A quantile regression analysis of the role of R&D spillovers at firm level

ABSTRACT

The aim of this paper is to analyse the role played by knowledge spillovers in the firm productivity of two European countries, namely France and Italy. A quantile regression approach has been used to investigate the spillovers-productivity nexus across countries and quantiles. Moreover, in order to distinguish between the different sources of knowledge spillovers, a distinction is made between R&D spillovers from firms in the same sector and geographical region, same sector but different regions and the same geographical area but different sectors. Results suggest heterogeneity across countries and quantiles.

KEYWORDS:

• R&D
• spillovers
• European manufacturing firms
• quantile regression

Acknowledgements

The author would like to thank Francesco Aiello, Lidia Mannarino, Valeria Pupo, Fernanda Ricotta and two anonymous referees for their valuable suggestions. The author is also grateful to the participants of the XXXIX AISRE Annual Scientific Conference (Bolzano, 17–19 September 2018) for their helpful comments on an earlier version.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes on contributor

Paola Cardamone is Associate Professor of Economic Policy at University of Calabria. Her research interests include R&D, innovation, productivity, knowledge spillovers, university-industry collaboration.

Notes

1 https://www.bruegel.org/publications/datasets/efige/

2 After merging the EFIGE dataset with balance-sheet data from the Bureau van Dijk’s Amadeus dataset, to estimate TFP, Bruegel researchers adopted the LP approach using a Cobb–Douglas production function and value added as proxy of output, deflated with industry-specific price indices. Labour and capital were measured by the number of employees and the real value of tangible fixed assets, respectively. Industry-specific production functions were considered (Altomonte and Aquilante 2012; Altomonte, Aquilante, and Ottaviano 2012).

3 In order to preserve anonymity, the EFIGE database just includes randomised regional and industry identifiers (according to 11 NACE-Clio categories). This means that users know that a given firm in a given country is in an ‘industry 2’ or/and in ‘region 3’, but they do not know what ‘industry 2’ or ‘region 3’ correspond to. Also regions were combined to 5 aggregated regions for France and 11 regions for Italy.

4 Since each NACE-Clio industry code could be mapped into different Pavitt classes, here each sector is defined combining NACE-Clio and Pavitt classification in order to better proxy the technological proximity of firms in each sector. Two firms are considered in the same sector if they belong to the same NACE-Clio and Pavitt industry. In addition, in the computation of intrasectoral-intraregional spillovers, firm i is excluded.

5 Using a sample of Italian manufacturing companies over the period 1996-2005, Antonelli and Scellato (2013) employed an analogous approach to investigate the role of knowledge interactions in firm TFP. The average TFP of the other firms in the same sector and region, that of firms located in the same region but operating in other sectors and the average TFP of firms in the same sector but located in other regions are considered. Results show positive and significant effects of the three average TFP measures.

6 Tables A1 and A2 report descriptive statistics and correlation matrices, respectively. Variance inflation factors for independent variables of Equations (1) and (2) are also computed and they are lower than 3.

7 Estimates are obtained through the ‘qreg2’ (Machado, Parente, and Santos Silva 2011) command in Stata.

8 Due to the cross-sectional structure of the data, most of the explanatory variables, such as R&D, are contemporaneous with the phenomenon to be explained. While one should be cautious in interpreting estimates in terms of causal relationships, they can be seen as associations between variables.

9 A number of contributions have discussed the negative effect of spillovers (e.g. De Bondt 1997; McGahan and Silverman 2006; Kafouros and Buckley 2008). Indeed, a firm’s R&D increases not only the stock of knowledge in a society but also its own products, processes and productivity. Consequently, it might happen that the increased productivity of one firm negatively affects the economic performance of others (Kafouros and Buckley 2008). Hence, the negative effect of competition could outweigh the positive influence of spillovers. As an example, Kafouros and Buckley (2008) found that UK low-tech manufacturing firms, because of their limited ability to absorb external knowledge, are negatively affected by R&D spillovers. Moreover, a negative competition effect through intra-industry trade is found by Bitzer and Geishecker (2006).