Patents and R&D: Searching for a Lag Structure

Abstract

This paper extends earlier work on the R&D to patents relationship (Pakes-Griliches 1980, and Hausman, Hall, and Griliches, 1984) to a larger but shorter panel of firms. Using both non-linear least squares and Poisson type models to treat the problem of discreteness in the dependent variable the paper tries to discern the lag structure of this relationship in greater detail. Since the available time series are short, two different approaches are pursued in trying to solve the lag truncation problem: In the first the influence of the unseen past is assumed to decline geometrically; in the second,the unobserved past series are assumed to have followed a low order autoregression. Neither approach yields strong evidence of a long lag. The available sample, though numerically large,turns out not to be particularly informative on this question. It does reconfirm, however, a significant effect of R&D on patenting (with most of it occurring in the first year or two) and the presence of rather wide and semi-permanent differences among firms in their patenting policies.