Pierre Therrien, MSc, Science and Innovation Sector, Industry Canada, 235 Queen St. Room 936B, Ottawa, ON K1A 0H5, Canada and Petr Hanel, PhD, Économique, Université de Sherbrooke and Centre interuniversitaire de recherche sur la science et technologie (CIRST)., Boul de l'Université, Sherbrooke, QC J1K2R1, Canada.
Research teams from nineteen OECD countries have been involved in a joint research project having for objective to implement an econometric model to examine the links between innovation and productivity. The firm data from innovation surveys carried out by national statistical agencies, were in some cases linked to additional data on firm’s activities and performances. The theoretical and econometric model, inspired by Crepon-Dugay and Mairesse (1998), was discussed and agreed by all research teams. Using the country specific data, the common ‘core’ model was estimated by a common econometric methodology. The specification of the three stage model takes care of the selection and simultaneity issues. To ensure international comparability, the ‘core’ model did not include variables for which data were missing in some countries. Its formulation left aside some potentially important relationships suggested by theoretical considerations and/or former research. In spite of these constraints, results are broadly in line with theoretical hypotheses and other studies and show a surprising degree of similarity between countries. The main recognized weaknesses of the core model is its controversial treatment of process innovations that comes out with a negative sign in the performance equation. To improve the process innovation side of the core model is one of the main objectives of the extended model presented here. Another objective is to asses the sensibility of firm performance to alternative variables not included in the core model. Instead of sales per employee, the performance indicator is labour productivity measured as value added per employee, and its change over time (2002-2004). Several binary (yes or no) explanatory variables are replaced by better measures available from the Canadian innovation survey linked to annual manufacturing census (the extent of foreign market orientation, various sources of innovation, government policies etc.). To examine more in depth the potentially different determinants of process innovations and their impact on productivity, the model is estimated for different sub-samples, e.g. depending of whether the innovative firms introduced only process innovation or introduced both product and process innovation. Different types of process innovations (improved method of producing vs. improved logistics; quality of innovation based on novelty) will also be tested. The estimated relationships are controlled for the size of firm and industry sector. They are estimated for manufacturing firms of all sizes and separately for the small and medium size firms employing less than 150 employees. To examine whether industry effects are more complex and pervasive than information captured by industry dummy variables the models are also estimated separately for ten industry groups.