The role of R&D spillovers in product and process innovation

Abstract

The goal of this article is to assess the effectiveness of innovation activities and technological diffusion in Italy. In particular, we provide an econometric evaluation of the impact of technological efforts on the innovative output. Using a panel of 1203 firms over the period 1998 to 2003, we estimate a bivariate probit, which models the probability of introducing product and process innovations. Results show that the probability of implementing a product or a process innovation is positively correlated with innovative activities and negatively correlated with technological spillovers.

Acknowledgements

The author would like to thank Francesco Aiello for his precious comments as well as for his suggestions with regards the drawing up of the article. Financial support received by University of Calabria (Progetto Giovani Ricercatori) is also gratefully acknowledged.

Notes

¹ Other studies analysing the probability of introducing product and process innovation are provided by Criscuolo et al. (2004) and by Inkmann and Pohlmeier (1995). Criscuolo et al. (2004), using a sample of 8172 UK firms over the period 1998 to 2000, find a marked positive correlation between researchers employed and the likelihood of introducing innovation. Inkmann and Pohlmeier (1995), using a sample of 1325 West German firms over the period 1984 to 1988, estimate the impact of product and process innovation of similar firms (in terms of firm size, firm demand expectations and industry affiliation) on the probability of carrying out product and process innovation. Results show that the probability of introducing process innovation decreases when the share of product-innovating firms and the average of technological distance of other firms in the same industry increase, while it is positively affected by the share of product-innovating firms multiplied by the technological distance. Similarly, the probability of introducing product innovation is negatively correlated with the share of process-innovating firms and the average of technological distance of other firms in the same industry, and positively correlated with the share of product-innovating firms multiplied by the technological distance.

² According to Jaffe (1986) and Cincera (2005), the Euclidean measure is sensitive to the length of the vector. The length depends on the level of concentration of the firm's research activities among the technological classes. With this measure, the more two firms are diversified, the lesser the length of their technological vectors. As a result, these firms will be located in the central region of the technological space. Hence, they will be close to each other even if their technological vectors are orthogonal (Cincera, 2005, p.12).

³ The DEA was first proposed by Charnes et al. (1978). It consists of a nonparametric approach used to estimate a production function in order to determine the minimum amounts of inputs necessary to produce a given amount of output. This is different from stochastic frontiers in that this method does not require the specification of a functional form of the production process. In this analysis, the DEA approach is implemented considering a problem oriented to the minimization of inputs and assuming variable returns to scale.

⁴ A quadratic measure of distance is preferred to a linear one, in order to take into account the fact that technological flows due to geographical proximity decrease more than proportionately when distance between firms increases. Indeed, it is reasonable to assume that beyond a certain distance, technological flows are only marginally affected by geographical proximity.

⁵ Istat – 8th General Industry and Service Census (http://dwcis.istat.it/cis/index.htm).

⁶ Imposing a rate of depreciation equal to 0,15 is a consolidated hypothesis in empirical analyses which use technological capital (Hall and Mairesse, 1995; Harhoff, 1998; Del Monte and Papagni, 2003; Parisi et al., 2006). In some of these studies (Hall e Mairesse 1995; Harhoff, 1998) a higher depreciation rate is also considered, equal to 25%, but results are not substantially different from those obtained imposing a depreciation rate equal to 15%.

⁷ Weights are given by , where F _it are the sales of the ith firm at time t (t = 1998, … , 2003) belonging to a group sized N (i=1, … , N).

⁸ In order to avoid some information being dropped, ICT and technological capital variables are incremented by 1 so that when they are zero we obtain ln(1) = 0. In this way, we maintain the relevant information without losing observations.

⁹ Dummies relative to the North–West, to the North–East, and to the Centre are included. The control group concerns firms located in the South of Italy.