| This dissertation analyzes the patent-R&D relationship, an empirical counterpart to the knowledge production function, using count panel data. The data is an original panel of 410 firms investing in R&D from 1984 to 1993 and applying for patents during that period. Econometric models for count panel data, including Poisson specifications with fixed effects and random effects and Negative Binomial models with fixed and random effects, are estimated, taking into account both the discrete nature of the count dependent variable and firm specific unobserved heterogeneity.; The number of patent applications submitted by firms are explained by current and lagged R&D spending levels in addition to lagged patent counts, current and lagged intra-sector R&D spillovers and patent count spillovers. Firm size, time and an R&D-time interaction term are considered as additional determinants. Competitive R&D spillovers, diffusion patent count spillovers, a positive effect of past patenting on current patenting and a declining effectiveness of R&D over time are suggested by the analysis of the full sample.; Analysis of a subset of the data reveals that university spillovers play a crucial role in the innovation process promoting patents with a magnitude on the order of that of current R&D. Moreover, university spillovers appear to be especially vital in the chemical and professional and scientific instruments sectors.; The impact of market structure characteristics on patenting is also analyzed for another subset of the data. The results provide only very weak evidence in support of the Schumpeterian Hypothesis based on the tiny positive effects of industry concentration and firm size.; The distinguishing feature of this dissertation is that it employs count panel data models to take a more comprehensive look at spillover externalities in the patent-R&D relationship. It expands the analysis to include intra-sector R&D and patent count spillovers as well as university spillovers and geographically mediated university spillovers. The findings show that university and intra-sector patent count spillovers can be classified as diffusion spillovers, positively affecting patent applications, while intra-sector R&D spillovers reflect the heightening competition among firms for patents in the US. |
