Abstract: | We assess the usefulness of patent statistics as an indicator of innovation, using a direct measure of innovation in the hard disk industry (1976–1998). Three findings emerge: (i) patents “predict” innovations better than a random guess, and a simple refinement makes them more useful; (ii) conditional on actually innovating, conglomerates and larger firms patent more than specialised startups and smaller firms; and (iii) patent reforms seem to make the patent–innovation relationship nonstationary. These results suggest that researchers should use caution when comparing patents of different types of firms and across years. |