How much should be spent in research and development (R&D)? How should R&D vary over
the business cycle? In this paper we answer both questions in the context of a calibrated
dynamic general equilibrium model with Schumpeterian endogenous growth. Firstly,
we demonstrate that, although the existence of distortions in a decentralized economy
produces underinvestment in R&D, a simple proportional subsidy to R&D spending alone
cannot restore the first best allocation. The optimal proportional R&D subsidy attains a
second best allocation in which R&D spending exceeds its first best level. Secondly, we
show how the observed procyclicality of R&D is socially inefficient. However, the welfare
loss due to this dynamic inefficiency is much smaller than the loss due to underinvestment
in R&D.