Breaking the waves: a Poisson regression approach to Schumpeterian clustering of basic innovations

The Schumpeterian theory of long waves has given rise to an intense debate on the existence of clusters of basic innovations. Silverberg and Lehnert have criticised the empirical part of this literature on several methodological accounts. In this paper, we propose the methodology of Poisson regression as a logical way of incorporating this criticism. We construct a new time series for basic innovations (based on previously used time series), and use this to test the hypothesis that basic innovations cluster in time. We define the concept of clustering in various precise ways before undertaking the statistical tests. The evidence we find supports only the 'weakest' of our clustering hypotheses, i.e., that the data display overdispersion. We thus conclude that the authors who have argued that a long wave in economic life is driven by clusters of basic innovations have stretched the statistical evidence too far.