Our objective in this paper is to explore in detail the second of the three essential functions of universities as elements of a system of innovation (Schartinger et al. 2002, Smith, 1995): the generation of a type of knowledge that is directly applicable to industrial production processes; we will use patents as indicator for this process. Several research studies have demonstrated that the analysis of patents is a sufficiently valid and objective method for determining technology transfer (Archibugi, 1992, Basberg, 1987, Boitani and Ciciotti, 1990, and Trajtenberg, 1990). For Meyer- Krahmer and Schmoch (1998). patents show the interest in commercial exploitation of a new technology (...) A high share of patents on the part of scientific institutions can be considered a good indicator for a close relationship of science and industrial laboratories in the technology field”. Also the accessibility of patents allows a more comprehensive treatment than surveys or case studies (Henderson et al., 1998). Unlike the previous studies on university patents that have utilised the university (Henderson et al. 1998; Coupe 2003) or university teachers (Agrawal and Henderson, 2002) as the unit of analysis in the output of patents, in this study we utilise the research group, as an intermediate unit between the university and the academic inventor. This is a realistic option considering that the majority of researchers are integrated into groups following common lines of research, which are also the basic unit for regional financing. With a sample of 1155 Research Groups from Andalucía (Spain), the questions to investigate are the following: What is the influence of the human resources (doctors, graduates and support personnel) on the generation of university patents? What influence does the scientific capacity of the research group have on the generation of university patents? What influence does external financing -public and private- have on the output of patents? To test for the factors that influence the output of patents by public research institutions, we follow the empirical tradition of the formulation of a "knowledge production function" begun with the work of Griliches (1979): Pati=f(Hi, If, Fi, Oi), Where Pat is the production of patents by the research group i, H is the human resources of the group, S is its scientific capacity, F is the financial resources and O represents other characteristics of the group. The empirical specification of this function, given the nature of patent data, suggests the formulation and calculation of two types of model: the first, utilising models of binary choice, will enable us to determine what characteristics are possessed by groups that obtain patents; the second, by means of a count model, will pick up the determinants of the production of patents by each research group (poisson/negative binomial o zero inflated posson/negative binomial models). Our working proposal is as follows. First we carry out a review of the literature on the production of technological knowledge in the universities, as measured by patents, and then we put forward the initial working hypotheses, the methodology and the data utilised. Next we present the results of testing our hypotheses. Finally we draw the principal conclusions and policy implications.
