Firms’ productivity growth and R&D spillovers: An analysis of alternative technological proximity measures

Abstract

This paper aim at assessing the impact of R&D spillovers on firms’ economic performance as measured by productivity growth. The construction of R&D spillovers is based on Jaffe's methodology (1988, 1996) which associates econometrics and data analysis. The main objective of the paper is to extend Jaffe's methodology by examining alternative methods for measuring R&D spillovers and to test their impacts in terms of the robustness of results. In particular, the method used to classify firms into technological clusters as well as the metrics implemented to appreciate firms’ technological proximities which enter the construction of spillovers are further investigated. In addition to R&D spillovers, firms’ own R&D capital, labour and physical capital are estimated by means of a Cobb–Douglas production function. The data set consists of a representative sample of 625 worldwide R&D intensive firms over the period 1987–1994.

Keywords:

• R&D spillovers
• Productivity growth
• Panel data
• Clustering methods

Acknowledgements

I thank Cristiano Antonelli (the Editor), two anonymous referees, Henri Capron, Lydia Greunz and Ruslan Lukatch as well as participants of the fourth INIR workshop at Université Libre de Bruxelles, August 2002, for helpful comments on earlier versions of this paper. This Paper is produced as part of a CEPR research network on ‘Product Markets, Financial Markets and the Pace of Innovation in Europe’, funded by the European Commission under the Research Training Network Programme (Contract no: HPRN-CT-2000-00061).

Notes

¹For a review of this empirical literature, see Feldman (2000) and Breschi and Lissoni (2001) inter alia. The literature on the spatial agglomeration of innovative activities also distinguishes between R&D spillovers associated with industrial specialisation (Marshall externalities) and diversity (Jacobs externalities) in a given region (see Greunz, 2004, for a discussion).

²These questions have already been examined in previous work (Capron and Cincera, 1998 and Capron and Cincera, 2001).

³See Berndt et al. (1995) for a discussion of different quality-adjusted prices indexes in the case of PCs.

⁴In fact, firms have to face relevant costs in order to build their absorptive capacity and to engage in cooperative and interactive activities (Cohen and Levinthal, 1989; Veugelers, 1997).

⁵Quoting Mohnen (1996, p. 51), ‘In a world of certainty and free disposal, R&D spillovers are expected to have beneficial effects, since it is reasonable to assume that firms do not adopt new ideas which reduce their profits. However, one can raise a number of arguments claiming that R&D can have detrimental effects on profit, productivity growth or welfare. For strategic reasons, firms may feel obliged to enter an R&D race without necessarily benefiting from it. R&D spillovers can increase or decrease the price that a producer can charge for his product, depending on whether the new product from outside R&D is substitutable or complementary to the firm's own product. New products can displace old ones. This process of creative destruction can be harmful if innovators do not have time to recover their R&D investments. Firms may have to incur heavy adjustment costs to learn the new technologies. Finally, R&D can reduce welfare when firms use R&D as a strategic tool to raise entry barriers, or when firms are obliged to duplicate R&D to stay in the race.’

⁶See Griliches (1979 1992), Mohnen (1991) or Cincera and van Pottelsberghe (2001) for reviews on this topic.

⁷Taking the example of Verspagen (1997, p. 49), ‘One may think of sectors such as rubber and plastic products which, by the chemical nature of their technology base, may benefit from technical knowledge on fertilisers, although their relationship in terms of user–producer interactions with the fertiliser industry will be marginal’.

⁸For studies based on US patents, see also Jaffe et al. (1993) and Jaffe and Trajtenberg (1996) and for European patents, Maurseth and Verspagen (2002).

⁹Mohnen (1996, p. 41).

¹⁰See Appendix A. Hence, the dimension of the technological space K is equal to 50 and each component of the technological vectors, t _ik , is defined as the share of patents applied by the ith firm in the kth technological class with respect to the total number of patents applications of that firm.

¹¹In some sectors patent protection is relatively inefficient and other protection mechanisms such as secrecy can be preferred. The effectiveness of the various protection mechanisms varies across industries and is very important for only a few of them, mainly chemicals and pharmaceuticals (Mansfield, 1986; Levin et al., 1987; Cohen et al., 2000).

¹²The fact that not all inventions are patented, nor all are patentable and that not all patented inventions are economically valuable are two other major drawbacks of patent statistics (Griliches, 1990; Jaffe and Trajtenberg, 2002).

¹³Licht and Zoz (1996) investigate the patents–R&D relationship for German firms by using patent applications in different patent offices. The results do not differ, at least for large firms, according to the patent office considered.

¹⁴As emphasised by Cohen and Levinthal (1989), firms will only be able to learn from external R&D if they themselves invest in R&D.

¹⁵Capron and Cincera (1998), extend Jaffe's methodology by distinguishing, in addition to the local and external stocks, also national stocks from international ones.

¹⁶For φ=0.2 and P _ij =0.1, for instance, the transformed proximity P _ij ^* increases to about 0.63.

¹⁷For Φ=2, for instance, all original proximities comprised between 0 and 0.25 yield a transformed proximity close to zero. In the case, only the R&D of technologically close firms (P _ij >0.25) enter the R&D spillover stock. This situation may arise when the R&D spillovers are effective only for technological proximities above a certain threshold.

¹⁸More details as regards the construction of the sample and the variables as well as descriptive statistics are available in Capron and Cincera (1998). In terms of R&D expenditures, the Japanese firms of the sample represent 38% of the corresponding national aggregate figure. For the European and US firms these percentages are 48 and 53, respectively.

¹⁹See Arellano and Bond (1991) for a discussion of the GMM first difference estimator. This estimator uses a set of lagged endogeneous variables as instruments. The validity of the set of instruments can be tested by using Sargan tests of overidentifying restrictions.

²⁰The literature on R&D spillovers suggests that bothe intra and inter-industry spillovers are important and affect positively economic growth. Yet, the magnitude of these effects varies considerably across industry sectors (See Forni and Paba, 2002, for some recent evidence on this question).

²¹The paper has since been published (Los and Verspagen, 2000).

Additional information

Notes on contributors

Michele Cincera

Email: [email protected]