This paper analyses the differences in total factor productivity between R&D investing firms and firms that do not undertake these activities. We allow for a two way relationship between R&D and productivity by applying non-parametric testing procedures and the concept of stochastic dominance. Firm data are drawn from the Spanish Encuesta sobre Estrategias Empresariales for 1991-1998. Our results show that R&D investing firms are more productive. Furthermore, the most productive firms selfselect themselves into R&D activities and by starting these activities they reach higher levels of productivity.

Key words: productivity and R&D, firm data, stochastic dominance, nonparametric analysis.
JEL classification: D21, D24, L10, C12, C14.