Swarming of innovations, fractal patterns, and the historical time series of US patents

While most of us who study intellectual and technical advancement believe that innovations tend to swarm, the details of this process are not well understood. The aggregate-level behavior of US patents is examined as a way to better infer the process that generates innovation. The amount of swarming decreases as the observational period increases, which indicates that the process of innovation is not perfectly self-similar. Instead, the effects of innovations are mostly contained within specialized areas, and do not often trigger further advances in other fields.