Innovation and Economic Growth

TABLE OF CONTENTS

1 Introduction
1.1 Research questions
1.2 The Term 'Innovation'
1.3 Theoretical Background

2 Principal Earlier Studies
2.1 Introductory Discussion
2.2 Four Empirical Studies

3 Comparison of Technology Stocks among 15 OECD Countries
3.1 Introduction
3.2 Interpretation
3.3 A European Technology Stock Map

4 Econometric Analysis
4.1 Introduction
4.2 The Model
4.3 Econometric Challenges
4.4 Interpretation

5 Innovation Policies in Europe
5.1 Introduction
5.2 Weaknesses of Europe’s Innovation Activities
5.3 Diversity and Convergence
5.4 Policy Actions Undertaken in Europe

6 Conclusions

Appendix I - Data Sources
Output (GDP)
Labour Stock
Capital Stock
Technology Stock

References

ACKNOWLEDGEMENTS

As with every piece of work I have undertaken, there are people I especially want to thank for their support. For the support in writing this thesis as well as in motivating me to do further academic research I would first like to thank Dr. B. L., my content supervisor, who provided me with excellent support over the whole period of this thesis. He was always a competent and very helpful person I could consult in case of problems and questions. Second, I want to thank Dr. E. L., my process supervisor, for accompanying me on the first steps of this thesis and the research proposal. The meetings with the small Bachelor thesis group supervised by Erik at the beginning of this semester were very supportive and interesting. Third, I am grateful to my friends Anselm Mattes, Michael Fuller, Arne Heutink and Ipek Özer who reviewed this thesis and provided me with useful comments. Fourth, the whole Bachelor thesis group has to be thanked as everyone provided interesting comments and helped to support each other.

1 Introduction

1.1 Research questions

The fact that some countries show higher growth rates than others has long been observed in the global economy. Even within the OECD area, where countries are regarded as being relatively homogeneous, growth levels differ remarkably. This leads inevitably to the question how different growth rates can be reconciled and what factors are responsible for different growth rates. The purpose of this thesis is to examine one of the factors, besides the classical input factors labour and capital, that is often regarded to play a significant role in economic growth namely innovation.

After a theoretical introduction in the first section of this thesis, the second section reviews principal earlier studies about the effect of innovation on economic growth to provide a methodological overview. In section three, research and development (R&D) expenditures (taken as a proxy for innovation) among 15 OECD countries are evaluated in the period 1973 - 1998 to analyse how the technology stock of these countries developed relative to their GDP growth and to answer the question which countries performed relative well with respect to innovation and which countries lag behind. Section four analyses the impact of innovation on economic growth by econometrically testing a neoclassical growth model where innovation is endogenized. The purpose of this econometric analysis is to show how much economic growth can be accounted to innovation when compared to the classical input factors capital and labour. Section five focuses on recent policy actions undertaken in the European Community to support innovation. The section analyses possible policy responses to the innovation phenomenon with respect to the European Community. Section six contains brief concluding remarks.

1.2 The Term 'Innovation'

Innovation is a multi-dimensional phenomenon, not easy to define. Therefore, it became necessary to find a definition for fitting to the content of this thesis. I opted for the definition of the Oslo Manual that was compiled by the OECD (1992, 1996), and that defines innovation as: “The market introduction of a product (good or service) new or significantly improved, or the introduction of new or significantly improved processes, based on new technological developments, new combinations of existing technologies or use of other type of knowledge acquired."

The European Commission's Green Paper on Innovation (1995) provides some examples, including the development of vaccines and medicines, improved safety in transport, (ABS, airbags), easier communications (mobile phones, videoconferencing), more open access to know-how (CD-ROM, multimedia), new marketing methods (home banking), better working conditions, more environment-friendly techniques, and more efficient public services.

1.3 Theoretical Background

Starting in 1950s it was Robert Solow (1957) who was the first to include an exogenous term labelled "technical change^[1] " in the neoclassical growth model that beforehand included only labour and capital as explanatory variables for economic growth. Motivated by the objective to describe an elementary way to segregate variations in output per capita due to technical change from those due to changes of the availability of capital, Solow's analysis came to the conclusion that technical change was responsible for the large majority (87.5 per cent) of economic growth. In Solow's model technology was interpreted as a "free good", meaning that it is accessible for everybody and everywhere free of charge. This also implied that technology and innovation were equally available on a global level and that they were regarded as a worldwide commodity.

While more recent studies gradually lowered Solow's conclusion that technical change was responsible for almost all economic growth, these studies still found a significant role for technological change (for example Jorgenson, 1990). However, the fact that according to the theory, all countries would converge towards the same level of productivity because of the public good character of innovation could not be observed in the real world.

New literature such as Römer (1986) or Lucas (1988) endogenized what was considered to be the underlying source of sustained growth in per-capita income, namely the accumulation of knowledge. As Aghion and Howitt (1992) describe, there are many channels through which societies accumulate knowledge, including formal education, on-the-job training, basic scientific research, learning by doing, process innovation, and product innovation. Consequently some neoclassical theorists followed a new path by considering innovation as an output of a separate endogenous technology-sector in the production function, supplying other sectors with new technology (see Romer, 1990). Innovation was considered as a nonrival, partially excludable good with a monopolistic competition equilibrium. Although commonly labelled "new growth theories" Fagerberg (1994) notes that some efforts to endogenize technological progress had already been undertaken in the 1960s (Denison 1962, 1967) long before these model caught new attention.

With respect to country differences, scholars developed models that accounted for technological spillovers and their effect on economic growth (Coe and Helpman, 1995). The idea behind technological spillovers was that part of technological knowledge is purchased on world markets whereas other parts are purchased on domestic markets. As far as the world market is involved technology spillover effects exist in countries affecting domestic economic growth. Innovation is then no longer a good equally available in all places of the world as in Solow's model but rather a good that involves learning effects and dissemination barriers across countries.

This thesis endogenizes innovation in a neoclassical production function in Section 4, thereby following the new growth theories described above. No attempt is made to explain spillover effects among countries.

2 Principal Earlier Studies

2.1 Introductory Discussion

This chapter is supposed to provide a literature review about four important studies of innovation and economic growth to supply background knowledge as well as earlier estimation results. By focussing on these four studies, some boundaries should be mentioned that limit the scope of the thesis in general and the scope of this chapter in particular. First, all studies address a total economy level, thereby neglecting estimations on firm or industry level. For comprehensive summary of firm, industry, and total economy studies, the reader is referred to the extensive review of Cameron (1998). Second, parts of the four reviewed papers are only treated in a rudimentary way, i.e. special innovation characteristics like R&D spillovers or innovation convergence are not explained extensively. What all four studies have in common is the approximation of innovation by using R&D expenditures as starting point. However, as Verspagen (1996) points out, there are some drawbacks of this approach. First, R&D is basically an input indicator and might not tell everything about the associated outputs. This relates to the fact that R&D as well as innovation is essentially a search process having fundamental different characteristics as other input factors. As Teece (1996) notes, innovation is characterised by uncertainty, path dependency, and a cumulative nature to name only the most significant characteristics in this context. Moreover, Verspagen (1996) acknowledges that there are various aspects of innovation that are not measured by official R&D statistics, even in well defined statistical concepts such as the OECD Frascati Manual (1994). One alternative to R&D expenditures is the use of patents as a proxy for innovation. However, major drawbacks of this approach are different patenting propensities among sectors and the fact that patent law as well as patent criteria differ among national patent agencies.

2.2 Four Empirical Studies

To start in chronological order, Griliches (1973) was one of the first scholars who attempted to estimate a relationship between innovation and growth by using R&D expenditures. Griliches exclusively estimates figures for the United States and did not use time series data, i.e. his estimates correspond only to one particular year. He used a growth accounting framework measuring total factor productivity and paid special attention to the separation of private and public R&D expenditures. This distinction is important, because, as Griliches reasoned, most productivity related R&D expenditures are provided by the private sector. Public R&D expenditures are often undertaken by military bodies, not providing any productivity effect for the economy^[2]. Concerning the results of his study Griliches estimates for the year 1966 showed a total contribution of R&D to growth of half a per cent per year. Furthermore, Griliches provides an estimated social rate of return of 23.4 per cent for R&D activities. Although Griliches did not directly provide time series evidence, his results show some reasonable evidence for a link between R&D expenditures and growth and highlight the importance of separating public and private R&D expenditures. The study can therefore be seen as one of the important early studies with respect to R&D expenditures and economic growth on a country level.

Patel and Soete (1988) estimated the impact of R&D for five OECD countries for the period 1956 - 85. Their regression equation looked as follows:

illustration not visible in this excerpt

where Y was labour productivity, c was a constant, R was the R&D capital stock, t was time, and a as well as ß were elasticities.

This implied that the capital stock was replaced by the simple temporary time trend t. According to Patel and Soete, the reason for not implementing the capital stock in the regression model, was a lack of data relating to capital stock measurements. Before reporting the results of this study it should be mentioned that econometric models like the one used by Patel and Soete often have to be corrected for autocorrelation^[3]. As this will be explained more detailed in the econometric analysis part in Section 4 of this thesis, the reader should for the moment only keep in mind that all of the results of Patel and Soete had to be corrected for autocorrelation. The following table reports the results of the regression results of Patel and Soete (1988).

Table 1 (Patel and Soete, 1988): Labour productivity regression results 1956-85

illustration not visible in this excerpt

With respect to the R&D elasticities, it is firstly notable that all countries exhibit a significant elasticity in the study of Patel and Soete. Second, the estimated R&D elasticities differ significantly between countries, with Japan showing the highest elasticity and the United States showing the lowest elasticity of the countries observed.

The reader should also note that the [illustration not visible in this excerpt] for Japan is significantly lower than for other countries meaning that the model is less explanatory for Japan compared to the other countries. As Patel and Soete point out, their study provides useful hints and indications of presumed economic relationships that might however be largely obscured by the difficulties in approximating some of the most crucial variables, a limitation already indicated in the introductory discussion of this chapter.

Coe and Helpman (1995) provided estimates for international R&D spillovers as well as a cumulated rate of return for R&D activities in G7 countries. Although this thesis does not put much effort to explain R&D spillovers, the reader should still keep in mind that these spillovers are far away from being negligible and can contribute significantly to economic growth. Coe and Helpman estimated that about one quarter of total benefits of R&D investments in G7 countries accrues to its trading partners^[4]. As Coe and Helpman used a regression model that accounted for foreign R&D and for domestic R&D effects, they also provided regression estimates on the impact of domestic R&D that are comparable to the results of the other studies in this section and to the results of the econometric tests performed in this thesis. The results showed a 23 per cent rate of return of domestic R&D activities on an aggregated G7 basis for the year 1990. Similar to Griliches (1973), Coe and Helpman did not provide time series data for the countries. However, the result of 23 per cent return in G7 countries is very similar to the estimate of Griliches (1973) who estimated a 23.4 per cent return for the United States in 1973. Verspagen (1995) used a similar approach as Patel and Soete (1988) and investigated the three European countries Germany, France, and the United Kingdom as well as a "European" time series that he derived by simply adding the values of the three countries investigated. In addition to explaining labour productivity by the R&D stock and a time trend as done by Patel and Soete, he incorporated the variable k, the capital- labour ratio. His estimation model, thus, looked as follows:

illustration not visible in this excerpt

where Y was labour productivity, c was a constant, к was the capital-labour ratio, R was the knowledge stock, t was time, and a, ß, and γ were elasticities.

Verspagen used four different estimates for the knowledge stock, two patent stock lifetime variables and two R&D stock variables. The two patent stock variables differed in the patent lifetime where Verspagen first assumed a short 10 year lifetime and second a 15 year lifetime. By estimating the R&D stock, Verspagen first used a moving average estimate and second a perpetual inventory estimate. The moving average stock was defined as being equal to [illustration not visible in this excerpt] where Xt were R&D expenditures in period t. The perpetual inventory stock was defined as [illustration not visible in this excerpt] with the first value of the series calculated by multiplying the R&D expenditures of the first year with the factor 5. The perpetual inventory stock, as calculated by Verspagen therefore assumed a depreciation of 15 per cent per annum. As the perpetual inventory stock method will be used in the econometric tests performed in this thesis in Section 4, only the results of this stock estimate will be reported here. However, it should be acknowledged that the results of other variables taken for the estimation of the knowledge stock in some cases provided quite different results.

[...]

^[1] Solow labels "technical change" as a short-hand expression for any kind of shift in the production function. This is a much wider concept compared to the term "innovation", mainly dealt with in this thesis.

^[2] Consider the research on a military aircraft as an example where productivity effects for the rest of the economy are very limited.

^[3] Wooldridge (2003, p. 845) defines autocorrelation as the correlation between the errors in different time periods in a time series or panel model.

^[4] Coe and Helpman use bilateral import shares to calculate foreign R&D stocks. This implies that trade fosters R&D spillovers and that R&D spillovers grow in line with increasing trade.