Data-intensive Innovation and the State:Evidence from AI Firms in China

Abstract:
Data-intensive technologies, like AI, are increasingly widespread. We argue that states may play an important role in shaping data-intensive innovation because: (i) states are key collectors of data and (ii) data is sharable across uses within firms, potentially generating economies of scope. We investigate a prototypical setting: facial recognition AI in China. Collecting comprehensive data on AI firms and government procurement contracts, we find economies of scope from government data: firms awarded contracts richer in government data produce both more government and commercial software. To study the aggregate implications of government data provision, we build a directed technical change model. We show that government data provision can increase growth and bias the direction of innovation, but welfare increases only when economies of scope are sufficiently strong and citizens’ and states’ preferences are sufficiently aligned. We conclude with three applications analyzing states’ choices of: industrial policy, surveillance levels, and privacy regulation.
Contact Emails:
scoco@ceibs.edu