Modeling Innovation Ecosystem Dynamics through Interacting Reinforced Bernoulli Processes

Understanding how capabilities evolve into core capabilities-and how core capabilities may ossify into rigidities-is central to innovation strategy (Leonard-Barton 1992, Teece 2009). A major challenge in formalizing this process lies in the interactive nature of innovation: successes in one domain often reshape others, endogenizing specialization and complicating isolated modeling. This is especially true in ecosystems where firm capabilities and innovation outcomes hinge on managing interdependencies and complementarities (Jacobides, Cennamo and Gawer 2018, 2024). To address this, we propose a novel formal model based on interacting reinforced Bernoulli processes. This framework captures how patent successes propagate across technological categories and how these categories co-evolve. The model is able to jointly account for several stylized facts in the empirical innovation literature, including sublinear success growth (successprobability decay), convergence of success shares across fields, and diminishing cross-category correlations over time. Empirical validation using GLOBAL PATSTAT (1980-2018) supports the theoretical predictions. We estimate the structural parameters of the interaction matrix and we also propose a statistical procedure to make inference on the intensity of cross-category interactions under the mean-field assumption. By endogenizing technological specialization, our model provides a strategic tool for policymakers and managers, supporting decision-making in complex, co-evolving innovation ecosystems-where targeted interventions can produce systemic effects, influencing competitive trajectories and shaping long-term patterns of specialization.