Models of Research and Productivity

In this chapter we develop the structure of our models for estimating the effects of U.S. public agricultural research on U.S. agricultural productivity. We begin by laying out a general model relating research spending to agricultural productivity. Because this general specification includes too many parameters to be estimated individually with any degree of precision, some restrictions must be imposed. A conventional approach is to model productivity as a function of knowledge stocks that are created as a result of past research and can be represented using a distributed lag model defined by a small number of parameters. In this chapter we describe our approach for creating research knowledge stocks, including the specification of the research lag structure and state-to-state (and federal-to-state) spillovers used to construct the knowledge stocks. In later chapters we evaluate the effects of specification choices on our estimates and the implied benefit-cost ratios.