On additive spillovers and returns to scale in R&D

This paper explores economically meaningful forms of cost functions for process research and development in the presence of imperfect appropriability of inventive output. We propose, as a central criterion to be satisfied by the knowledge production process, that a given R&D investment should always produce more cost reduction if devoted to one lab rather than two independent labs operated under natural spillovers. With input spillovers this postulate is broadly satisfied. However, with output spillovers this is generally not the case for R&D technologies having decreasing returns to scale. We identify economically plausible restrictions on the size of spillovers and on the properties of the R&D cost function so as to bring about compatibility with the above postulate. Some empirical evidence in support of both the criterion and of its theoretical implications is discussed.