The growth in agricultural R&D investments around the world has slowed considerably in the past 20-30 years, even though most experts agree that there is substantial underinvestment in agricultural R&D. Introducing a simple economic model of the ex ante selection of R&D projects allows us to make a more insightful interpretation of the available ex post rate-of-return evidence. Two sets of factors can be identified that determine the level of investment in agricultural R&D: (1) the ex ante choice set of R&D projects for a given domain and time and (2) the extent to which the assumptions of full information and selection rationality apply. There are important differences in innovation opportunities between industries, between countries, and through time. Apart from purely technological opportunities, factors that also play a role are the size and structure of the market, the rate and speed of technology adoption, risk and uncertainty, and R&D effectiveness and efficiency. If improved, each of these factors could either increase R&D benefits or reduce R&D costs, creating a larger choice set of profitable R&D projects. By assuming less than full information and selection rationality, various explanations for underinvestment in agricultural R&D can be tested. The model also allows us to gain insight into the political economy aspects of R&D project selection by examining situations in which two distinctive R&D choice sets are competing with each other for funding, or in which projects are selected according to the different perspectives of social planners, farmers, or consumers.