The Value of Data; Simone Galperti (University of California San Diego)

Abstract

Personal data is an essential input of many modern industries. Yet, its value is hard to establish and formal markets for data are still lacking. Consider a platform that mediates trade between a seller and a population of buyers using individual data records of their personal characteristics. After formulating this as an information-design problem, we use linear-programming duality to characterize the unit value that the platform derives from each buyer's specific record. We find that this value differs from the payoff that the platform directly earns from the trade between the seller and a buyer, which would be a biased measure of her record's value. This bias reflects unaccounted externalities between records, which arise because the platform pools records to withhold information from the seller. We then characterize the platform's willingness to pay for more records---e.g., more buyers joining the platform---and for better records---e.g., more information about existing buyers. Our analysis establishes essential properties of the "demand side" of data markets. Our methods apply generally to a large class of principal-agents problems.