The importance of R&D and Innovation for Productivity: A Reexamination in Light of the 2000 French Innovation Survey.

Abstract

This paper consists in a detailed robustness investigation of the Crepon, Duguet and Mairesse [1998] (CDM) econometric framework, which encompasses as a reduced form equation the production function extended to R&D, as pioneered by Zvi Griliches. We focus on the sensitivity of the estimated productivity elasticities of innovation and R&D, with respect to different model specifications and estimation methods, and to the different innovation output quantitative and qualitative indicators available in the 1998-2000 French Community Innovation Survey. The main conclusion we draw is that as long as we properly correct for endogeneity and also, when appropriate, for selectivity, we find again estimates of the productivity elasticities of R&D and of the corresponding median rates of return to R&D, which are close to the usual estimates from the extended Cobb-Douglas production function. We interpret the need to instrument innovation and R&D as revealing important measurement errors in the innovation intensity variables, as well as but to a lesser extent in the innovation binary indicators, the R&D intensity variable and binary indicator. Although innovation surveys unveil interesting information and the CDM model is a good framework, which can both be used to advance our understanding of the firm R&D and innovation activities and performances, much more remains to be done to improve the quality of the data and relevance of the analysis.