either category. Explanations for differences in agricultural R&D intensities across countries should
be sought among the more country-specific factors.
This characterization of the ranked distribution of R&D projects provides a rough outline of
the factors that shape the portfolio of innovation opportunities. Depending on the specific industry and
country, a further detailing of these factors should provide a better understanding of the innovation
opportunities within reach, as well as what could be done to enhance them.
4. Empirical evidence
Rates of return on public agricultural R&D investments vary widely from negative to positive rates of
100% and more. On average, however, the reported rates range between 40% and 60% (Alston et al.
2000). In the literature, this high average has been used frequently as an indication of underinvestment
in agricultural R&D (Ruttan 1980, Pinstrup-Anderson 2001). The argument made here is that the
mode of the ex post rate-of-return results provides substantially more information about relative
underinvestment in agricultural R&D than the mean. Hence, it is necessary to give a new
interpretation to the reported rate-of-return results.
The most recent compilation of rates of return to agricultural R&D has been published by
IFPRI (Alston et al. 2000). The focus of the IFPRI study, which is a meta-analysis of more than 1,800
agricultural R&D rate-of-return calculations, is to understand differences in rate-of-return results due
to differences in methods, research focus, location, time, etc. For the purposes of the current study,
however, a large number of the rate-of-return observations had to be eliminated, for reasons of
comparability. For example, all nominal (rather than real) rates of return were eliminated as well as all
rates of return pertaining to “all agricultural R&D,” “crop & livestock R&D,” or “all crop R&D.”
Such R&D programs are far too aggregate to provide meaningful information about the marginal
R&D project. Even after this correction, the set of rate-of-return results is still somewhat biased
towards research programs rather than discrete research projects. The latter would be preferred in
order to get as close a correspondence with the ex ante choice situation as possible.
Most rate-of-return studies provide multiple rates for the same R&D project or program,
depending on the assumptions made, such as the time lag between R&D investment and impact. The
12