|dc.description.abstract||The central aim of this study is to analyze the research production and R&D activities of Colorado State University (CSU) across its different colleges, departments and other research units, and to evaluate how those activities impact the Colorado economy's agriculturally-related sectors. The study consists of three main chapters, to introduce the dynamics of university knowledge transfer to local agricultural economies. Chapter 1 explores CSU research production and technology-transfer activities, using a unique panel date set for each of 54 academic departments over the period of 1989-2012. In order to estimate the empirical knowledge-production function (KPF), this chapter attempts to build a negative binomial panel regression model with a polynomial distributed lags (PDL) of research expenditures. Three categories of research outputs are modelled, including (1) published journal articles, (2) industry collaboration, and (3) technology transfer mechanisms. In the regression results, publications are clearly the most common research outputs of the university, with a more systematic relationship between research inputs and publications than the other two types of research outputs. Moreover, it appears to exhibit decreasing returns to scale, whereas the collaborative and tech transfer research outputs appear to show increasing returns to scale. In the results of a seemingly unrelated regressions (SUR) model among the three different types of research outputs, publications and the tech transfer mediated research outputs are the primary research outputs in the university and have the maximum impact from past research expenditures. Furthermore, results indicate that collaboration mediated outputs are substitutional relative to the more formal tech transfer outputs. Chapter 2 explores the agency of knowledge production, viewing scientific research teams as "quasi-firms" arising as independent knowledge-creating entities within the university context. First, the findings from the ego-centric social network showed that the participation of outside members makes it possible to increase the size of the ego-centric teams and the growth patterns of the percent share are an obvious parallel to the patterns of team size. Particularly, the growth rate of team size is opposite of the percent share of ego's home department co-authors with upward and downward tendencies, respectively. Second, the findings from the regression results showed that the number of CSU departments per team is statistically significant in the team's assembly mechanism for both the article teams and patent teams. Thus, it seems reasonable to conclude that cross-functional team formation is more effective and common in the university research team formation and has a positive impact on the size of research teams. Finally, the quality of research teams' knowledge production tells us that the group with multiple departments per article has a higher research impact than a group using a single department per article. By the same token, larger-sized teams have higher impacts than relatively smaller-sized teams, as well as field variety. The group with multiple references per patent had a higher impact than the groups with a single or no reference per patent. This result tells us that the citation mapping from backward citations to forward citations is a significant factor for testing the research teams' impacts on the economic and social benefits with respect to knowledge spillover. Chapter 3 has focused on CSU's knowledge spillovers within agriculturally related fields and technologies. The findings indicated that academic knowledge spillovers are geographically bounded, but they are not strictly limited to the regional scale. Crucially, the impact of university spillovers on agriculturally-related industries depends upon which type of knowledge dissemination channel is utilized by university researchers. Broadly speaking this chapter evaluates four types of channel—including the publication mechanism, the industry collaboration and extension mechanism, the technology patenting/licensing mechanism, and the venture creation mechanism—each of which are variously adapted to transmitting different degrees of sticky (tacit) versus slippery (codified) knowledge. The results showed that in both aggregate level of technology and six different technological categories, the spillover impacts of journal publications, are rarely localized within Colorado; rather, the geographic scope of these impacts are national and even global. However, the extent to which the spillover impacts of patented knowledge is localized within Colorado is open to question because it is possible to control permissions for use, but at the same time it is impossible to limit everyone’s awareness and use of it, particularly in foreign jurisdictions where patents are not taken out by the university. However, the collaboration mechanism requires closer interaction and greater geographic proximity, which usually prevents global dissemination. Thus, we observe geographic proximity is significantly important for these channels. However, there are even distinctions within these. For example, we find industry coauthorship on articles to be less likely to be localized than privately sponsored grant awards. Nevertheless, the stickiness of these channels might depend also on the different technological categories. As mention as above, the geographic proximity is important only in aggregate level of technology, but it can be varied across the different technological categories, especially the slippery form of knowledge in animal health and nutrition health technology. Finally, university start-ups are highly geographically bounded near universities because in the early stages start-up companies need support from their host university.