Research Subsidy Spillovers, Two Ways

Abstract

We study how the outputs of research spill over technological and geographic space in the context of the U.S. Small Business Innovation Research program. We infer input-output links using text analyses and identify the marginal costs of producing patents using noncompetitive grant matching policies. Due to technological spillovers, the costs of spurring specific types of patents are substantially larger than the costs of spurring any patent. Due to geographic spillovers, roughly 80% of the patents generated by the program are from inventors who never participate in the program. The cost effectiveness of research subsidies crucially depends on which outputs count.