Elsevier

Research Policy

Volume 30, Issue 7, August 2001, Pages 1139-1158
Research Policy

The persistence of innovative activities: A cross-countries and cross-sectors comparative analysis

https://doi.org/10.1016/S0048-7333(00)00139-6Get rights and content

Abstract

This paper examines the persistence of innovative activities at the firm level in a comparative perspective. A new data set is used composed of six panel data, one for each of the following countries: France, Germany, Italy, UK, Japan and the USA. For each country, we use data on patent applications to the European Patent Office in the period 1978–1993 by 1200–1400 manufacturing firms. Using a transition probability matrix (TPM) approach, we find evidence for the existence of persistence in innovative activities, although, it is not very high in the aggregate and it declines as time goes by. However, both great innovators and non-innovators have a high probability to remain in their state and persistent innovators originate a disproportionate share of innovative activities. In this sense, persistence in innovative activities is quite strong.

These tendencies apply to all countries considered here, although, clear country-specific properties are observed. Moreover, there is heterogeneity also across industrial and size classification. Intersectoral differences are invariant across countries, suggesting that persistence is (at least partly) a technology-specific variable. Persistence tends to increase with firm size, but the relationship between firms’ size and persistence is strongly-country-specific and it is not a simple one.

Introduction

In this paper, we ask whether innovative activities are persistent at the firm level. To this task, we use data on patent applications to the European Patent Office over the period 1978–1993 for six samples of firms in the following countries: France, Germany, Italy, Japan, UK and the USA.

The issue of persistence in innovative activities is relevant in the context of the discussion about the properties of the patterns of innovative activities. The absence of persistence is likely to be associated to “creative destruction” (or Schumpeter Mark I model), whereas, the existence of significant degrees of persistence would contribute to generate processes of “creative accumulation” (or Schumpeter Mark II model). A fortiori persistence in innovative activities may bear significant implications for both theory and policy-making.

The observation of persistence would lend some support to the “competence-based” theories of the firm at the microeconomic level (Nelson and Winter, 1982, Teece and Pisano, 1994)1 and to the theories of endogenous growth at the macroeconomic level. Conversely, the existence of persistence in innovative activities would weaken those interpretations of the processes of growth of firms, industries and countries (ranging from simple Gibrat-type processes to the models of the real business cycle) where dynamics is essentially driven by small uncorrelated shocks. More generally, understanding whether innovative activities are persistent or not at the firm level would constitute an important piece of evidence for finding and improving current theories of industrial dynamics and evolution, where some forms of dynamic increasing returns play a major role in determining degrees of concentration and its stability over time, rates of entry and exit and so forth (Klepper, 1996, Jovanovic, 1982, Ericson and Pakes, 1992, Hopenhayn, 1992, Nelson and Winter, 1982, Dosi et al., 1995, Malerba et al., 1997).2

However, very little is known about the relative empirical relevance of persistence in innovative activities. Cumulativeness — which is closely related but not identical to persistence — has been shown to constitute a fundamental property of innovative activities at the firm level by an enormous body of case-studies. Moreover, there is considerable evidence that firms’ innovative activities are cumulative in the sense that their technological specialisation tends to remain stable over prolonged periods of time (Granstand, Patel and Pavitt, 1997). Similarly, there is substantial evidence that the technological specialisation of countries tend to persist for prolonged periods of time (see, for instance, Archibugi and Pianta, 1992, Cantwell, 1996 and Patel and Pavitt (1992)). However, only few studies have tackled the issue of persistence directly.

Recently, a few studies have started to provide some interesting — and somewhat contrasting — results.3 Malerba and Orsenigo (1999) examine the patterns of innovative entry, exit and survival, using European Patent Office data for six countries. They find that innovative activities are characterised by high degrees of turbulence. The population of innovators changes substantially over time, through processes of entry and exit. A large fraction of new innovators is composed by occasional innovators. They also constitute a significant part of the whole population of innovators but a much lower share of the total number of patents at any given time. Only a fraction of entrants survive and succeed in remaining innovative after their first patent. When they do, however, their technological performance improves consistently in the following years.4 Moreover, large innovators tend to remain large for long periods of time. These results would seem to suggest that although, turbulence is a pervasive and quantitatively important phenomenon, nevertheless, innovative activities are generated — in general — by a relatively stable core of large (both in terms of patents and employees) and persistent innovators, who account for a very large share of total patents. Around this core, one observes a large turbulent fringe of small, occasional innovators, who often patent only once and cease to patent thereafter.

In another paper, Geroski et al. (1997) estimate a Proportional Hazards Model to model the probability that the spell of time in which a firm innovates will end at any particular time. Geroski et al. find little evidence of persistence at the firm level. Very few firms innovate persistently and this happens only after a threshold level of initial patenting which only very few firms ever reach. However, persistent firms account for a very large share of total patents in the sample.

Cefis (1996) uses instead, non-parametric techniques (transition probability matrices) on a sample of patents applied for by 577 UK firms. She finds little persistence in general, but also strong persistence among the greatest and the smallest innovators. Moreover, she finds substantial heterogeneity in the degree of persistence across sectors and firms’ size.

In the present paper, the Cefis’ approach is used and improved at least in two senses. First, we examine six countries, instead of just one (UK). Second, we focus explicitly on the differences in persistence across industries and firms’ size. The reasons for a cross-country analysis are straightforward. Differences in persistence across countries are to be expected, in view of the substantial heterogeneity in the institutional set-ups, in the technological specialisation, etc. of different countries.

Similarly, investigating for differences across firms’ size classes is a quite natural extension of conventional tests of the Schumpeterian hypothesis.

Cross-sectors analysis is less obvious, though. Whilst some cross-industries differences are to be expected, the issue here concerns whether these differences are systematic across countries or not. In other words, we are interested in determining if persistence in innovative activities is technology- or industry-specific.

In previous papers (Malerba and Orsenigo, 1995, Malerba and Orsenigo, 1996, Malerba and Orsenigo, 1999, Breschi et al., 2000, Malerba et al., 1997), one of us provided empirical evidence that different technological classes exhibit different patterns of innovative activities, one resembling the creative destruction model, the other the creative accumulation model, with little variation across countries. It was then showed that the patterns of innovative activities are linked to the nature of the relevant technological regime, defined by opportunity and appropriability conditions, degrees of cumulativeness of technological advances, and the nature of the knowledge base (for further discussions of the notion of technological regimes, see, e.g. Nelson and Winter, 1982, Winter, 1984, Dosi, 1988, Malerba et al., 1997). These results suggest that it might be misleading to derive strong conclusions from aggregate data. Rather, we expect innovative activities to be systematically characterised by different degrees of persistence across technologies and industries, as well as across countries.

Two caveat are necessary before beginning our discussion.

First, we do not investigate the sources of persistence. Indeed, persistence may arise from very different mechanisms and it can be analysed at different levels of aggregation. Persistence may be the outcome of “success-breeds-success” processes like those used in the Nelson and Winter (1982) models: innovative success yields profits that can be reinvested in R&D, thereby, increasing the probability to innovate again. Persistence may also be the outcome of the intrinsically cumulative nature of learning processes (Rosenberg, 1976, Nelson and Winter, 1982). The generation of new knowledge builds upon what has been learned in the past, not only in the sense that past knowledge constrains current research, but also in the sense that knowledge generates questions which, in turn, generate new research. Moreover, research is often characterised by dynamic increasing returns, in the form of learning-by-doing and learning-to-learn, and today’s research generates tomorrow’s new opportunities (Klevorick et al., 1993, Cohen and Levinthal, 1989).5

Innovative persistence may also derive from organisational features at the firm level, ranging from the establishment of R&D facilities at a fixed cost — which then produce a relatively stable flow of innovations — to firm-specific technological and organisational capabilities of heterogeneous agents. Such capabilities might reflect intrinsic and fixed characteristics of individual firms (e.g. entrepreneurial or managerial talent) or they might be interpreted as the outcome of processes of accumulation of competencies.

In practice, it is very difficult to distinguish between the various possible sources of persistence in innovative activities. Likewise, it is also very hard to identify whether the observed persistence is the outcome of “unobserved heterogeneity” among agents or it reflects strictly “time-dependent” processes. At this stage of the analysis and with our data-set, we can only look if the data reveal persistence in a relatively weak and generic definition, i.e. some degree of continuity in innovative activities over time, irrespective of its origin. However, given the state of the current debate, we believe that it might be quite important to establish the extent to which innovative activities exhibit some form of persistence, even in this weaker definition.

The paper is organised as follows. Section 2 describes the data set and Section 3 presents the methodology of the analysis. In Section 4 we discuss the results and in Section 5 we present the sensitivity analysis. Section 6 concludes the paper.

Section snippets

The data

Data on patent applications were used as a proxy of innovative activities in order to investigate persistence in innovative activities. Data on patent applications come from EPO-CESPRI data bank which collects information about all the applications made to and the patents issued by the European Patent Office.

Patent applications represent an indicator of innovative activities carried out inside the firm (Acs and Audretsch, 1989, Griliches, 1990), but they are not interpreted here as a measure of

Methodology

The empirical hypothesis whether there is persistence in innovative activities might be tested using two different approaches. The first one is the standard autoregressive analysis, where the autoregressive parameter can be interpreted as a measure of persistence. Given the shortness of patent time series, standard econometric methods give biased estimates of the persistence parameter.

Overall sample

Let us begin the discussion of the results by looking at overall sample. Table 1 shows the autoregressive parameter, ρ, in the six countries for transition periods of, respectively, 1 year, 5 years, 10 years. Two main results emerge clearly. First, at the aggregate level, persistence is not high but not negligible, confirming the previous results by Cefis (1996). Second, persistence declines significantly as the transition period lengthens.

Sensitivity analysis

Some auxiliary assumptions have been formulated in developing the analysis. The sensitivity analysis explores whether the results previously obtained are robust to changes in some of these assumptions. Specifically, careful consideration is directed to (i) the potential existence of structural breaks within the 16 years of the sample, meaning that the transition probabilities are not time invariant, and (ii) the possibility that the data must be represented with second-order Markov chains.

Conclusion

The main results of this paper can be summarised as follows:

  • 1.

    Data indicate the existence of some degree of persistence in innovative activities at the firm level, although it is not very high in the aggregate and declines as time goes by. However, data show also a marked bimodality and graduality in the transitions from one state to another. That is to say, both great innovators and non-innovators have a high probability to remain in their state and in general “big jumps” are unlikely. In this

Acknowledgements

We would like to thank Ashish Arora, Giovanni Dosi, Alfonso Gambardella, Paul Geroski, Soren Johansen, Franco Malerba, the participants at the seminar held at University of Trento and at the conference: The Economics of Science and Technology: Micro-Foundations and Policy, Urbino, 1998, and an anonymous referee for comments and helpful suggestions. The financial supports of the EU TRM program, ERBFMBICT 96-0805, and of MURST, is gratefully acknowledged.

References (42)

  • Bottazzi, G., Dosi, G., Lippi, M., Pammolli, F., Riccaboni, M., 2000. Process of corporate growth in the evolution of...
  • Breschi, S., Malerba F., Orsenigo L., 2000. Schumpeterian patterns of innovation and technological regimes. The...
  • Cantwell, J., 1996. The globalisation of technology: What remains of the product cycle model? Industrial Research and...
  • Cefis E., 1996. Is there any persistence in innovative activities? Discussion Paper no. 6/1996, University of Trennto,...
  • Cefis E., 1999. Persistence in innovative activities. An empirical analysis. Ph.D. Thesis, European University...
  • Cefis, E., Espa, G., 1998. Assessing accuracy in transition probability Mmatrices. Discussion Paper no. 1/1998,...
  • W.M. Cohen et al.

    A reprise of size and R&D

    Economic Journal

    (1996)
  • W.M. Cohen et al.

    Innovation and learning: the two faces of R&D. Implications for the analysis of R&D investment

    Economic Journal

    (1989)
  • Dosi, G., 1988. Sources, procedures and microeconomic effects of innovation. The Journal of Economic Literature,...
  • G. Dosi et al.

    Learning, market selection and the evolution of market structure

    Small Business Economics

    (1995)
  • Ericson, R., Pakes, A., 1992. An Alternative Theory of the Firm and Industry Dynamics. Cowles Foundation Discussion...
  • Cited by (0)

    View full text