Research

This paper argues that product quality improvement is a key driver of robot adoption in U.S. manufacturing. Using microdata, I find that firms’ robot use increases with a stronger strategic focus on enhancing product quality and the introduction of higher-quality goods. Establishments with greater robot intensity produce goods with higher unit prices (a proxy for quality) relative to peers within the same product category. Following robot adoption, establishments experience increases in both unit prices and revenues, consistent with an outward shift in demand toward higher quality. Motivated by this evidence, I develop a multi-sector model in which rising automation is driven by non-homothetic forces: producing higher-quality goods is more robot-intensive, and as income rises, demand shifts toward higher quality. I calibrate key industry-level parameters governing non-homotheticity using indirect inference and quantify the extent to which these forces account for the rise in automation as part of a broader structural change in a growing economy.

This paper studies the effects of artificial intelligence (AI) adoption on firm growth, with a focus on its role in product innovation and marketing. Using annual panel data on Korean firms, I document that AI is primarily adopted to support product development. Estimates from a local projection difference-in-differences (DiD) design show that firms adopting AI subsequently experience faster growth, along with increases in both product innovation and marketing activity. Although AI-adopting firms were, on average, already growing faster than their industry peers, the relatively faster growth observed after adoption persists even after conditioning on pre-adoption growth trajectories or capital investment around the time of adoption. These findings suggest that selection effects alone may not fully account for the post-adoption growth differentials observed among AI adopters.

Under the production approach to markup estimation, where markup is estimated as the ratio of a variable factor's output elasticity to its cost share of revenue, differences in production technology across firms and over time may be mistakenly attributed to higher markups if these technological differences are not accounted for in estimating output elasticity. I analyze how allowing firms that adopt advanced automation technologies to operate under different production functions affects markup estimation. Using a firm panel that tracks the adoption status of advanced technologies, I investigate how estimated markups change when accounting for technology adoption, whether firms with automated technology, on average, charge higher markups, and whether markups increase following technology adoption.