Abstract: We study the relative importance of within-firm and within-city agglomeration for innovation. Using novel data linking U.S. patents to detailed employment histories, we construct an inventor-level panel that tracks individuals across firms and locations. Inventor productivity rises with the number of inventors at the firm but falls with non-inventor employment. Expressing changes in citation-weighted patent output relative to its sample mean, our most saturated specification shows that 10 percent more firm inventors raises citation-weighted patent output by 2.2 percent, while 10 percent more non-inventors lowers it by 0.7 percent. We establish causality using a mover design and a shift-share instrumental variables strategy based on variation in the supply of graduates who eventually became inventors across university programs. The IV estimate implies that 10 percent more inventors raises output by 8 percent. By contrast, the positive innovation spillovers at the city level disappears when the inventors at the same firm are excluded, while within-firm spillovers remain strong. The firm, rather than the city, appears to be the primary locus of innovation spillovers.
Real Estate Department