During the past 40 years, the returns to agricultural R&D have been on average in the range of 40-60% (Alston, et al 2000, Evenson 2001). Many agricultural economists see this high average as convincing evidence that there is significant underinvestment in public agricultural R&D (Ruttan 1980, Pinstrup-Andersen 2001). This paper sheds new light on the underinvestment hypothesis by introducing a simple model of the selection of R&D projects and confronting it with the rate-of-return evidence accumulated over the years worldwide. The model assumes that the distribution of all possible R&D projects on an expected rate-of-return (ERR) scale declines asymptotically. Under the neoclassical conditions of full information and profit maximization, R&D project selection starts with the project with the highest ERR and continues until the budget is finished or the last project hits the social cutoff rate, whichever comes first. Hence the underinvestment gap can be defined as the difference between the ERR of the marginal R&D project (the actual cutoff rate) and the social cutoff rate. Only three variables need to be known to estimate the underinvestment gap: the social cutoff rate, the actual cutoff rate, and the slope coefficient. Taking less than full information and economic rationality into account, the paper discusses how the latter two can be derived from a sufficiently large and representative sample of ex-post rates of return on agricultural R&D. Important findings of the model are: · Not the mean but the mode of the ex-post rate-of-return distribution is the relevant variable for assessing underinvestment in agricultural R&D. · Under the assumption of full information and profit maximization, developed countries could have invested about 40% more in public agricultural R&D and developing countries about 137% more. In terms of agricultural R&D intensity (i.e., R&D expenditures as a percentage of AgGDP), developed countries could have invested 2.8% rather than 2.0%, and developing countries 1.0% rather than 0.4% in 1981-85. · Low investment in public agricultural R&D in developing countries is caused foremost by a relatively smaller portfolio of profitable R&D projects to choose from. Underinvestment certainly plays a role (the gap is bigger for developing countries), but it explains only a small part of the difference in agricultural R&D intensity between developed and developing countries. · While efforts to reduce the underinvestment gap should continue (e.g., better priority setting and mobilization of political support), more emphasis should be placed on designing policies that help to shift (the portfolio of) R&D projects higher up on the ERR scale, even at the risk of increasing the underinvestment gap. Key words: agricultural R&D, underinvestment, rate of return, research intensities