The paper offers a novel justification for the non-obviousness patentability requirement. An innovation involves two stages: research results in a technology blueprint, which development transforms into a profitable activity. An innovator, who is either efficient or inefficient, must rely on outside finance for the development. Only patented technologies are developed. Strengthening the non-obviousness requirement alleviates adverse selection by discouraging inefficient innovators from doing research, but creates inefficiencies by excluding marginal innovations. We show that it is socially optimal to raise the non-obviousness requirement so as to exclude bad innovators; we also provide several robustness checks and discuss the policy implications.