Is there persistence in innovative activities?

This paper examines firm innovative persistence using patent applications of 577 UK manufacturing firms. Non-parametric techniques show the empirical distributions of patents are neither Geometric nor Poisson. There exists a threshold effect represented by the first patent: the probability to go from zero to one patent is uniformly much lower than from n to n+1 patents, with n<1. Transition Probability Matrices show little persistence in general, but strong persistence among 'great' innovators that account for a large proportion of patents requested: innovative activities, at least which are captured by patents, are persistent. There is heterogeneity across industrial and size classification.