1. Introduction
Starting with a study on hybrid corn by Griliches (1958), rate-of-return studies have become standard
practice in documenting the economic impact of agricultural R&D. Although the estimated ex post
rates vary quite substantially, the average tends to range in the order of 40-60% (Alston et al. 2000,
Evenson 2001). Though many have questioned the accuracy of these rates and expressed doubt about
as to how representative the selected projects are, a widely shared belief is that the estimated rates are
robust enough to accommodate such criticisms and still be in a range that is substantially above the
social cutoff rate. Based on this evidence, Ruttan (1980) argued that there is serious underinvestment
in public agricultural R&D. This argument has become a widely accepted opinion (if not fact) among
agricultural economists. Therefore, any slowdown in the growth or, even worse, any contraction of
public agricultural R&D expenditures is reason for serious concern.
But what do we actually know about this underinvestment? How real is it? Is it higher in
developing countries than in developed countries? Is it higher for some types of research than others?
And, how much more should have been invested in agricultural R&D? In order to answer these
questions, the underinvestment argument needs to be defined more clearly. This paper attempts to do
so by introducing a simple model that represents the ideal, economic version of the selection of R&D
projects. The outcome in real life is, of course, different because the assumptions of full information
and strict rational maximizing behavior only hold partially. Nevertheless, this paper argues that in
retrospect it should still be possible to detect the economic forces that determine the selection.
The structure of this paper is as follows. Section 2 starts with a brief introduction of the model
and discusses a new way of interpreting a representative sample of ex post rates of return on
agricultural R&D. Section 3 focuses on the factors that determine the characteristics of the portfolio
of R&D projects to choose from. The available statistical evidence and its interpretation are presented
in section 4, while in section 5 the characteristics derived from the rate-of-return sample are linked
with reported agricultural R&D investment levels. Having determined the order of magnitude of the
investment gap, section 6 assesses the various possible explanations of why the investment gap exists.
Section 7 summarizes the main conclusions.