Towards Data Auctions with Externalities

The design of data markets has gained importance as firms increasingly use predictions from machine learning models to streamline operations, yet need to externally acquire training data to fit such models. One aspect that has received limited consideration is the externality a firm faces when data is allocated to competing firms. Such externalities couple firms' optimal allocations, despite the inherent free replicability of data. In this paper, we demonstrate that the presence of externalities increases the optimal revenue of a monopolistic data seller by letting firms pay to prevent allocations to other competing firms. This is shown by first reducing the combinatorial problem of allocating and pricing multiple datasets to the auction of a single digital good. We achieve this by modeling utility for data solely through the increase in prediction accuracy it provides. Then, we find the welfare and revenue maximizing mechanisms, highlighting how the form of firms' private information - whether they know the externalities they exert on others or vice-versa - affects their overall structures. In all cases, the optimal allocation rule is a single threshold (one per firm), where either all data is allocated or none is.