The length of patents is a key policy tool for innovation and long-run growth. This paper empirically exploits an anticipated policy change that generated quasi-experimental variation in US patent length across technical fields. A difference-in-difference analysis identifies two empirical facts. First, the news of a future patent term extension causes a drop of R&D and innovation until policy implementation. Second, the drop continues even after the implementation of the new longer term. The latter effect is driven by an externality, whereby the fall of the pace of development of existing projects at the news and before the implementation slows down the diffusion of knowledge through patents, leading to less research ideas in the same technical field by the time the implementation comes. The paper proposes a semi-endogenous growth model where research and development are distinct. The former generates new abstract ideas, and development transforms the stock of ideas into patented products. A faster pace of development increases the productivity of research through an externality. The model can replicate the documented empirical patterns, and a structural estimation implies a long-run elasticity of innovation to permanent patent term changes of +0.35. Normatively, the model finds an optimal patent length of 28 years in the absence of policy anticipation. However, in the presence of anticipation, strong technological spillovers have relevant consequences for the impact of the policy on welfare and output.
Download the paper
