This study analyzes the dynamic effects of a patent reform in a model of cumulative innovation. It finds that the incentives at different stages of the innovation process are not uniform. The first innovator captures most of producers' surplus from a policy reform. The high incentive of the first innovator may dominate the potentially negative incentives of subsequent innovations. The optimal patent policy from a dynamic social welfare perspective can be quite different from that of steady-state social welfare. Concerns about the problems of patent thickets or anti-commons may not be too serious if dynamic social welfare is considered.