Skip to main content

Advertisement

SpringerLink
Log in

The interactions between national systems and sectoral patterns of innovation

A cross-country analysis of Pavitt’s taxonomy

  • Regular Article
  • Published:
Journal of Evolutionary Economics Aims and scope Submit manuscript

Abstract

Do national and sectoral innovation systems interact with each other? The paper explores this unexplored question by carrying out a cross-sector cross-country analysis of European systems of innovation in the 1990s. The empirical study takes Pavitt’s (Res Policy 13:343–373, 1984) taxonomy as a starting point, and it investigates the cross-country variability of Pavitt’s sectoral patterns of innovation. The analysis leads to three main results. First, the various technological trajectories show large differences across countries, due to the influence of national innovation systems. Second, there is evidence that the interaction between national systems and sectoral patterns of innovation constitutes an independent source of variability in the sample. Third, the analysis leads to the identification of eight sector- and country-specific technological trajectories in European manufacturing industries, and, based on that, proposes a refinement of Pavitt’s taxonomy. The refined taxonomy, in a nutshell, suggests that sectoral systems must be supported by and interact with their respective national systems in order to become industrial leaders.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2

Similar content being viewed by others

Notes

  1. Due to some missing values for some of the variables for Germany and Spain, the results of the test presented in this section do not include these countries, and therefore refer to a sample of eight countries.

  2. These are the so-called Pavitt’s “measured characteristics” (see Tables 1 to 3 of his 1984 article).

  3. For a complete list of sectors included in each category of the taxonomy, see Appendix 1.

  4. Equations 1 and 2 are nonlinear, and require an iterative solution. This is based on the method of maximum likelihood. The solution is commonly found by the Newton’s method in a relatively small number of iterations.

  5. The choice of the baseline category does not affect the results of the MNL test, so that any other category could have been chosen instead.

  6. In a MNL model, each estimated coefficient measures the proportional change in the ‘log of the odds-ratio’ of the dependent variable when the kth regressor changes by one unit. In other words, if the estimated coefficient β k is positive (negative), the likelihood of that response category will increase (decrease) by a factor of β k for any unit change of the kth regressor.

  7. The relative position of different European countries in the various categories of Pavitt’s taxonomy will be analyzed in further detail in Section 4.

  8. A related exercise has recently been presented by Evangelista and Mastrostefano (2006). Their paper analyzes the extent of country-, sector- and firm-specific sources of variability in a cross-section of manufacturing industries in Europe. However, their exercise differs from the one presented here in two main respects. First, the present paper focuses on the cross-country variability in relation to Pavitt’s sectoral groups. Second, our analysis of variance does not only focus on the country- and sector-specific components, but it does also consider an interaction term between these factors.

  9. A recent paper by Dopfer et al. (2004) discusses the interactions between the micro, meso and macro levels of analysis in evolutionary economics. The theoretical discussion presented there constitutes an interesting and general framework to link the various levels of analysis in evolutionary theorizing. Differently from the use made in their paper, however, in the present work the term meso refers to the sectoral level of analysis, i.e. the study of the patterns and evolution of different industries.

  10. A specific example of this in relation to the Norwegian case is discussed by Narula (2002).

  11. In this cluster analysis, manufacturing sectors have been grouped according to the four categories of Pavitt’s taxonomy, so that the results presented in this section refer to a sample of 40 observations (i.e. four industry groups in ten European countries).

  12. The CIS-SIEPI database has been constructed as a result of the EU-funded SIEPI project (“The Structure of Innovation and Economic Performance Indicators”). The dataset contains CIS2 data at a higher level of sectoral disaggregation because the data have been obtained directly from national sources (i.e. from the statistical offices of the ten countries included in the database).

References

  • Andersen ES (1992) Approaching national systems of innovation from the production and linkage structure. In: Lundvall BA (ed), National systems of innovation: towards a theory of innovation and interactive learning. Pinter, London

    Google Scholar 

  • Archibugi D (2001) Pavitt’s taxonomy sixteen years on: a review article. Econ Innov New Technol 10(5):415–425

    Article  Google Scholar 

  • Archibugi D, Pianta M (1994) Aggregate convergence and sectoral specialization in innovation. J Evol Econ 4:17–33

    Article  Google Scholar 

  • Balzat M, Hanusch H (2004) Recent trends in the research on national innovation systems. J Evol Econ 14:197–210

    Article  Google Scholar 

  • Balzat M, Pyka A (2006) Mapping national innovation systems in the OECD area. Int J Technol Glob 2(1–2):158–176

    Google Scholar 

  • Begg I, Dalum B, Guerrieri P, Pianta M (1999) The impact of specialization in Europe. In: Fagerberg J, Guerrieri P, Verspagen B (eds) The economic challenge for Europe—adapting to innovation based growth. Edward Elgar, Cheltenham

    Google Scholar 

  • Breiman L, Friedman J, Olshen R, Stone C (1984) Classification and regression trees. Wadsworth, Belmont

    Google Scholar 

  • Castellacci F (2007a) Evolutionary and new growth theories. Are they converging? J Econ Surv 21(3):585–627

    Article  Google Scholar 

  • Castellacci F (2007b) Technological regimes and sectoral differences in productivity growth. Ind Corp Change 16(6):1105–1145

    Article  Google Scholar 

  • Castellacci F (2008) Technological paradigms, regimes and trajectories: manufacturing and service industries in a new taxonomy of sectoral patterns of innovation. Res Policy 37(6–7):978–994

    Article  Google Scholar 

  • Cefis E, Orsenigo L (2001) The persistence of innovative activities: a cross-countries and cross-sectors comparative analysis. Res Policy 30:1139–1158

    Article  Google Scholar 

  • Dopfer K, Foster J, Potts J (2004) Micro-meso-macro. J Evol Econ 14:263–279

    Article  Google Scholar 

  • Dosi G (1988) Sources, procedures, and microeconomic effects of innovation. J Econ Lit XXVI:1120–1171

    Google Scholar 

  • Edquist C (1997) Systems of innovation: technologies, institutions and organisations. Pinter, London

    Google Scholar 

  • Edquist C (2005) Systems of innovation: perspectives and challenges. In: Fagerberg J, Mowery DC, Nelson RR (eds) The Oxford handbook of innovation. Oxford University Press, Oxford

    Google Scholar 

  • Evangelista R (1999) Knowledge and investment. The sources of innovation in industry. Edward Elgar, Cheltenham

    Google Scholar 

  • Evangelista R, Mastrostefano V (2006) Firm size, sectors and countries as sources of variety of innovation. Econ Innov New Technol 15(3):247–270

    Article  Google Scholar 

  • Fagerberg J (1995) User–producer interaction, learning and comparative advantage. Camb J Econ 19:243–256

    Google Scholar 

  • Fai F, von Tunzelmann N (2001) Industry-specific competencies and converging technological systems: evidence from patents. Struct Chang Econ Dyn 12:141–170

    Article  Google Scholar 

  • Freeman C (1987) Technology policy and economic performance: lessons from Japan. Pinter, London

    Google Scholar 

  • Laursen K (2000) Do export and technological specialisation patterns co-evolve in terms of convergence or divergence? Evidence from 19 OECD countries, 1971–1999. J Evol Econ 10:415–436

    Article  Google Scholar 

  • Laursen K, Meliciani V (2000) The importance of technology based inter-sectoral linkages for market share dynamics. Weltwirtscaftliches Archiv 136(4)

  • Laursen K, Meliciani V (2002) The relative importance of international vis-à-vis national technological spillovers for market share dynamics. Ind Corp Change 11(4):875–894

    Article  Google Scholar 

  • Laursen K, Salter A (2005) The fruits of intellectual production: economic and scientific specialisation among OECD countries. Camb J Econ 29:289–308

    Article  Google Scholar 

  • Lundvall BA (1992) National systems of innovation: towards a theory of innovation and interactive learning. Pinter, London

    Google Scholar 

  • Lundvall BA, Borrás S (2005) Science, technology and innovation policy. In: Fagerberg J, Mowery DC, Nelson RR (eds) The Oxford handbook of innovation. Oxford University Press, Oxford

    Google Scholar 

  • Malerba F (1993) The national system of innovation: Italy. In: Nelson R (ed) National innovation systems: a comparative analysis. Oxford University Press, New York and Oxford

    Google Scholar 

  • Malerba F (2005) Sectoral systems: how and why innovation differs across sectors. In: Fagerberg J, Mowery DC, Nelson RR (eds) The Oxford handbook of innovation. Oxford University Press, Oxford

    Google Scholar 

  • Malerba F (2006) Innovation and the evolution of industries. J Evol Econ 16(1–2):3–23

    Article  Google Scholar 

  • Malerba F, Orsenigo L (1995) Schumpeterian patterns of innovation. Camb J Econ 19:47–65

    Google Scholar 

  • Malerba F, Montobbio F (2003) Exploring factors affecting international technological specialization: the role of knowledge flows and the structure of innovative activity. J Evol Econ 13:411–434

    Article  Google Scholar 

  • Marsili O, Verspagen B (2002) Technology and the dynamics of industrial structure: an empirical mapping of Dutch manufacturing. Ind Corp Change 11(4):791–815

    Article  Google Scholar 

  • Miles I (2005) Innovation in services. In: Fagerberg J, Mowery DC, Nelson RR (eds) The Oxford handbook of innovation. Oxford University Press, Oxford

    Google Scholar 

  • Mowery D, Nelson R (1999) The sources of industrial leadership. Cambridge University Press, Cambridge

    Google Scholar 

  • Mowery D, Sampat B (2005) Universities in national innovation systems. In: Fagerberg J, Mowery DC, Nelson RR (eds) The Oxford handbook of innovation. Oxford University Press, Oxford

    Google Scholar 

  • Murmann JP, Homburg E (2001) Comparing evolutionary dynamics across different national settings: the case of the synthetic dye industry, 1857–1914. J Evol Econ 11:177–205

    Article  Google Scholar 

  • Narula R (2002) Innovation systems and ‘inertia’ in R&D location: Norwegian firms and the role of systemic lock-in. Res Policy 31(5):795–816

    Article  Google Scholar 

  • Nelson RR (ed) (1993) National innovation systems: a comparative analysis. Oxford University Press, New York and Oxford

    Google Scholar 

  • Nelson R, Winter S (1977) In search of a useful theory of innovation. Res Policy 6:36–76

    Article  Google Scholar 

  • Nelson R, Winter S (1982) An evolutionary theory of economic change. The Belknap Press of Harvard University Press, Cambridge

    Google Scholar 

  • Pavitt K (1984) Sectoral patterns of technical change: towards a taxonomy and a theory. Res Policy 13:343–373

    Article  Google Scholar 

  • Peracchi F (2001) Econometrics. Wiley, Chichester

    Google Scholar 

  • Porter M (1990) The competitive advantage of nations. Macmillan, London

    Google Scholar 

  • Powell W, Grodal S (2005) Networks of innovators. In: Fagerberg J, Mowery DC, Nelson RR (eds) The Oxford handbook of innovation. Oxford University Press, Oxford

    Google Scholar 

  • Scott Long J (1997) Regression models for categorical and limited dependent variables. Advanced quantitative techniques in the social sciences, vol. 7. Sage Publications

  • Tidd J, Bessant J, Pavitt K (1997) Managing innovation—integrating technological, market and organizational change. Wiley, Chichester

    Google Scholar 

  • Verspagen B (2008) Challenged leadership or renewed vitality? The Netherlands. In: Edquist C, Hommen L (eds) Small Country Innovation Systems. Globalization, Change and Policy in Asia and Europe. Edward Elgar, Aldershot

    Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Department of International Economics, Norwegian Institute of International Affairs (NUPI), PO Box 8159, Dep. 0033, Oslo, Norway

    Fulvio Castellacci

Authors
  1. Fulvio Castellacci
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Fulvio Castellacci.

Appendices

Appendix 1: The dataset and the sectoral classification

The empirical analysis carried out in this paper has made use of the CIS-SIEPI database. This contains data from the Second Community Innovation Survey (1994–1996) on innovative activities of manufacturing industries in ten European countries (Germany, Spain, France, Italy, Netherlands, Norway, Portugal, Sweden, UK and Austria). Compared to other CIS-related data sources (e.g. Eurostat), the CIS-SIEPI database contains data at a higher level of sectoral disaggregation (22 manufacturing industries, instead of 10 as in most other sources), and it therefore makes it possible to obtain a more accurate picture and to shed new light on sectoral patterns of innovation in Europe.Footnote 12

In the empirical analysis, the 22 manufacturing industries have been assigned to the four categories of Pavitt’s taxonomy by following Pavitt’s (1984) original paper, as well as other subsequent empirical analyses that have made use of the taxonomy (Begg et al. 1999; Laursen and Meliciani 2000; Marsili and Verspagen 2002). The sectoral classification used in the paper is then the following.

  • Specialized suppliers: Machinery and equipment; medical and optical precision instruments.

  • Science-based: Electrical; radio and TV; office, accounting and computing; chemicals; coke, refined petroleum products and nuclear fuel.

  • Scale intensive: Motor vehicles and trailers; other transport; rubber and plastics; basic metals; fabricated metal products; food and beverages.

  • Supplier-dominated: Textiles; wearing; leather and footwear; wood and related; pulp and paper; printing and publishing; other non-metallic mineral products; furniture; recycling.

The industries assigned to each of the four categories are consistent with previous works (see Laursen and Meliciani 2000, Appendix 1). The only exception refers to the sector coke, refined petroleum products and nuclear fuel. We have decided to include it in the science-based category because: (1) there exists a significant scientific component in the production of nuclear fuel; (2) the industry is characterized by large firms (as measured by our variable SIZE), which is one of the main characteristics of science-based sectors; (3) the interactions between University and innovative firms are strong (in terms of our variable SCIENCE); and (4) the industry is closely related and to some extent similar to the science-based chemicals sector (see Marsili and Verspagen 2002).

Appendix 2: The CART methodology

The classification and regression tree algorithm (CART) is a flexible non-parametric method of multivariate analysis (Breiman et al. 1984). It can be used for classifying a set of N cases into J categories based on a vector X of characteristics, or, alternatively, for predicting to which category a case belongs based on its vector X of characteristics.

The dependent variable in CART is categorical (j = 1 to J), while the explanatory variables XBiB (i = 1 to M) can be both categorical and scale. The general idea of CART is to construct a hierarchical classification of cases, where each step of the algorithm splits a group of cases into two sub-groups (nodes) based on one single predictor variable XB. The CART algorithm can be described as follows.

  1. (1)

    The initial node (root node) comprises all N cases in the sample. It is split into two nodes, NB1B and NB2B, on the basis of the predictor variable XBiB that makes it possible to achieve the best split (searching among all possible splits, and all predictor variables used as inputs in the analysis). The criterion to search for the best split is to reduce the node’s impurity measure, i.e. to reduce the number of cases not belonging to a given category. A node is pure when all cases belonging to it refer to the same category. The two most used criteria for splitting are the Gini and the Twoing methods. The results presented in Section 4 are based on the former.

  2. (2)

    The same splitting rule is subsequently applied to all successive non-terminal nodes. A node is terminal when it is not possible to improve the misclassification rate by splitting it further into two subnodes. The resulting tree, TBmaxB, tends to be very large, because no cost for splitting has initially been specified. This means that splitting cases is costless, and that the tree will thus tend to have many branches and several terminal nodes.

  3. (3)

    The tree TBmaxB, therefore, does not provide either a correct idea of the right-sized tree, or an accurate and honest estimate of its misclassification rate. For this reason, the tree must be pruned, i.e. the branches that are superfluous must be cut. This is achieved in two ways. First, the algorithm specifies costs associated with each successive split, so that the higher the number of splits, the greater the overall cost. Second, the CART selects the best pruned subtree among all possible pruned subtrees. This selection is obtained by using two alternative methods: (1) test sample estimates, where a new sample is used to assess the precision of each subtree obtained through the analysis of the learning sample (this is the preferred method when a large sample is considered); (2) v-fold cross-validation, where the learning sample is partitioned into V equal parts, and the vBthB fraction is used to evaluate the precision of the (1-v)BthB larger part (this method leads to better results in relatively small samples, and we have therefore used that in our analysis). Both criteria lead to an estimation of the number of misclassified cases, so that the best pruned subtree is the one that minimizes the estimated misclassification rate.

The classification tree diagram reported in Fig. 2 (Section 4) is the final result of the CART algorithm, and represents, therefore, the best pruned subtree. The right tree size, i.e. the number of branches and terminal nodes described in Table 4, has therefore been found out endogenously by the algorithm through an extensive examination of all possible splitting conditions at each step, and all possible pruned subtrees.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Castellacci, F. The interactions between national systems and sectoral patterns of innovation. J Evol Econ 19, 321–347 (2009). https://doi.org/10.1007/s00191-008-0113-9

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00191-008-0113-9

Keywords

JEL Classification

Search

Navigation