Entry into Innovation Areas

Abstract

This paper investigates the multi-area entry decisions of publicly listed firms in the United States electronics industry from 1990 to 2019. I apply Latent Dirichlet Allocation to over 630,000 patent titles and abstracts to endogenously define twenty innovation areas and merge these with market-implied patent valuations. The data reveal three empirical patterns. First, innovation occurs across multiple domains simultaneously. Second, the number of active competitors varies across areas. Third, the industry focus shifted from physical hardware to connectivity technologies. To explain these patterns, I estimate a static simultaneous-move game of firm entry using a partial identification approach with moment inequalities, which accommodates multiple equilibria and the combinatorial choice set. The 99 percent confidence sets support four conclusions. First, firms enter areas where expected patent valuations are higher. Second, holding those valuations fixed, additional rivals reduce expected profit. Third, firms also benefit from occupying more areas at once. Fourth, a large per-area entry cost, together with rival congestion, still limits how many areas a firm enters. Starting from observed portfolios, between 13 and 16 percent of firm-years already have at least one extra area that looks worth occupying at the estimated costs. A 25 percent reduction in the per-area cost raises that share to between 22 and 27 percent, and a 50 percent reduction raises it to between 33 and 41 percent. A 50 percent reduction in how costly rivals are raises that share to between 15 and 21 percent. In a case study of targeted cost reductions, entry rises in semiconductor areas such as “Semiconductor Memory Architecture” and “Semiconductor Lithography and Fabrication Processes.” The offset is lower entry in other fields, including “Wireless Base-Station and Mobile Communications.”