Innovation and productivity: a firm-level analysis for French Manufacturing and Services using CIS3 and CIS4 data (1998-2000 and 2002-2004)

Abstract

We investigate the effect of innovation on labour productivity in France, using a general framework that accounts for research activities and for both product and process innovation. To control for selection and endogeneity effects, we estimate a nonlinear simultaneous-equations econometric model which comprises five equations. The first two account for selection into R&D activities and for the determinants of R&D intensity. The third and fourth equations allow us to estimate the “knowledge production function”, distinguishing between product and process innovation. The final equation identifies the impact of each type of innovation on labour productivity. This model is estimated by Maximum Likelihood on the third and fourth waves of the French component of the Community Innovation Survey. We find consistent results across both periods and for both manufacturing and services: product innovation appears to be the main driver of labour productivity, whereas the influence of process innovation is either not significant or close to zero. Sensitivity analyses suggest that the estimates of the effect of product innovation are fairly robust, but that those of process innovation are far less stable. We conclude with some cautionary words relative to the joint modelling of the impacts of product and process innovation on productivity.