The remainder of the paper is organized as follows: Section 2 reviews previous works and places our paper within that literature. Section 3 develops the econometric methodology, departing from the previous published work on agricultural knowledge that is reviewed in Section 2. Section 4 describes the data and variable creation. Section 4 reports the empirical results, and Section 5 presents the conclusion and policy implications.The knowledge production function has various applications at societal and sectoral levels. A recent published theoretical framework addressing the role of knowledge in society’s growth was developed by Dolgonosov . Distinguishing between technological knowledge and general total knowledge, the author demonstrated that knowledge is essential to allow sustainable population growth within the carrying capacity of the planet. The role of knowledge production is essential, especially with the increasing population and environmental load. This framework suggests that society could introduce policies to improve the efficiency of knowledge production in various sectors. The literature distinguishes also between knowledge of various qualities. Cammarano et al. introduced the notion of quality of innovation output, using patent data from bio-pharmaceutical and equipment-producing companies. The analysis suggests a more productive knowledge process in which innovative firms use knowledge and information produced by external sources. Working on a related industry, Lauto and Valentin estimated a knowledge production function for what was coined the new science development model for clinical medicine,hydroponic dutch buckets in which research can be conducted in a transnational effort, or locally.
This is a very interesting distinction that may indicate the efficiency of transnational simultaneous research benefitting from a variety of conditions and its superiority to knowledge spillover of research conducted separately. However, the authors find that by its nature, transnational research may have lower efficiency and impact because it includes diverse aspects in quantitative comparisons. Some surprising findings are offered by Roper and Hewitt-Dundas , who estimates the interaction between knowledge stocks and flows and their impact on the firm’s innovation. They found that negative rather than positive effects between knowledge stocks and innovation , and knowledge flows dominate the effects of knowledge stocks on the innovation of the firm. Several works address the issue of networking and proximity among the knowledge creation centers , and the effects of collaboration within and between regions on knowledge productivity . Both works were applied to Europe. Ramani et al. develop a model of knowledge production function that can be estimated at both the firm and the sector level and apply it to the bio-food industry. The production function in this work allows to distinguish between the absorptive capacity to exploit inter- and intrasectoral spillovers. Marrocu et al. found that technological proximity outperforms the geographic proximity, suggesting that networking has a limited role in enhancing knowledge creation. The most relevant finding of De Noni et al. to our work is that the impact on knowledge productivity is stronger in the case of collaboration between regions with diversified knowledge base. From a different perspective, Verspagen and De Loo addressed the spillover effect of knowledge, both across sectors and over time using a knowledge flow matrix. The methodology is very relevant for knowledge production investments, but it is heavily dependent on data that might not be readily available everywhere. Two examples of recent studies that address spillover effects in knowledge production are Wang et al. and Neves and Sequeira . Wang et al. estimated the spillover effects in the semiconductor industry to find that the strength of the networking ties between companies explain the level of spillover effect in the knowledge production process. Spillover effects are expected to be stronger in weaker network ties. Neves and Sequeira conducted a meta-analysis of data from 15 published works to find expected, but reassuring results.
They quantify level of spillover effects and discover that the spillover effect will be larger when they include in the estimation of the knowledge production foreign inputs, and it will be lower when only rich economics are included in the estimation. Finally, universities are considered a hub for knowledge production, based on research conducted in addition to their role as educational institutions. Gurmu et al. used patents issued to universities during 1985–1999 as a measure of knowledge. They explained variation in knowledge by field of knowledge, R&D expenditures , as well as detailed human capital variables, and several control variables. Their results indicated marginal contribution of each research variable to the production of knowledge. While the literature review is by no means inclusive, it represents the many efforts that have been made in the literature for understanding the determinants of knowledge production. We will rely on these works while developing our analytical framework.The literature suggests that agriculture-related R&D inputs result in the production of knowledge, which upon application leads to improvement in productivity in the agricultural sector. Alston et al. , Birkhaeuser et al. , and Evenson estimated the impact of R&D and extension-related expenditures on agricultural productivity. The underlying theory is that expenditures made towards R&D and outreach impact productivity, and that impact of research expenditures is differential; old expenditures have a lower impact on current productivity. Evenson and Birkhaeuser et al. reported positive impacts of both R&D and cooperative extensions on productivity for studies from around the world. While these studies provide strong evidence of a long-term impact of R&D-related expenditure as well as the impact of farmer-extension agent contacts on productivity, there is a gap in our understanding of how well these proxies for agricultural knowledge represent actual knowledge produced. This is understandable because measurement of knowledge produced from investments in R&D is conceptually and computationally complicated. Griliches discussed the issues of measurement of knowledge production between public and private sector investments in R&D. He claimed that patents are a good approximation of knowledge and innovation, especially because of the commercial value attached to it. An industry or a firm likes to file for patents to have sole right on its invention and is paid for its use by others. Pavitt mentioned that patents are good proxy measures of innovative activities. Other studies have used patents as proxies for knowledge production.
Data on patents are well documented in the United States and in the rest of the world and are easily obtainable without the hassle of conversion of units. In the industrial sector, knowledge produced through research is mostly owned as private property by the innovating firm because of the related commercial incentive of private property ownership. This makes patents the most appropriate proxy variable for knowledge production function analysis in the case of private sector research. However, publicly funded research and especially agricultural research creates knowledge, most of which is publicly available. Pardey and Dinar used publications as a proxy for knowledge production. Publications are more prevalent in public research agencies, where research results are typically published in journals. Dinar used peer-reviewed journal publications in different fields as the dependent variable for his study of the agricultural research system in Israel. According to Pardey , publications have been chosen over patented and non-patented output like mechanical innovation processes or new biological material, books, State Agricultural Experiment Station bulletins, and newsletters. Publications capture the knowledge output of a station completely because they establish intellectual property rights of the researchers over their work,bato bucket which in turn affect their salary scale, promotion rate, and tenure status. Link analyzed the determinants of inter-farm differences on the composition of R&D spending, namely basic and applied R&D. He regressed these R&D components on profits, diversification, ownership structure, and subsidies. Jaffe found a significant positive impact of university research on corporate patents for a number of technical areas, such as drugs and medical technology, and electronics, optics and nuclear technology in the United States. The literature on the topic leads us to two main observations: a dearth of papers that deal with the analysis of the knowledge production function and the study of the impact of production inputs on knowledge produced; and the choice of variables representing knowledge produced through investments in R&D only provides a partial picture of the true process. There is little attempt to compute a comprehensive knowledge production variable that captures knowledge produced through all avenues. UCCE follows an input-output framework for research, which involves utilization of research inputs such as manpower and infrastructure, for the production of knowledge to be disseminated to potential clientele from a variety of different sectors. This knowledge is produced through basic and applied research, and extension work, which are targeted to address the needs of the clients at the county level. Agricultural knowledge that is generated by UCCE is public in nature and is freely available to all.
Because of this, it seems appropriate to use various types of peer-reviewed publications by advisors as the representative variable for knowledge. But publications are only a part of the total knowledge produced; there are other modes by which knowledge is produced and disseminated by UCCE. These need to be incorporated into the analysis to capture a more complete representation of the generated knowledge. To achieve this, we collected data on eleven different modes by which UCCE produces knowledge, all of which are aggregated to the county level to create a knowledge index that captures all UCCE knowledge produced.ichotomous variables representing county fixed effects are introduced in the model to control for factors that are common to a county, and possibly impact productivity. Year fixed effects can control for random shocks, e.g., budget surplus leading to a recruitment of more skilled advisors in a particular year, which may have led to larger number of total knowledge produced across all counties in a single year.This is included to capture possible diminishing marginal returns to infrastructure. Expenditure on infrastructure can be beneficial to knowledge production, but after a certain degree of provision the marginal effect may diminish. It makes little sense to keep building laboratories and offices if there are no researchers or staff to fill them. We follow Roper and Hewitt-Dundas , who introduced the plant size as a quadratic Schumpeterian resource indicator, which has also been shown by Jordan and O’Leary to have an inverted-U shaped relationship with knowledge production. A similar specification by Charlot et al. lumps all R&D costs in a quadratic relationship due to economies and diseconomies of scale. The quadratic specification of infrastructure expenses means that over-investment in research infrastructure may turn to be counter-productive and to result in diminishing marginal productivity of knowledge production. To test this hypothesis, the square term for log of infrastructure expenses was included in our model. The choice of the log-log model for the empirical analysis is to facilitate the computation of output elasticity for each of the inputs of production.The University of California Cooperative Extension was established a century ago with the purpose of educating the citizens about agriculture, home economics, mechanical arts and other practical professions.2 Through the course of almost a century since the Smith Lever Act of 1914, the UC Cooperative Extension has grown into an elaborate system that has branched out from handling mainly farm related issues to many other aspects concerning the farm, as well as the overall society. Extension advisors communicate practical research-based knowledge to agricultural producers, small business owners, youth, and consumers, who then adopt and adapt it to improve productivity and income. Today the UCCE works in six major areas,including Agriculture, 4-H Youth Development, Natural Resources, Leadership Development, Family and Consumer Sciences, and Community and Economic Development. This paper focuses on UCCE activities in agriculture. The University of California Division of Agriculture and Natural Resources headquartered in Oakland, California, is the source of data for the analysis in this paper. We collected annual budget data from the database for all UCCE county offices for the period of 2007 to 2013.Our data set includes complete data for seven years for 47 county offices, which serve the 58 counties in California. There are six groups of two counties each, which are served by a single county office. And there is one office that serves four counties. Upon comparing older UCCE budget data with real expenditures, we found that they follow similar time trends for each county office and could be used as proxies for expenditures. This data was converted into constant 2013 US dollars, using GDP deflator data from the World Bank database and is presented as such hereafter.Henceforth, we will refer to the UCCE budget as expenditures, to avoid ambiguity.