Determination of R&D investment in French firms: a two-part hierarchical model with correlated random effects

Abstract

The aim of this paper is to identify the determinants of Research and development (R&D) investment in French firms. For analysis of covariate effects on R&D expenditures, a two-part model that combines a probit regression on the decision to perform R&D and a linear regression on log-positive expenditures is estimated. This two-part hierarchical model with a correlated random effect's structure accounts for both the skewed nature of firms’ R&D expenditures data in French regions and the fact that firms’ expenditures are correlated within the regions. In this study, we propose to identify which factors have affected French R&D investment in French manufacturing sectors in 2005. Estimations are based on a sample of 3308 French firms within 19 regions of France.

Keywords:

• R&D expenditures
• French firms
• French regions
• MCMC
• two-part hierarchical model

JEL Classification :

• O3
• D21
• R10
• C11
• C15
• C34

Acknowledgements

The authors are grateful to Serge Vendemini for providing database and to Georges Bresson for his helpful comments. We would also like to thank two anonymous reviewers for their comments and constructive suggestions.

Notes

France is lagging behind the USA, Japan and Germany.

The 22 administrative French regions minus Corsica, Languedoc-Roussillon and Limousin.

For an extended survey, see Cohen (1995).

Here, the two dimensions of the panel concern regions and firms.

NUTS2 French regions correspond to administrative regions. Regions considered are enumerated in Appendix 2.

All variable concerning firms are extracted from Amadeus.

International if the firm is active only at the international level and 0 if not.

For a review of concentration measures and their properties Hart (1975).

This dummy equals 1 if the firm is active in a high-tech industry.

R&D subsidies are public regional assistance provided for industries aiming to support the technological development of small and medium firms and accelerating the transfer of the public research towards the industry (which is about innovation of product, process or service).

Industry subsidies are public regional assistance given for industries that are mainly intended to finance the operations of creation, maintenance, modernization, adaptation or transmission of local companies in the commercial sector, craft industry and services, in order to preserve or develop a number of regional firms.

Data on patents are extracted from REGIO (Eurostat).

Data concerning patents are extracted from REGIO, which is a Eurostat database.

Random effects are on regions and not on firms.

The quantity reported in the MC error column gives the MC standard error of the mean, which is an estimate of (σ is the standard deviation and N is the number of simulations).

Those products concern essentially French know-how goods, existing patented innovations, …

Corresponding regions are reported in Appendix 2.