Reconciling the Firm Size and Innovation Puzzle by

Since Schumpeter, there has been a long-standing debate regarding the optimal firm size for innovation. Empirical results have settled into a puzzle: R&D spending increasing with scale while R&D productivity decreases with scale. Thus large firms appear irrational. We propose the puzzle stems from the fact that product and patent counts undercount large firm innovation. To test that proposition we use recently available NSF BRDIS survey data of firms R&D practices as well as a broader measure of R&D productivity. Using the broader measure, we find that both R&D spending and R&D productivity increase with scale—thus resolving the puzzle. We further find that while large firms and small firms differ in the types of R&D they conduct, there is no type whose returns decrease in scale—there are merely types for which the small firm penalty is less severe.