Data sharing or algorithm sharing?

Data combination and analytics can generate valuable insights for firms and society as a whole. Multiple firms can do so by means of new technologies that bring the algorithm to the data (“algorithm sharing”) or, more conventionally, by sharing the data (“data sharing”). Algorithm-sharing technologies are gaining traction because of their advantages in terms of privacy, security, and environmental impact. We present a model that allows us to study the economic incentives generated by these technologies for both firms and a platform facilitating data combination. We find that, first, the platform chooses data sharing unless algorithm sharing’s analytics are sufficiently superior to those associated to data sharing. Second, we identify the properties of the analytics benefit function that ensure that algorithm sharing results in a higher total data contribution. Third, we highlight scenarios in which, in presence of data externalities, there can be a mismatch between the choice of the platform and the preference of a social planner