Long-run patterns in the discovery of the adjacent possible

The notion of the "adjacent possible'' has been advanced to theorize the generation of novelty across many different research domains. This study is an attempt to examine in what way the notion can be made empirically useful for innovation studies. A theoretical framework is construed based on the notion of innovation as a search process where knowledge is recombined to discover the adjacent possible. The framework makes testable predictions about the rate of innovation, the distribution of innovations across organizations, and the rate of diversification or product portfolios. The empirical section examines how well this framework predicts long-run patterns of new product introductions in Sweden, 1908-2016, and explores the long-run evolution of the product space of Swedish organizations. The results suggest that, remarkably, the rate of innovation depends linearly on cumulative innovations, which explains advantages of incumbent firms but excludes the emergence of winner-take-all distributions. The results also suggest that the rate of development of new types of products follows Heaps' law, where the share of new product types within organizations declines over time. Finally, the study also demonstrates that the structure of the Swedish product space carries important information about adjacent possible innovations.