Brewing the future : stylized facts about innovation and their confrontation with a percolation model

Innovations are known to arrive more highly clustered than if they were purely random and independent. Their distribution of importance (as measured in returns or citation rates) is highly skewed and appears to obey a power law or lognormal distribution. Technological change has been seen by many scholars as following ‘technological trajectories’ in some space of technological characteristics and being subject to ‘paradigm’ shifts from time to time. Innovations appear to arrive in clusters. Thus the innovation process is clearly more highly structured than a simple random process, but is still characterized by high unpredictability and risk. We first summarize some of these empirical observations, drawing on wellknown as well as innovative statistical measures. We then briefly review a ‘percolation’ model of the innovation process (Silverberg 2002, Silverberg and Verspagen 2002) and analyze its statistical properties on simulated data with respect to these measures. The model is able to generate similar patterns of clustering in both ‘space’ and time, highly skewed distribution ranging between a pure-Pareto in the tails to a lognormal, and structured technological trajectories.