
CASE Network Studies and Analyses 284 - Differences in productivity and its d...
 
Monitoring And Evaluation For World Bank Agricultural Projects
Monitoring And Evaluation For  World Bank Agricultural  ProjectsMonitoring And Evaluation For  World Bank Agricultural  Projects
Monitoring And Evaluation For World Bank Agricultural Projects
 
Howe
HoweHowe
Howe
 
Literature Review for Knowledge Exchange and Enterprise Network (KEEN) Resear...
Literature Review for Knowledge Exchange and Enterprise Network (KEEN) Resear...Literature Review for Knowledge Exchange and Enterprise Network (KEEN) Resear...
Literature Review for Knowledge Exchange and Enterprise Network (KEEN) Resear...
 
Incr global renewable energy mkt share un report - beijing re-report
Incr global renewable energy mkt share   un report - beijing re-reportIncr global renewable energy mkt share   un report - beijing re-report
Incr global renewable energy mkt share un report - beijing re-report
 
tesis_doctorado_nacional_jlcb
tesis_doctorado_nacional_jlcbtesis_doctorado_nacional_jlcb
tesis_doctorado_nacional_jlcb
 
CASE Network Reports 81 - The Development Gap between the CIS and EU
CASE Network Reports 81 - The Development Gap between the CIS and EUCASE Network Reports 81 - The Development Gap between the CIS and EU
CASE Network Reports 81 - The Development Gap between the CIS and EU
 
Research
ResearchResearch
Research
 

Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework

  • 1. Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework With an overview of the Colombian case-study by Andrés Barreneche García B.Sc., University of the Andes (Colombia), 2008 A Dissertation Submitted in Partial Ful llment of the Requirements for the Degree of M A E Main Supervisor: Prof. Dr. Yoichi Koike Second Supervisor: Prof. Dr. Yongjin Park Ritsumeikan University Graduate School of Economics July 2010
  • 2. ii
  • 3. Linking National Systems of Innovation and Economic Growth under the Knowledge Economy Framework With an overview of the Colombian case-study by Andrés Barreneche García∗ B.Sc., University of the Andes (Colombia), 2008 Main Supervisor: Prof. Dr. Yoichi Koike Second Supervisor: Prof. Dr. Yongjin Park Abstract This dissertation applies the Knowledge Economy (KE) Framework developed by the W B as a means to assess the e ect of National System of Inno- vation (NSI) performance on economic growth. The KE approach integrates the NSI concept as one of the four factors deemed to enhance economic output in terms of knowledge creation, di usion and adaptation; these are: the Economic Regime, the Innovation System, Education and Information and Communica- tion Technologies. This framework is employed for an empirical study about the connection between KE variables and economic growth with a sample of 75 coun- tries (developed and developing) in the [1998, 2007] period. This work concludes that higher NSI performance, as a function of foreign technology transfer (man- ufactures imports and FDI) and knowledge appropriation (R&D expenditure and high-technology exports) variables, is conducive to superior increments of GDP. Furthermore, this dissertation advances the discussion of the Colombian case- study and diagnoses that the country has failed to harness opportunities of foreign technology transfer as a consequence of the disjunction between NSI actors. ∗ E-mail: barreneche@gmail.com iii
  • 4. “We have the good fortune to live in democracies, in which individuals can ght for their perception of what a better world might be like. We as academics have the good fortune to be further protected by our academic freedom. With freedom comes responsibility: the responsibility to use that freedom to do what we can to ensure that the world of the future be one in which there is not only greater economic prosperity, but also more social justice.” Joseph Stiglitz. Nobel Prize Lecture, December 8th, 2001. iv
  • 5. Acknowledgements I would like to thank: Professor Yoichi Koike, for his guidance, support, helpful comments, patience, and for giving me the opportunity to cultivate my ideas while keeping me on track. Ritsumeikan University: professors, sta members and colleagues, for contribut- ing towards a gratifying academic experience in Japan. The Government of Japan (MEXT), for funding me with a scholarship. Alejandro Hoyos Suárez, for his friendship and for providing me with thought- ful advice in the writing of this dissertation and through my studies in the master’s program. My family: my parents Juan José Barreneche Silva and Maria Cristina García de Barreneche and my brother Alejandro Barreneche García, for the uncondi- tional love they have provided me in spite of the thousands of kilometers that have separated us. Sebastián Perez Saaibi, for his encouragement and unquestionable companionship as a fellow Colombian expatriate. v
  • 6. Contents Abstract iii Acknowledgements v Table of Contents vi List of Tables viii List of Figures ix 1 Introduction 1 2 Background 3 2.1 Endogenous Growth and Innovation . . . . . . . . . . . . . . . . . . 3 2.2 The ‘National System of Innovation’ Approach . . . . . . . . . . . . 5 2.2.1 Empirical Studies of NSI . . . . . . . . . . . . . . . . . . . . . 8 2.3 Leveraging on the Knowledge Economy Framework . . . . . . . . . 9 2.4 Research Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3 Data and Methodology for Analysis 14 3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 3.1.1 Rationale for Variable Selection . . . . . . . . . . . . . . . . . 15 3.2 The Construction of KE Pillar Indices . . . . . . . . . . . . . . . . . . 19 3.3 Speci cation of Models . . . . . . . . . . . . . . . . . . . . . . . . . . 22 4 Results and Evaluation 25 4.1 KE Pillar Indices vs GDP per Capita . . . . . . . . . . . . . . . . . . . 25 4.2 Econometric Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 5 An Overview of the Colombian National System of Innovation 35 vi
  • 7. 5.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 5.2 Current Status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5.3 Prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 5.4 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 6 Conclusions 47 A Constructed KE Pillar Indices 50 B Statistical Tests 54 B.1 Endogeneity: Durbin-Wu Hausman Test . . . . . . . . . . . . . . . . 54 B.2 Heteroskedasticity: White Test . . . . . . . . . . . . . . . . . . . . . . 55 C Estimations with GDP per Capita as the Explained Variable 56 C.1 Speci cation of Models . . . . . . . . . . . . . . . . . . . . . . . . . . 56 C.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 Bibliography 58 vii
  • 8. List of Tables Table 3.1 Summary Statistics for the Selected KE Variables . . . . . . . . 16 Table 3.2 The Constitution of the KE Pillar Indices . . . . . . . . . . . . . 21 Table 3.3 Summary Statistics for the Regression Variables . . . . . . . . . 22 Table 3.4 Econometric Analysis: Model De nitions . . . . . . . . . . . . 23 Table 4.1 OLS Regression Results for GDP Growth. . . . . . . . . . . . . 30 Table 5.1 Detailed Innovation System Indicators for Colombia . . . . . . 43 Table 5.2 Retrospective Estimates for Colombia’s GDP Growth (Model 4) 45 Table A.1 KE Pillar Indices for [1998, 2002] . . . . . . . . . . . . . . . . . 50 Table A.2 KE Pillar Indices for [2003, 2007] . . . . . . . . . . . . . . . . . 52 Table B.1 GDP Growth OLS Regressions with added Proxy Residuals . . 54 Table B.2 White Test Summary Results . . . . . . . . . . . . . . . . . . . . 55 Table C.1 Model De nitions for GDP per Capita . . . . . . . . . . . . . . 56 Table C.2 OLS Regression Results for GDP per Capita . . . . . . . . . . . 57 viii
  • 9. List of Figures Figure 4.1 GDP per Capita and Innovation System . . . . . . . . . . . . . 26 Figure 4.2 GDP per Capita and Economic Incentives . . . . . . . . . . . . 26 Figure 4.3 GDP per Capita and Governance . . . . . . . . . . . . . . . . . 27 Figure 4.4 GDP per Capita and Education . . . . . . . . . . . . . . . . . . 27 Figure 4.5 GDP per Capita and ICT . . . . . . . . . . . . . . . . . . . . . . 28 Figure 4.6 Marginal E ect of the Innovation System Scores (Model 4) . . 31 Figure 4.7 Regression’s Residuals vs Innovation System Scores (Model 4) 32 ix
  • 10. Chapter 1 Introduction has become understood to emerge through the interactions of a vari- I ety of agents such as rms, universities and governmental bodies. These actors are considered to have particular roles in processes where knowledge is created, adapted, di used and incorporated into a speci c good or service. The synergies taking place in a given country have been notably identi ed and studied by re- searchers using the National System of Innovation (NSI) concept. This comprehen- sion has provided tools for science, technology and innovation (STI) policy design. As follows, these instruments have been widely adopted by public administrators from a diversity of countries, ranging from OECD founding members such as France and Finland, to developing countries like Korea and Brazil. The development of an NSI theory has been, however, hampered by the inherent di culties for empirical analysis. In contrast to other elds such as nancial economics, the interactions involved in innovation are di cult to parametrize and thus analyze quantitatively. As a consequence, the impact of policies derived from the NSI concept has yet to be fully understood. This dissertation leverages the Knowledge Economy (KE) Framework developed by the W B as a means to assess the e ect of NSI performance on eco- nomic growth. The framework centers on the following idea: the manner in which applicable knowledge is produced and owing within society is crucial for increas- ing economic output. The KE approach integrates the NSI concept as one of the four components, referred as KE Pillars, deemed to enhance growth in terms of knowledge creation, di usion and adaptation; these are: the Economic Regime, the Innovation System, Education and Information and Communication Technologies. This framework allows an empirical study about the connection between KE vari- 1
  • 11. ables and economic growth for a sample of 75 countries (developed and developing) in the [1998, 2007] period. This analysis yields a signi cantly positive impact com- ing from the level of NSI performance, as a function of foreign technology transfer (manufactures imports and FDI) and knowledge appropriation (R&D expenditure and high-technology exports) variables, on increments of GDP. Furthermore, this dissertation advances the discussion of the Colombian case-study and diagnoses that the country has failed to harness opportunities of foreign technology trans- fer as a consequence of the disjunction between NSI actors. The organization of document is described below. Chapter 2 elaborates on the main problem of concern: to establish a quantitative link between the NSI concept and economic growth. A hypothesized solution is proposed as an application of the KE Framework. The main concepts and the previous studies, in which this dissertation intends to build upon, are revised. Chapter 3 procures to describe the data and methodology selected to approach the hypothesis. The W B data for representative variables of the KE Framework is presented. It then explains how this information is prepared for an empirical study. Chapter 4 is where the the empirical analysis takes place. The gathered evidence is fully described and evaluated in detail. Chapter 5 reviews the Colombian case-study considering the current approach and the gathered evidence. It seeks to place the exposed link between the NSI approach and economic growth at the service of policy-making in the context of this particular country. Chapter 6 synthesizes the claims, the process for their validation and the ndings of this dissertation. It also mentions the recognized research opportunities for further development of the NSI concept and its potential applications. 2
  • 12. Chapter 2 Background chapter surveys the theoretical preliminaries and previous works required T to de ne the research hypothesis which aims to connect the NSI approach with economic growth. On this account, three main subjects are covered. First, a concise background of economic theory is presented regarding endogenous growth, in order to understand the role of innovation in the development of markets. The concept of a NSI is the second topic of discussion. The trajectory of this approach is described through the exposure of representative studies, in which this dissertation is based on. Lastly, the KE Framework is introduced in order to lay the grounds for the quantitative analysis that will be undertaken in the forthcoming chapters. 2.1 Endogenous Growth and Innovation The Solow model is signi cant for economic growth theory, not only because of its revealed ndings, but also due to the shortcomings that can be derived from it. This model has brought a general understanding on how savings, population growth and technological progress a ect the level of an economy’s output and its growth over time [Mankiw, 2006]. However, the standard implementation of this model cannot explain the di erences between per capita production in high-income countries and that of the least developed countries, or why the average growth of GDP per capita is much higher in the present time than 200 years ago [Jones and Manuelli, 2005]. Total Factor Productivity has come as a way to recognize these di erences, although the reasons behind them are still a matter of debate under this approach. These limitations of the Solow model have been the main inspiration for the subsequent 3
  • 13. development of Endogenous Growth Theory (EGT). According to [Jones and Manuelli, 2005], EGT models have focused on the pro- duction and dissemination of knowledge, whether by inclusion under the assump- tions or directly in the system’s speci cation. Studies of this branch have arrived to a common conclusion: asymmetries in development rely on the di erences between social institutions across time and countries (e.g. countries with inadequate protec- tion of property rights will grow at a slower pace). In the case of the basic Solow model, the role of knowledge (technology) was considered along with labor and capital, but left unexplained and assumed exogenous. There are two main criticisms for considering knowledge as an externality, which endogenous growth centered in resolving: the mentioned di culty to explain observed long run di erences in economic growth and the fact that changes of productivity are a result of conscious decisions made by socio-economic agents. Innovation1 is a process in which new products and services are developed through R&D activities that originate in market competition. This permanently on- going process results in technological progress which is, in turn, the centerpiece of endogenous growth theory. Most economists accept technological change and inno- vation as the principal constituents of economic growth [Aghion and Howitt, 2005]. In other words, growth can di cultly be sustainable in the absence of steady tech- nological improvements [Barro and Sala-i-Martin, 2003]. This idea is supported by the fact that innovation has been integrated to many contemporary growth models. Paradoxically, innovation has only received scholarly attention until recent years. Furthermore, economic studies of innovation have been centered in microeconomics [Fagerberg, 2006]. However, a particular approach regarding the procurement of in- novation under the macroeconomic perspective has been developed based on the concept of a NSI. 1 This work implicitly uses the following de nition of innovation. It is a product or service which ful lls the following conditions: i. it contains a technological novelty either by new devel- opment, combination or application of one or more technologies; ii. it addresses a speci c mar- ket need; and iii. it generates pro ts, meaning the investment involved yielded positive returns [Escorsa, 1998]. 4
  • 14. 2.2 The ‘National System of Innovation’ Approach In the eld of economics, the NSI concept gained attention with the arrival of the- ories which highlighted the role of technology, such as the EGT. The NSI approach has received researchers’ interest due to its focus on the endogenous building of ca- pabilities for development and also because it provides a speci c role for government policy towards the technological catch-up process [Gancia and Zilibotti, 2005]. The connection between the NSI approach and economic growth is still rather unexplored. This is mainly explained by two reasons. First of all, the measure- ment of innovation systems for practical purposes has been a matter of debate [Holbrook, 2006]. This issue is important because this dissertation aims to focus in quantitative rather than qualitative analysis; the measurement problem, along with its present resolution, will be discussed and illustrated with empirical studies later on. Secondly, the NSI approach has been found insu cient to explain growth by itself, which is why this concept has been mainly used to study industrial de- velopment dynamics qualitatively [Lundvall, 2007]. For this limitation, it will be described below how the KE Framework provides a robust platform in order to link the NSI concept to economic growth. The concept of a NSI began to be developed more than 20 years ago. Since then, it has been employed to analyze industrially advanced countries such as Nor- way, Sweden and Australia; to explain successful development case-studies in Asia: Japan, South Korea, Taiwan, China and India; and likewise in Latin America: Brazil and Chile [Feinson, 2003]. Among this diversity of studies, the de nition of a NSI presented below is commonly seen. It corresponds to the rst introduction of the term in print by Christopher Freeman2 , in his book about the favorable outcome of Japanese technological and economic policy in the 1960s and 1970s [Edquist, 1997]. De nition of National System of Innovation: “The network of institutions in the public and private sectors whose activities and interactions initiate, import, modify and di use new tech- nologies.” Source: [Freeman, 1987]. 2 According to Freeman, the rst person he heard use the expression of ‘National System of Innovation’ was Bengt-Åke Lundvall. 5
  • 15. Although being increasingly popular within scholars and policymakers, the NSI approach has experienced complications since its birth. Generally, theories come within a speci c eld of science i.e. addressed and built by academics and scien- tists from a particular discipline which, to a certain extent, share similar interests, methodologies, terminology, etc. This has certainly not been the case for the NSI approach. Attempts to de ne, describe and explain this concept have ascended from a variety of elds, such as engineering, economics and management. Furthermore, the study of NSI is not only decentralized in terms of discipline, but also geograph- ically: pertaining books and papers are being published worldwide. However, the exibility of the term can be seen as an advantage, as it allows scholars to adapt the analytical tool for studying di erent contexts. The rst study by Freeman mentioned above, which began popularizing the concept, was con- ducted to understand the successful development experience of Japan. Ever since, the NSI approach has been widely used to understand how knowledge is produced and applied in industrially advanced countries and how developing economies catch- up in this process. Some exemplary studies are depicted below. [Freeman, 1995] reports how, since the 1970s, empirical evidence has begun to be gathered regarding R&D investment and innovation, particularly in in Japan, the United States and Europe. The data permitted to demonstrate how the success of innovation depends on R&D expenditure. Furthermore, not only the links between rms were found to have a critical importance, but also the external relationships with other types of institutions such as universities. In the 1950s and 1960s, the Japanese success was simplistically endorsed to product imitation and the importa- tion of foreign technology. However, when Japan’s exports started outperforming those of the United States, this explanation turned inappropriate. This overcom- ing is associated with relatively higher levels of industrial R&D spending in Japan. Nevertheless, this factor does not o er a su cient explanation. The Soviet Union and other Eastern European countries proved that dedicating R&D resources with- out any elaboration did not guarantee innovation, di usion and productivity gains. This elaboration refers to the linkages within the innovation system i.e. the impact of technical innovations in society depended on how suitable these are for domestic business, combined with the e orts devoted by rms to adopt them. A more recent NSI case-study is Korea. [Feinson, 2003] states that this country’s experience displays the bene ts of dynamic, responsive science policies towards the 6
  • 16. technological catch-up process. By articulating the NSI agents, the Korean govern- ment was able to drive the transition from a subsistence farming economy to one in which technology is acquired, di used throughout the nation and employed in favor of innovation. Korea’s rst stage was to promote technology in ows. For this purpose, the traditional path of promoting FDI and licensing was not followed. Al- ternatively, policy at this stage concentrated on establishing turnkey businesses. The steel, paper, chemical and cement industries were all founded in the form of turnkey factories, which were domestically expanded afterwards. Rather than fostering licensing, policymakers opted for the promotion of import capital goods which embody technology; the importation of this kind of goods may have been the most productive method of technology transfer. At that time, Korea probably relied more on this channel than any other newly industrialized country. The second stage was to assimilate the imported technology into its domestic production lines. This was addressed by public funding and a series of incentives towards R&D, including tax breaks and exemption from military services for key personnel. The third stage consisted in an outward orientation in the form of export promotion policies. These included the liberalization of access to imported intermediate products, the facili- tation of banking loans for working capital destined to export-related investments and the elimination of restrictions to foreign capital. The NSI approach has also been used to evaluate STI policies in countries of the Latin American region. For example, [Holm-Nielsen and Agapitova, 2002] studied how Chile has increased its competitiveness due to a favorable macroeconomic en- vironment for STI. However, in this country, research institutions have remained rather disjointed from the productive sector, which wastes potential for continu- ous product innovation and hinders increments of living standards. The analysis suggests that Chile needs to make its NSI more e ective in two main ways: by strengthening the venture capital market and introducing more measures for pro- moting networking and cooperation between science and industry. This achieve- ments are deemed to increase the returns to R&D investments. The unconstrained aspect of the de nition for NSI has allowed a wide variety of analyses. However, this exibility brings complications because just as a NSI diverges between countries and regions, so does the approach from their analysts [Lundvall, 2007]. In this ambivalence and lack of consensus relies the criticisms of the NSI concept. In natural sciences, agreeing in strict de nitions is seemed 7
  • 17. as crucial to allow scienti c progress. Modern economics is characterized for the emulation of this rigidity. This might be the reason why the NSI approach has slowly penetrated economic theory. The components, attributes and relationships that compose a NSI are tremendously di cult to quantify because the level of anal- ysis corresponds to that of an entire nation. To overcome for this limitation, the quanti cation of National Systems of Innovation has been focused not in its com- ponents but in the overall performance of the system [Carlsson et al., 2002]. Data recording has recently begun in this respect, across a wide range of countries. As follows, concrete indicators regarding manufactures imports and exports, FDI, R&D expenditure can be used to measure how a given NSI is performing. This quantita- tive approach is undertaken in the KE Framework and employed here, as explained in the forthcoming Section 2.3. However, it is pertinent to revise before how previ- ous works have pursued to identify statistical evidence for a relationship between NSI performance and economic growth. These studies re ect various important lessons which are recurrently taken into account in this dissertation. 2.2.1 Empirical Studies of NSI [Freeman, 2002] discusses the relevance of innovation systems for economic growth over the last two centuries. Based on this timespan, the study uses gathered indi- cators which describe NSI performance and aims to identify to which extent their variations resulted in faster or slower rates of growth. While labor and capital pro- ductivity are employed for the rst century, the second is analyzed using more spe- cialized indicators such as manufactures exports, information and communication technologies expenditure and R&D personnel, among others. The analysis is rather limited by data availability. However, it argues that these indicators show a clear pattern of how, after the industrial revolution, the NSI approach can be employed quantitatively to understand the divergence in paths of economic growth. [Rodríguez-Pose and Crescenzi, 2006] analyzes the link between R&D invest- ment and patents with economic growth. The study focuses on the connection be- tween the e ciency of innovation systems and the geographical di usion of knowl- edge spillovers. With an econometric analysis, that includes only European coun- tries, the work highlights two main results. First, the interaction between research and socio-economic institutions determined the potential for maximizing the capac- 8
  • 18. ity to innovate. Secondly, proximity had an important role in allowing the di usion of economically productive knowledge and its impact in overall growth. [Krammer, 2008] executes a cross-country analysis for Eastern Europe to explore what enables countries to innovate more than others at a national level. As a proxy for innovation, Krammer uses the number of international patents granted by the US patent o ce. His results suggest R&D commitments and the ‘innovative tradi- tion’ were key for increasing the knowledge stock. Openness and the protection of intellectual property rights determined higher international patenting, while struc- tural industrial distortions had a negative in uence in the quantity of patents. Lastly, [Fagerberg and Srholec, 2007] uses a broader set of data for studying the role of innovation within a set of capabilities for development. The work bases its methodology on a factor analysis that compromises 25 indicators and 115 countries from 1992 until 2004. By this method, four types of capabilities were identi ed: the development of the innovation system, the quality of governance, the character of the political system and the degree of openness. Of these, the innovation system and governance were found to be of noteworthy pertinence for economic growth. As it will be shown later in Chapter 3, this dissertation will particularly build upon this study. In contrast, however, di erent indicators are selected, arranged and statistically analyzed here, using the KE Framework. This method of analysis points to similar conclusions to those of [Fagerberg and Srholec, 2007]. In synthesis, the trajectory of the NSI approach has been more qualitative than quantitative. This approach aims to explain the dynamics of industrialization, tech- nological catch-up and development. Empirical analysis has been made di cult due to the problems for measuring innovation systems. However, researchers have pro- posed performance as a plausible scale of reference, which has allowed some studies to emerge. Still, there is a need for more quantitative research to validate the NSI concept. This dissertation adds to these e orts by turning to the KE Framework, described in the following section. 2.3 Leveraging on the Knowledge Economy Framework This section describes how the KE Framework integrates the NSI approach, comple- menting it in order to allow a robust analysis of its role towards economic growth. 9
  • 19. For this purpose, this section includes relevant excerpts from the book Building Knowledge Economies: Advanced Strategies for Development [World Bank, 2009a]; in this publication, the Wold Bank compiled its work on the KE Framework. The W B has addressed the KE Framework through its Knowledge for Development (K4D) Program. This program contributes to the framework by pro- ducing publications and distributing those of third-party specialists in this eld of study. Its aim is to promote the framework’s awareness among policymakers world- wide. Through the KE literature and the e orts for data compilation from the K4D Program, the W B has constructed the KE Framework to analyze, study and devise policy recommendations for knowledge-driven economic growth. The Knowledge Economy Framework... “...describes how an economy relies on knowledge as the key engine for growth. It is an economy in which knowledge is acquired, created, disseminated and applied to enhance economic development.” Source: [World Bank, 2009a]. The KE Framework, as depicted in its name, highlights the increasingly protag- onist role knowledge has in an economy. Countries worldwide, both industrially advanced and developing, have been recognizing know-how and expertise as criti- cal as other economic resources. Industrial production requires appropriate policies that re ect the current interconnected and globalized economic context. According to [World Bank, 2009a], the KE Framework rests on four pillars: the Economic Regime, the Innovation System, Education and Information and Com- munication Technologies. These have been previously supported as foremost for economic development by an ample literature and empirical works. Their de nition and pertinence are described below. Economic Regime. It is de ned as the set of economic and institutional incentives designed to promote an environment that permits knowledge creation, assim- ilation and di usion. This pillar covers a broad set of macroeconomic issues and policies, such as trade, nance and banking and governance. Due to this broadness, this pillar is sometimes divided into two sub-pillars, Economic In- 10
  • 20. centives and Governance, to facilitate analysis3 . The former is related to how resources can mobilize within an economy, while the later deals with how the political circumstances and its stability provide an appropriate business climate. A favorable Economic Regime is required to obtain better policy results from the other, more functional pillars. Industrially advanced countries generally have solid institutions based on democracy and free markets. Governments promote the development of their institutional regimes by improving labor and nancial markets, and by strengthening governance (e.g. increasing the enforcement of contracts and controlling corruption). Innovation System. It consists of rms, research centers, universities, think tanks and other institutions within a given country, that import or produce knowl- edge and adapt technologies to the local context4 . STI activities require pub- lic support in an ample range of ways such as the funding of basic research and the facilitation of knowledge di usion. The latter is of particular impor- tance for developing countries, where knowledge and technology, the inputs for innovation, arrive from abroad in the form of FDI and manufactures im- ports, among other channels. Indigenous knowledge capabilities should also receive attention. The importance of this pillar relies on the empowerment for achieving desired social and economic outcomes through the application of knowledge. Education. This pillar is related to the human skills required for the acquisition and exercise of knowledge. The preparation of the labor force includes the primary, secondary and tertiary levels of education, vocational training and continuous learning. The focus on a given level of education depends on the country’s stage in economic development. A member from the Least Devel- oped Countries group should give more attention to primary education, as literacy and arithmetic skills are required before more advanced competences are gained. As the country’s economy grows, the relevancy of continuous 3 This sub-pillar distinction will be taken into account in this dissertation. In Chapter 4 the re- sults will show that it is important to analyze both Economic Incentives and Governance separately than as a whole. 4 W B ’s nomenclature omits the word ‘national.’ However, by comparing the de ni- tions of ‘Innovation System’ and NSI, it can be a rmed that both terms are concurrent. 11
  • 21. learning increases, as this type of education is necessary for innovation re- sulting from the constant adaptation of knowledge. Education creates jobs, reduces poverty levels and increases empowerment. It is a fundamental pillar for the KE. Information and Communication Technologies (ICT). ICT encompass the types of technologies that enable the di usion of knowledge. ICT, bearing tele- phone, television, radio and Internet networks, are critical for the economies of today, based on globalization and information. These reduce transaction costs signi cantly by providing accessibility to knowledge. A strong ICT pillar allows rapid and reliable exchange of information within a country and across its borders. Recent advances are a ecting how knowledge is acquired, created, shared and applied, which has positively impacted manufacturing, trade, gov- ernance and education activities, among others. Regarding this pillar, policies consider telecommunication legislation, along with the investment required for building and capitalizing ICT through the socio-economic dimension. 2.4 Research Hypothesis The problem for the measurement of innovation systems and the lack of a robust framework, mentioned in Section 2.2, must be solved for studying the connection between NSI theory and economic growth. These issues are both addressed by the KE Framework i.e. by adopting the accountability of performance, as the achieve- ments of a given NSI are analyzed using indicators such as R&D expenditure and high-technology exports. The KE Framework also integrates NSI theory with the other three5 pillars mentioned above, which allows to study economic development from a knowledge-based perspective. With the concepts that have been revised up to this point, this dissertation’s research hypothesis is structured below. 5 Four (a total of ve KE Pillars), if Governance and Economic Incentives are considered as separate pillars, as it would be the case later on in this dissertation. 12
  • 22. Research hypothesis: “A positive connection between NSI performance indicators and eco- nomic growth can be quantitatively found under the KE Framework. This connection reveals challenges and opportunities for a developing economy such as Colombia, along with STI policy recommendations.” The relevancy of this hypothesis lies in three main aspects. First of all, innova- tion has been considered to be a central issue for EGT and in Economics in general. Exploring the dynamics of technological progress using new metrics represents an attractive contribution to the eld. Second, it would contribute to the e orts for the empirical veri cation of the NSI approach reviewed above, by the means of a new methodology. Thirdly, the validation of the hypothesis statement would favor the position of the NSI concept as one of the centerpieces in the development process. Understanding the role of National Systems of Innovation would nurture policy- makers in the areas of STI. The subsequent pages aim to address this hypothesis. In particular, the next chapter describes the gathering of data and its analysis, taking into account the literature revised until this point. 13
  • 23. Chapter 3 Data and Methodology for Analysis , it was discussed how the KE provides the required framework for P understanding the role of a given NSI in its economy. Upon this background and seeking to contribute towards a deeper apprehension of the NSI concept and its validity, a research hypothesis was de ned. Aiming to link NSI performance with economic growth, this chapter states and thoroughly describes the employed data, and then declares the methodology for the respective statistical analysis. With this purpose in mind, the subject of the KE data is discussed at rst. The source and the process of selection and recollection are concisely portrayed. After, the issue of a dataset with an excessive number of variables is exposed. This is resolved via a Principal-Component Factor Analysis, which groups the variables of a given pillar in the construction of an associated index. With the constructed indices, the OLS regression models are stated for the hypothesis’s testing. 3.1 Data To investigate any e ect of NSI performance in economic growth, data was gathered seeking to satisfy the following two criteria: a diversity that covers all the features of the KE and the availability of observations for signi cant amount of time. Through the K4D Program, mentioned in Section 2.3, the W B classi es a variety of pertinent statistics from the World Development Indicators (WDI) under the four main KE Pillars [World Bank, 2010b]. The program’s dataset makes refer- ence to more than 100 WDI variables. Data recollection began with a depuration of this source, aiming to comply with the two criteria stated above. In particular, 14
  • 24. the process took into account the fact that several of the referred KE variables have started being recorded, across a signi cant amount of countries, only until recently. These variables were discarded, for the limited observations would not allow a sig- ni cant timespan for analysis. The selection process yielded a total of 19 variables for 75 countries in the the period [1998, 2007]. To balance the dataset and compensate for missing gures, the timespan was divided into two 5-year intervals. These are: [1998, 2002] and [2003, 2007]. Observations were de ned as the average values of the variables for each of these two periods. The summary statistics of these variables are displayed in Table 3.1. 3.1.1 Rationale for Variable Selection Although most of the variables are already classi ed under the KE Framework by the K4D Program, it is necessary to discuss why each is signi cant for the pillar it represents. Starting with the Economic Regime Pillar, there is market capitalization of listed companies, domestic credit provided by the banking sector and domestic credit to the private sector, all measured in % of GDP. The rst variable is de ned as the sum of the product between share price and the number of shares outstanding, for all companies listed in the country’s stock exchange. The second variable refers to the totality of credits conceded to various sectors on a gross basis, excluding those provided to the central government. The third includes nancial resources provided to the private sector (e.g. loans, non-equity securities and trade credits). These three variables have been employed in studies concerning nancial market development and economic growth; although the importance of the former in the latter has been a matter of debate, several studies have evidenced on a signi cantly positive e ect [Levine, 1997];[Levine and Zervos, 1998]. There are six variables for the Governance Pillar. Their de nitions are presented as stipulated in [Kaufmann et al., 2009]. Voice and Accountability captures the per- ceptions to which citizens from a given country are able to participate in elections, along with freedom of expression, freedom of association, and free media. Political Stability re ects perceptions of the probability that the government will be destabi- lized or overthrown by unconstitutional or violent ways, including political violence and terrorism. Government E ectiveness captures the perception of the quality of 15
  • 25. Table 3.1: Summary Statistics for the Selected KE Variables Obs [1998,2002] Obs Mean Std. Dev. Min Max Obs [2003,2007] ∗ 100 Economic Incentives (Values in % of GDP) Market capitalization of 225 53.21 60.05 0.07 434.31 49.78 listed companies Domestic credit provided by 362 57.66 53.63 -57.35 304.29 50.00 the banking sector Domestic credit to the 362 46.6 44.58 0.72 220.73 50.00 private sector Governance (Indices) Voice and Accountability 396 -0.03 1 -2.19 1.66 49.24 Political Stability 389 -0.06 0.98 -2.78 1.64 48.07 Government E ectiveness 397 -0.02 1 -2.15 2.26 49.12 Regulatory Quality 391 -0.04 1 -2.46 1.96 49.10 Rule of Law 394 -0.05 0.99 -2.33 2.07 48.73 Control of Corruption 391 -0.02 1 -1.79 2.49 49.10 Innovation System Manufactures imports 336 66.67 11.04 21.16 90.89 50.60 (% of merchandise imports) High-technology exports 328 9.76 12.59 0 73.09 50.91 (% of manufactures exports) Foreign direct investment, 347 4.96 5.58 -6.58 39.35 49.86 net in ows (% of GDP) Research and development 200 0.87 0.92 0.01 4.47 52.50 expenditure (% of GDP) Education Public spending on educa- 305 4.69 2.09 0.6 15.57 52.79 tion, total (% of GDP) School enrollment, 355 72.54 31.06 5.93 156.48 49.86 secondary (% gross) School enrollment, tertiary 318 26.67 23.32 0.14 91.35 50.31 (% gross) ICT (Values per 100 people) Personal computers 366 11.9 17.3 0.01 84.69 49.45 Mobile phones and landlines 398 51.03 49.19 0.17 186.37 50.25 Internet users 393 14.67 18.98 0 81.21 49.62 Source: calculations based on [World Bank, 2009b]. 16
  • 26. public services, the civil service and the extent of its independence from political pressures, along with the credibility towards the government’s formulation and im- plementation of policies. The Regulatory Quality indicator perceives the ability of the government to devise and carry out robust policies and regulations that allow and foster the development of the private sector. Rule of Law captures the impres- sion on how socio-economic agents have con dence in and abide to the rules of society i.e. speci cally, the quality of contract enforcement, property rights, the po- lice and the courts, as well as the protection from crime and violence. Lastly, the Control of Corruption captures perceptions of the extent to which public power is safeguarded from rent-seekers, considering all levels of corruption, and the seclu- sion of the State from private interests. Over the last decade, governance has been a central topic of growth promotion policies, especially in developing countries. According to [Gray, 2007], the most prevalent approach in governance policy-making is known as the ‘good governance’ agenda, which contains the six variables previously mentioned. Representatives of this agenda highlight its importance not only in the satisfaction of citizens’ aspira- tions regarding public institutions, but also as a means to foster economic growth and as a sustainable mechanism to reduce poverty. While the link between institu- tions and growth was a central matter of classical economics, the notion of ‘good governance’ had its grounds laid only until the 1970s and 1980s. The creation of quantitative measurements has been key to structure a consensus of the positive relationship between governance and economic growth. Regarding the (National) Innovation System Pillar, manufactures imports (% of merchandise imports, foreign direct investment (net in ows, % of GDP), high- technology exports (% of merchandise exports) and research and development ex- penditure (% of GDP) were selected as representative variables under the NSI per- formance approach. The case-studies included in Section 2.2 show that the rst two variables are pertinent channels of foreign technology transfer for the catch- up process. Manufactures imports incorporate foreign technology and represent intermediate capital goods necessary for producing value added exports. FDI is rel- evant as a source of capital for export promotion albeit does not necessarily ow into sectors intensive in technology, which is why public policy is sometimes em- ployed to foster investments that imply technology transfer. The cited case-studies also show that other two variables of the Innovation System Pillar measure the in- 17
  • 27. digenous appropriation of knowledge i.e. these are related to how the mentioned technology transfer channels are being utilized in the economy for producing inno- vations and towards the promotion of technological capabilities. High-technology exports account for this explicitly, as it refers to domestic production which em- ploys indigenously developed or adapted technology. Regarding R&D expenditure, the referenced authors of NSI studies recognize it as decisive in the adaptation of foreign technology to the local context. By understanding the signi cance of these four indicators, this dissertation seeks to elaborate on previous empirical research and explore the link between NSI performance and economic growth. Before proceeding with the remaining KE Pillars, it is important to acknowledge that, similar to other quantitative studies, the variables selected here emphasize formal modes of learning and innovation based in science and technology activi- ties [Lundvall, 2007]. This emphasis is re ected here in the selection of indicators of R&D and capital-embedded industrial goods. However, innovation strategies based on experience and the “doing, using and interacting” learning mode are rather overlooked. This is explained by the lack of standardized variables to represent experience-based innovations. In the case of the Education Pillar, three representative variables were selected. Public spending on education (% of GDP) adds up expenditure on education of public authorities at all levels, along with the subsidies to private education at the primary, secondary, and tertiary levels. Secondary school enrollment (% gross) and tertiary school enrollment (% gross) are the ratio of total enrollment, regardless of age, to the population of the age group that o cially corresponds to the level of education. According to [World Bank, 2009a], secondary education completes the provision of basic education that began at the primary level and lays the founda- tion for future learning. It yields both individual and social returns and provides an important amount of human capital required for countries’ economic growth. The role of tertiary education is crucial. Universities and research institutions have to address the call for creating a pool of experts capable of acquiring science and tech- nology and adapting it to the domestic context. Regarding the link between educa- tion and economic growth, [Teles and Andrade, 2004] states that while the evidence has been asymmetrical it mainly points towards a positive causal relationship. The same study identi ed a positive relation between public spending on education and economic growth. The reported signi cance of the relationship, however, varied 18
  • 28. depending on the composition of governmental spending between basic and higher education i.e. it lost its signi cance when the latter was not promoted. For the remaining Pillar, ICT, the variables measured per 100 people are: per- sonal computers, mobile phones and landlines and Internet users. As reported by [Batchelor et al., 2005], previous studies agree on how ICT can help develop- ing countries address a wide range of socioeconomic activities: the use of ITC en- hances the production of goods and the provision of services and thus increases productivity. There is less agreement, however, on how much a priority it should be to promote the increase of ICT infrastructure. These technologies are increas- ingly being seen as means to other development requirements rather than as an end themselves. Policies associated with this KE Pillar have been focused on alleviating the wide disparities in access; the poor is the part of society most out of reach from ICT. 3.2 The Construction of KE Pillar Indices Even after depuration, the dataset from Table 3.1 is still composed of too many variables for an econometric analysis. [Fagerberg and Srholec, 2007] faced a sim- ilar problem in its attempt to explore the relationship between NSI and economic growth. To face a dataset with numerous variables, the work employed a Principal- Component Factor (PCF) Analysis. As described in the study, this process is based on the idea that variables from the same category are likely to be signi cantly cor- related and thus can be reduced into a smaller number of indicators, which re ect the variance dimension of the data. The PCF Analysis assigns speci c “loadings” which weigh in the calculation of the factor score for each country. Countries re- ceive scores for each of its KE Pillars by adding the product of the pillar variables’ values and the corresponding coe cients, which are derived from the loadings. As it was mentioned in Chapter 2, [Fagerberg and Srholec, 2007] identi ed four factors from a set of variables: the innovation system, governance, the political sys- tem, and openness. In contrast, this dissertation uses a di erent set of representative variables for the KE Pillars. Using the K4D Program’s classi cation, individual PCF Analyses were carried out for each of the pillars. This ensured that the result- ing indicators kept the structure suggested by the KE Framework. Consequently, 19
  • 29. the indicators for each of the pillars contain information only from their respective variables. The Economic Regime Pillar is divided into two sub-pillars, as suggested by the W B : Economic Incentives and Governance [World Bank, 2009a]. The PCF Analysis executed here successfully identi ed an underlying structure for each of the KE Pillars and generated proxy indices. For every sampled country, an associated index (value) was calculated. The resulting set of KE Pillar Indices can be viewed in Appendix A. The PCF Analysis maximizes the amount of overall variance (accounting by all the variables as a group) that is to be captured by the index. The correlations in Table 3.2 indicate the “relevance” each variable has in its corresponding index. For example, in the case of Education, both secondary and tertiary school enrollment have a higher weigh in the calculation of the associated index values, compared with public spending variable. The constructed Education Index is more correlated with the rst two variables because the two enrollment indicators are more correlated with each other in comparison to the scal indicator1 . The variance explained by the Innovation System Index is 40.18%. While this value is rather low, it is still signi cant on what is considered best practices for PCF Analysis, as stated in [Costello and Osborn, 2005]. Furthermore, correlations in this index are all above the 32% recommended borderline. The values show that, in this constructed index, the more relevant variable is high-technology exports, followed by manufactures imports, R&D expenditure and, lastly, FDI in ows. It is critical to recognize that this particular Innovation System Index is designed to be functional only in the context of the KE and thus should not be employed in- dependently. There are other innovation indices more suitable for a comparative analysis or ranking purposes; a prominent one is the National Innovative Capac- ity (NIC) Index used in the Global Competitiveness Report [Porter and Stern, 2002]. Stand-alone innovation indices are, however, not suitable for this dissertation as these consider a series of factors that are better classi ed in other KE Pillars e.g. in the case of the NIC index: venture capital availability (Economic Incentives), the quality of institutions (Governance), human capital (Education), and the social penetration of information and communications infrastructure (ICT). The KE ap- proach allows the separation of these factors and thus an independent analysis of 1 For more details on PCF Analysis, please refer to [Smith, 2002]. 20
  • 30. Table 3.2: The Constitution of the KE Pillar Indices Economic Incentives Index Education Index Variance Explained: 81.12% Correlation Variance Explained: 63.38% Correlation Market capitalization of listed Public spending on education, 0.79 0.45 companies total Domestic credit provided by 0.94 School enrollment, secondary 0.93 the banking sector Domestic credit to the private 0.97 School enrollment, tertiary 0.91 sector Governance Index ICT Index Variance Explained: 87.84% Correlation Variance Explained: 91.09% Correlation Voice and Accountability 0.89 Personal computers 0.95 Political Stability 0.86 Mobile phones and landlines 0.94 Government E ectiveness 0.97 Internet users 0.97 Regulatory Quality 0.95 Rule of Law 0.98 Control of Corruption 0.96 Innovation System Index Variance Explained: 40.18% Correlation Manufactures imports 0.69 High-technology exports 0.76 Foreign direct investment, 0.35 net in ows Research and development 0.65 expenditure NSI performance. To compare the NIC Index with the Innovation System Index, developed in this dissertation, the correlation was calculated between the 2001 value of the former2 and the average for the years [1998, 2002] of the latter3 . A correlation of 65% sug- gests that the native Innovation System Index captures a considerable portion of the NIC dataset, while some of the remaining percentage is likely to be balanced with the information included in the other pillars. As mentioned earlier when elaborat- ing on its variables, the Innovation System Pillar Index should be interpreted as an attempt to represent the general impression of how a given NSI performs through its development. The indicator intends to re ect the several case-studies presented in Section 2.2. 2 Based on data from [Porter and Stern, 2002]. 3 Using the values of Appendix A. 21
  • 31. 3.3 Speci cation of Models The KE Pillar Indices, which contain the information of the variables in Table 3.1, can now be used as proxy variables in an econometric analysis for testing the hy- pothesis devised in Chapter 2. The summary statistics of the variables to be included in the upcoming regressions are presented in the following Table 3.3. Table 3.3: Summary Statistics for the Regression Variables Obs Mean Std. Dev. Min Max Explained Variable Annual GDP Growth (%) 134 1.27 0.67 -1.90 2.57 Proxy Variables (KE Pillar Indices) Economic Incentives 134 0 1 -1.25 2.85 Governance 134 0 1 -1.88 1.65 Innovation System 134 0 1 -2.05 2.26 Education 134 0 1 -2.95 2.35 ICT 134 0 1 -1.27 2.48 Control Variables GDP per Capita 134 8.60 1.35 5.53 10.61 (constant 2000 US $) Dummies (Binary Variables) [1998, 2002] Observation 134 0.52 0.50 0 1 Sub-Saharan Africa 134 0.07 0.26 0 1 Latin America & the Caribbean 134 0.15 0.36 0 1 East Asia & Paci c 134 0.06 0.24 0 1 Middle East & North Africa 134 0.05 0.22 0 1 South Asia 134 0.01 0.12 0 1 Europe & Central Asia 134 0.17 0.38 0 1 The logarithm of GDP growth is selected as the explained variable and the ve KE Pillar Indices are included as explanatory proxy variables. Furthermore, eight control variables are included in the analysis. Six of them are regional dummies (binary variables), which group developing countries according to W B ’s geographic classi cation. These are: Sub-Saharan Africa, Latin America & the Caribbean, East Asia & Paci c, Middle East & North Africa, South Asia and Eu- rope & Central Asia [World Bank, 2010a]. The null case of the regional dummy 22
  • 32. variables corresponds to high-income countries. The mean of the regional dum- mies represent their share in the dataset (e.g. 15% of the considered countries are from Latin America & the Caribbean). Adding all the means result in the propor- tion of developing countries in the [1998, 2007] sample: 51%. Thus, the remaining 49% of observations conform the sampled high-income countries. Another control variable is the logarithm of GDP per capita, to consider conditional convergence4 . Lastly, there is one more dummy that indicates if the data-point corresponds to a [1998, 2003] observation, to consider time e ects i.e. temporal variations that are not captured by the KE Pillars and the other control variables. The econometric analysis is undertaken in a set of four models. These models are represented in Table 3.4. All of these have the logarithm of GDP growth lngdpgi as the explained variable and have the KE Pillar Indices as proxy variables, a constant and an error term, represented as Pki , C and i, respectively. The basic model includes only the ve proxies, while the subsequent models incorporate the control variables progressively. Table 3.4: Econometric Analysis: Model De nitions 5 6 lngdpgi = βki Pki + β6i f irsti + β7i lngdpli + β(k+7)i Rki + β14 C + i k=1 k=1 Model 1 Model 2 Model 3 Model 4 The second model adds the f irsti dummy variable, which equals to one when the observation corresponds to the [1998, 2002] period and zero otherwise, thus tak- ing into account time e ects. The third augments the analysis with the variable lngdpli (logarithm of GDP per capita) to consider the in uence of conditional con- vergence. Finally, the fourth model adds the six regional dummy variables, written as Rki , to check for the geographical particularities that might a ect growth and are not captured by the other variables. 4 The theory of conditional convergence states that, given certain conditions, poorer countries grow faster than their richer counterparts, until all economies reach the same level of GDP per capita. Developing countries have the potential to increase their economic output levels at a faster rate, due to the fact that the e ects of diminishing returns are not as consolidated as in higher income countries. 23
  • 33. To recapitulate, this chapter described how data was extracted using W B guidelines and databases, in order to robustly represent the KE Pillars of 75 countries worldwide for a ten-year period of analysis: [1998, 2007]. This time interval was divided into two consecutive quinquennial periods: [1998, 2002] and [2003, 2007]. For each of these, averages of available data were calculated. The resulting variables were then employed for the construction of the associated KE Pillar Indices using PCF Analysis. These indices are to be used as explanatory proxy variables for GDP growth, along with eight control variables. The coe cients from the models stipulated above are to be estimated through regressions. In the next chapter, these results are displayed and analyzed in detail. 24
  • 34. Chapter 4 Results and Evaluation on the data and the methodology for its analysis, both presented in the B last chapter, this part of the dissertation seeks to exhibit the relevant outcomes in the validation of the hypothesis: the existence of a statistical link between NSI performance and economic growth under the KE Framework. This chapter is di- vided into two parts. First, scatter plots are included for each of the constructed KE Pillar Indices and GDP per capita. These give a rst view on how each index is relevant within the economic activity. The second part focuses on giving out information from the econometric analysis of the previously de ned models. 4.1 KE Pillar Indices vs GDP per Capita Before revising the econometric analysis, it is appropriate to get an initial sense of the roles the Innovation System Index and the other KE Pillar Indices have on eco- nomic growth. For this purpose, the present section will use scatter plots. These show a graph with the scores obtained by the 75 sampled countries on each pillar (see Appendix A) on the x-axis and their respective log of GDP per capita on the y-axis. To check for di erences between the two time periods, the dots are clas- si ed accordingly. A tted line is included to illustrate the general trend of the relationship between the index’s scores and the production levels. Regarding the main Pillar Index of concern, the Innovation System, Figure 4.1 displays a positive relationship with GDP per capita, with a correlation equivalent to 0.6373. Countries that scored a better NSI performance i.e. a better acquisition, production and absorption of applicable knowledge, at the same time experienced 25
  • 35. Figure 4.1: GDP per Capita and Innovation System higher levels of income. The tted lines suggest that the relationship strengthened between the rst period and the second. For the Economic Incentives Index, Figure 4.2 shows an apparently similar as- sociation. The correlation is slightly stronger compared to the Innovation System index: 0.6926. The gure depicts that countries which scored high in this index, with an environment suitable for a better allocation of resources represented by superior levels of domestic investment, had greater GDP per capita levels between [1998, 2007]. The slope decreased from the former quinquennial period to the latter, although not by much. Figure 4.2: GDP per Capita and Economic Incentives 26
  • 36. Figure 4.3: GDP per Capita and Governance The case of the Governance Index, shown in Figure 4.3, exhibits the largest cor- relation among all pillars: 0.8714. Compared to the rest graphs, the Governance tted value lines are steepest. It shows that countries with better institutions si- multaneously experienced superior positions of GDP per capita. This relationship appears to have slightly decreased over the time of analysis. Figure 4.4 shows the scatter plot for the Education Index. This index has a cor- relation of 0.7546 with GDP per capita. The slopes of the tted value lines remained practically the same. The data from the analyzed time interval supports the idea that richer nations have a more skilled labor force, which e ciently generates and Figure 4.4: GDP per Capita and Education 27
  • 37. applies knowledge. Lastly, Figure 4.5 is the respective graph for the ICT Index. As the other KE Pil- lars, it also displays a strong positive correlation with GDP per capita: 0.8392. The slopes of the tted value lines, however, presented the most enunciated decrease between [1998, 2002] and [2003, 2007]. This change might have a ected the econo- metric analysis, as mentioned later on. Still, it can be said that better infrastructure for the communication and di usion of information and knowledge is strongly re- lated to higher income per capita levels. Figure 4.5: GDP per Capita and ICT Although ones in greater measure than others, all the KE Pillar Indices display a positive correlation with economic growth. The fact that the Innovation Systems Index scores account for the lowest correlation with GDP per capita (albeit still high), is noteworthy. In a broader perspective, these graphs support the notions of the authors mentioned in Chapter 2: the NSI concept and the KE framework are relevant towards economic output. The Innovation System’s performance, along with the set of Economic Incentives, the level of Governance, Education and the expansion of ICT are playing signi cant roles in the economies of today. For a deeper understanding on how these economies grow in relation to the KE Pillars, the following section presents and discusses the results of the econometric analysis based on the models proposed in Chapter 3. 28
  • 38. 4.2 Econometric Analysis The KE Pillar Indices constructed in Section 3.2 are now employed for regressions with economic growth as the explained variable, using the models de ned in Section 3.3. Table 4.1 contains the results from the regressions against the logarithm of GDP growth. The table points out the estimated coe cients for each of the models. All the estimations are based on the Ordinary Least Squares (OLS) method. However, Model 1 and Model 3 were estimated using the ‘Huber-White Sandwich Estimator’ of variance in order to calculate robust standard errors, as evidence of heteroskedas- ticity was found for these models1 . Columns (1) through (4) correspond to standard OLS regressions, while (5) and (6) use a stepwise estimation to identify the speci ca- tion with the best statistical properties2 . Column (5) begins the estimation process with Model 3 and excludes the quartile of countries with lowest GDP per capita, while (6) starts with Model 4 without any variation to the sample. It is necessary to acknowledge that the stepwise regressions are included speci cally to provide secondary evidence for the relationship between the Innovation System Index and GDP growth. The resulting estimations (5) and (6), unlike (1) through (4), are not intended to be representative for the KE Framework as the stepwise regressions discarded some of the KE Pillar Indices. The possibility of endogeneity3 was addressed by the means of the Durbin-Wu- Hausman Test, which is conformed by two steps. In the rst one, each potentially endogenous proxy variable was regressed on all exogenous variables (the other proxies), along with the variables that were used in the construction of the regressed index. The resulting residuals are, in the second step, added to the new regression of the original model [Wooldridge, 2002]. If any residual coe cient was to come as signi cant in one of these latter regressions, endogeneity of the corresponding proxy variable needs to be accepted and the associated model should be estimated by two-stage least squares in order to achieve consistent results. The resulting co- 1 The results of White’s heteroskedasticity tests are included in Appendix B.2. 2 The stepwise estimation seeks to dismiss variables that do not provide explanatory power to the model given a particular signi cance level, in this case 10%. The process begins with the full model and checks whether the calculated p-value of a variable falls farther from the selected frontier. It then excludes the most statistically meaningless variable and starts over. At each step the procedure also inspects if a variable that was discarded earlier has become signi cant. 3 Endogeneity occurs when there is a causality loop between the explained and the explanatory variables. 29
  • 39. Table 4.1: OLS Regression Results for GDP Growth. C V Standard Regressions Stepwise Regressions (1) (2) (3) (4) (5) (6) Model 1 Model 2† Model 3 Model 4† Model 3‡ Model 4† Economic -0.213** -0.194** -0.150* -0.125 -0.151** -0.220*** Incentives (0.0852) (0.0860) (0.0781) (0.0942) (0.0700) (0.0737) -0.14 0.17 0.271** 0.280** 0.219* Governance (0.119) (0.127) (0.127) (0.141) (0.113) Innovation 0.04 0.119* 0.139** 0.243*** 0.178** 0.201** System (0.0747) (0.0683) (0.0686) (0.0847) (0.0749) (0.0876) -0.02 0.01 0.05 0.01 Education (0.0100) (0.0802) (0.0911) (0.0959) ICT 0.190 -0.306** -0.20 -0.329* (0.119) (0.148) (0.147) (0.174) [1998, 2002] -0.771*** -0.717*** -0.819*** -0.500*** -0.544*** Observation (0.165) (0.143) (0.173) (0.108) (0.105) log(GDP -0.221** -0.207* -0.361*** per capita) (0.0933) (0.119) (0.112) Sub Saharan -0.08 0.440*** Africa (0.252) (0.148) Latin America -0.27 the Caribbean (0.212) East Asia -0.65 Paci c (0.491) Middle East 0.21 0.552*** North Africa (0.219) (0.162) South 0.42 1.100*** Asia (0.309) (0.236) Europe 0.19 0.467** Central Asia (0.209) (0.181) 1.279*** 1.684*** 3.554*** 3.518*** 4.703*** 1.404*** Constant (0.0568) (0.0734) (0.793) (1.0950) (0.999) (0.0793) Observations 134 134 134 134 100 134 R-squared 0.09 0.25 0.29 0.37 0.30 0.31 Note: Standard errors in parentheses *** p 0.01, ** p 0.05, * p 0.1 † Using the Huber-White Sandwich Estimator of variance. ‡ Excluding the quartile of poorest countries. 30
  • 40. e cients of this test are included in Appendix B. In this occasion, no evidence of endogeneity was found. Straightforwardly, the most notable result is the recurrent signi cance of the Innovation System proxy variable through the regressions, suggesting a positive e ect from this particular Pillar Index upon GDP growth. Out of all the KE Pillars, it displays the most signi cant evidence: in columns (2), (3), (4) (5) and (6) with p-values lower than 10%, 5%, 1%, 5% and 5%, respectively. These results indicate that countries with better NSI performance experienced greater GDP growth in the period of analysis. This positive relationship is depicted in the following Figure 4.6. The graph shows the marginal e ect of the Innovation System Index regressor on GDP growth, after taking into account the associations between the other variables included in the regression (column (4); Model 4). The slope of the tted line corresponds to the Innovation System Index’s β calculated in the regression. As a support for the va- lidity of this model’s particular speci cation, another plot is included as Figure 4.7. The regression’s residuals do not display any apparent pattern with the Innovation System Index, supporting the absence of endogeneity discussed earlier. Figure 4.6: Marginal E ect of the Innovation System Scores (Model 4) To analyze the relationship between Innovation System’s performance and GDP growth, it is necessary to recall Table 3.2 (p. 21), which shows that the index is more correlated to high-technology exports, manufactures imports and RD expenditure (in that order) than FDI in ows. Although this structure should be considered as 31
  • 41. Figure 4.7: Regression’s Residuals vs Innovation System Scores (Model 4) it provides insights about its constitution, the Innovation System Index, due to its nature (calculated by PCF Analysis), must be appreciated as a whole. Determining which of the indicators that belong to this index is more critical for growth falls out of the scope of this approach. The Governance Pillar Index returned signi cant coe cients less consistently as with the IS index: in columns (3), (4) and (5) with respective p-values lower than 5%, 5% and 1%. Still, the unchanging positive sign supports the theory; ‘good governance’ has been relevant for higher growth. The outcome of the coe cients for the ICT and Economic Incentives indicators are more paradoxical, being both signi cantly negative in some iterations: (2) and (4) for the former and the latter in all but (4). This is explained, to an extent, due to the e ect of conditional convergence. The regression table portrays how intro- ducing the variable log (GDP per capita) in (3) reduces the signi cance of both ICT and Economic Incentives indicators, the latter losing all explanatory power. Coher- ently, these two pillars are strongly correlated with GDP per capita, 0.84 and 0.69 respectively. Particularly for the ICT Index, it is worth to revisit Figure 4.5. The scatter plot shows how the slope of the tted line became atter over time. This adjustment is most likely explained by the characteristic of the ICT variables employed here. ‘Personal computers,’ ‘Internet access’ and ‘mobile phones and landlines’ are all technologies that mature and continue to fall in price, thus become more accessible 32
  • 42. to poorer countries. As follows, a more dynamic ICT Index that contains this e ect might be more suitable for the present approach. Education did not yield a signi cant result. There are two main possible expla- nations for this. First, the calculation of the variance in ation factors4 for each of the explanatory variables suggested the presence of multicollinearity. It does not appear to be so severe, as the signs of the coe cients in Table 4.1 seldom changed. However, it might have deterred the signi cance levels. The second issue is that the Education Pillar Index variables represent current investment and enrollment, which do not re ect so strongly in present growth, but rather have a more important e ect in future increments of GDP. Regarding the control variables, the signi cance of the f irst dummy variable’s negative coe cient indicates a strong and generalized trend of greater growth for the years [2003, 2007] in comparison to [1998, 2002]. Also, although in some iter- ations more so than others, there were meaningful negative coe cients for lngdpl, which support the theory of conditional convergence. Finally, the regional dum- mies had a more secondary role on the model. In column (4), although not yielding signi cant coe cients, they eliminated the explanatory power of the Economic In- centives proxy variable and reduced that of lngdpl. Thus, the standard regression of Model 4 suggests that a portion of the negative e ect from the Economic Incentives Pillar and the ‘conditional convergence’ observed in the previous columns is related to regional particularities. Interestingly, the stepwise regression in column (6) traded lngdpl for the regional dummy variables; this trade-o , however, did not produce a big change the signi - cance and values of the other coe cients. For Latin America the Caribbean and East Asia Paci c no signi cant coe cients were produced, this suggests that the particularities of these regions are expressed by the proxy variables of the Inno- vation System and the Economic Incentives pillars, the latter containing the e ect of conditional convergence. Excluding the poorest 25% of countries in column (5) reduced the explanatory power of the Governance Index, suggesting its importance 4 ˆ vif (Bi ) = 1 2 where Ri corresponds to the R-squared of the OLS regression in which the 1−Ri , 2 ˆ explanatory variable associated with Bi becomes the explained variable, as a function of all the other explanatory variables of the original model. A large R-squared suggests a high goodness of t and so, in this case, multicollinearity in the original model. As the R-squared increases, so does vif . The “rule of thumb” states that if the vif for a particular explanatory variable exceeds ve, multicollinearity is present. 33
  • 43. in the excluded countries. To check for the consistency of the constructed KE Pillar Indices, the economet- ric analysis is replicated with GDP per capita as the explained variable. The set of regressions is included in Appendix C. First of all, it is necessary to highlight the well documented likelihood for the results with respect to GDP per capita to su er from endogeneity. With this setup, it is di cult to know whether the KE Pillar In- dices a ect the levels of GDP per capita, or if the relationship is opposite e.g. richer countries can a ord better public education. In the previous regressions no evidence of endogeneity was found i.e. the tests did not show that fast growth rendered better performance of the pillars. Even though a causal relationship cannot be identi ed in the regressions with GDP per capita as the dependent variable, these illustrate a positive connection between all KE Pillars. In this chapter, the coe cients of models derived from the KE Framework were estimated using the Pillar Indices constructed in Chapter 3. The Innovation System Index, as a function of inward FDI, manufactured goods imports, high-technology exports and RD expenditure, was found to have promoted economic growth dur- ing [1998, 2007] in the 75 sampled countries. This relationship reached con dence levels lower than 1% when taking into account time e ects, conditional conver- gence and regional particularities. Furthermore, under this approach the ‘good gov- ernance’ index displayed a similar positive e ect, albeit less prominently. These results are concurrent to those of [Fagerberg and Srholec, 2007] which also calcu- lated innovation system and governance indices, albeit using di erent variables, complementary factors for economic growth (i.e. the political system and openness) and other methodological di erences as to this dissertation. For the remaining KE Pillars, the evidence was inconclusive. The constructed NSI performance index is an insightful proxy variable for ex- plaining how economies have grown between. This evidence is pertinent for an developing country like Colombia, whose policymakers have traditionally focused KE e orts towards the Economic Incentives Pillar and neglected the consolidation of the Innovation System Pillar. The following chapter explores the signi cance of the exposed link, between NSI performance and GDP growth, for this particular country. 34
  • 44. Chapter 5 An Overview of the Colombian National System of Innovation point towards a positive e ect of NSI performance on economic growth. R What can can be inferred from this particular link, in favor of policy-making? This chapter focuses on answering this question. Due to the fact that each NSI re- lies heavily upon a particular context, a speci c country is chosen as a case-study. Colombia’s NSI is to be analyzed using the considered theoretical background and the gathered evidence. First of all, the general circumstances in which the system began to be recognized as a public institution are described. Secondly, attention is given to various studies which characterize the present state of Colombia’s NSI. Thirdly, the current policy framework and instruments are described, in order to grasp the system’s outlook. Finally, based on the revision of this case-study, recom- mendations are provided in relation to the observed link between NSI performance and economic growth. 5.1 Background Since 1991, Colombia has decidedly shifted into a full market-oriented economic ap- proach. The Import Substitution Industrialization model, commonly seen through- out the region, was discarded in the mid-1970s and full edged liberalization was gradually undertaken. In 1999 the country experienced a recession, caused by the generalized capital out ows experienced across the developing world at that time, which was aggravated in Colombia by an internal mortgage crisis. Although the downturn worsened the country’s poverty gures, it was followed by a recovery 35
  • 45. stage i.e. the economy experienced accelerated growth between 2002 and 2007; for this period, the increase of real GDP averaged 5.32%. This was mainly due to more favorable economic conditions abroad and a series of policies that enhanced the Colombian business climate, perhaps the most signi cant one being the sus- tained progress in domestic security. Nevertheless, with the current global economic downturn, the most recent gures are rather timid. Colombia’s economy grew only 2.53% in 2008 and the following year GDP growth was practically nonexistent. To enter the globalized economy, the Colombian State tore down tari barriers and other protectionist measures. This new approach to international trade, while rightfully seeking to improve domestic productivity levels, a ected many Colom- bian rms which could not compete with the foreign companies that entered the country. Policymakers, aware of this, have appointed export promotion mecha- nisms, which mainly seek to improve the country’s level of competitiveness. The current policy approach to competitiveness can be roughly understood by two doc- uments written in the second half of the 2000s by the Consejo Nacional de Política Económica y Social (C ; National Council for Socio-economic Policy), a gov- ernmental institution which provides the framework for Colombia’s development policies. The following extracts are representative of their respective documents. “...[in Colombia] a series of measures and projects must be estab- lished and carried out in order to advance competitiveness in interna- tional markets. These measures may go from the construction and the improvement of the physical infrastructure or the training of the labor force, to the reorganization of institutions or the eliminations of [bureau- cratic] procedures. All these projects [...] seek to eliminate the obstacles faced by the productive sector during its operation...” Source: translated from [Mincomercio and DNP, 2004]. “A nation’s competitiveness is de ned as the degree to which a coun- try can produce goods and services capable of competing successfully in globalized markets and, at the same time, improve the population’s in- come conditions and the quality of life. Competitiveness is the result of the interaction of multiple factors related to the conditions faced by business which a ect their performance e.g. infrastructure, human re- sources, science and technology, institutions, the macroeconomic envi- ronment, and productivity.” Source: translated from [Mincomercio and DNP, 2006]. 36
  • 46. These two extracts from the policy documents show that, in Colombia, there has been a bias towards exogenous mechanisms for increasing competitiveness i.e. by infrastructure improvements (e.g. highways, access to utilities and airports) and through the re nement of institutions (e.g. a more e cient bureaucracy, decreasing corruption and a more e ective enforcement of legal contracts). In both documents, the role of innovation is seldom mentioned and the respective policy guidelines are practically absent. As follows, the background of Colombia’s approach to compet- itiveness is, in terms of the KE Framework, focused in improvements of the Eco- nomic Incentives, Governance, ICT and Education Pillars, leaving the Innovation System Pillar as secondary at best. Contrary to other countries under the export-promotion scheme, Colombian pol- icy has lagged in addressing the role innovation has as an endogenous mechanisms that favors competitiveness. A domestically developed innovative product or ser- vice will (by de nition) outperform its competition, while implicitly contributing towards technological learning inside the rm and thus enhancing its productivity1 . The fact that innovation has been belatedly adopted as a State policy has brought upon several weaknesses in the Colombian NSI, as reported by the several studies included in the following section. In Colombia, the institutional framework for a NSI was introduced in 1995. It was preceded by the Sistema Nacional de Ciencia y Tecnología (SNCYT; National Science and Technology System), conceived in 1990. The SNCYT aimed to integrate a diversity of institutions which shared a common vision, mission and objectives. It included rms, universities, public research institutions, research centers, and tech- nological institutions, among other actors. The innovation system was born as a sub-system of the SNCYT, at a similar time as in other Latin American countries. Both systems are, in practical terms, considered as one: the Sistema Nacional de Ciencia, Tecnología e Innovación (SNCTI; National Science, Technology and In- novation System)2 [Monroy Varela, 2006]. To conserve the terminology employed in this dissertation, the SNCTI will be referred as the Colombian NSI, henceforth CNSI. The public organization in charge of devising and executing policies that aim to structure and strengthen the system is C , an institution that seeks to 1 For a detailed description of this process see [Kim, 1997], pp. 88-90. 2 This is due to the fact that both systems are fundamentally composed by the same actors, share common concepts, basic strategies and challenges. 37
  • 47. lead the generation an utilization of knowledge in favor of socio-economic develop- ment in Colombia. C de nes the CNSI as follows. “The National System of Science, Technology and Innovation is an open system composed by policies, strategies, programs, methodologies and mechanisms in favor of the management, promotion, nancing, pro- tection and di usion of scienti c investigation and technological inno- vation, as well as the public, private or mixed organizations which carry out or promote the development of scienti c, technology and innovation activities.” Source: translated from [Colciencias, 2010]. According to [Monroy Varela, 2006];[Forero, 2000], the CNSI has faced di cul- ties since its conception. The most conspicuous have been: an unstable and weak budget, inadequate policy formulation (e.g. shortsightedness and vertical decision- making), a lack of public directives, the social apathy towards the socio-economic value of scienti c research, a stagnant scienti c community, the phenomenon re- ferred as “brain drain,” and the disjunction among the system’s actors. The follow- ing sub-section presents representative studies which diagnose the CNSI’s weak- nesses. 5.2 Current Status Several studies have faced the task of diagnosing the CNSI. Three in-depth sur- veys are worth mentioning, as these reveal important aspects of the system. The Encuesta de Desarrollo Tecnológico en el establecimiento industrial colombiano (EDT; Technological Development Survey on the Colombian industrial settlement) began to be recorded since 1996 through the e orts of Colombia’s Departamento Nacional de Planeación (National Planning Department) and with the support of C . The introduction of an institutional NSI framework matched with an increased interest in the measurement and tracking of the industry’s techno- logical development. For this purpose, the rst EDT was conducted in 1996 and these e orts were continued with iterations of the survey in 2003, 2005 and 2006. However, the methodology has been constantly amended. While the survey meets higher standards today and has a considerably larger sample size (6080 rms in 2006 [DANE, 2010] as to 885 in 1996 [DNP, 1997]), its more recent results are hard to 38
  • 48. be compared with those of 1996 [Vargas Pérez and Malaver Rodríguez, 2004]. One telling indicator which has remained practically unchanged is the percentage of in- novative rms3 : 11.3% (1996) and 11.8% (2003)4 . [Monroy Varela, 2006] is another survey-based study, which queried about the articulation of the CNSI. It found that while the knowledge-producing agents of the system knew about the existence of a CNSI framework, business owners were mostly unaware of it. This feature suggests the system’s bias towards the supply- side of knowledge, which is particularly problematic for the development of inno- vations i.e. indigenous knowledge and technologies are not being incorporated into Colombian products or services. The survey shows that this bias is re ected on the disjunction of the system: agents primarily interact with others of the same type. The third survey reported in [Malaver Rodríguez and Vargas Pérez, 2006] was carried out in 2005, aiming to advance the results of the EDT albeit delimited to Bo- gotá (the capital city district) and its circumvent department of Cundinamarca (one of Colombia’s thirty-two administrative divisions). It found that less than a third of the 400 sampled rms in the aforementioned region interacted with other agents of the CNSI in order to complement their e orts in producing innovations and improve their technological capabilities. Furthermore, less than 10% of the companies turned to the public institutions of the CNSI. However, the low share of rms that engaged the system found few di culties in doing so, meaning that the problem lies in the absence of associations rather in than the links themselves. Once again the cultural problem is brought to focus, as 44% of the sampled rms considered unnecessary to reach the CNSI, while 19% did not innovate because it was not considered pro table or signi cant to do so. [Torres et al., 2007] gathered data through the Premio Innova (Innova Award), a technological innovation competition for Colombian small and medium enterprises (SMEs), created in 2004 as a mechanism to promote innovation in technological start-ups5 . The work analyzes how the participating companies are seeking to in- novate and typi es the interactions between the CNSI agents. The study is based 3 The survey classi es a rm as innovative if, at the corresponding period of inquiry, it has launched at least one good or service signi cantly improved by international standards [DANE, 2010]. 4 As reported by [Vargas Pérez and Malaver Rodríguez, 2004] and [DANE, 2010]. 5 The award aims to bring attention to Colombian innovations, benchmark technological capa- bilities among di erent sectors, and to encourage an innovation culture in SMEs. 39