The OECD working party on ‘Using microdata in economic research: innovation and performance’ has performed a cross-country analysis of the relation between R&D, innovation and productivity. This note discusses the general model, econometric issues and the approach taken. Some modifications are also discussed and tested to investigate the sensitivity of the results of the base model used by the OECD. Firstly, we look at the sensitivity of the results to alternative assumptions of the error structure, regarding correcting for selectivity bias as well as correlation over the error terms in each equation in the system. Secondly, we look at the outcomes for different instrument sets. Thirdly, alternative ways to calculate predictions for the endogenous variables are investigated, including as well innovators as non-innovators. Finally, we look at using a different measure of productivity as the dependent variable, namely Total (or Multi) Factor Productivity (TFP).

Results are presented based on the Dutch CIS IV data. The results for the common approach are compared with the alternatives. Based on this, suggestions for further improvements of the model are given.