R&D, innovation and productivity, and the CIS: sampling, specification and comparability issues

We construct and estimate a modified version of the Crépon, Duguet and Mairesse (1998) model to investigate the links between R&D, innovation output and productivity. We propose a model specification that, among other features, aims to better capture the variance component of the cooperation and obstacles to innovation variables. Results based on Finnish CIS data support this alternative specification and highlight the role of process innovations. We also address comparability over time within the same country, by comparing Finnish CIS3 and CIS4 estimates. Despite CIS harmonisation, both sample selection issues and nuances in the specification of the variables prove to have a profound effect on the results, and question the validity of this currently popular comparison exercises.