A critical assumption made here is that the modes of the optimal and suboptimal distributions
should more or less coincide. At first, this may seem a rather bold assumption, but the opposite
assumption, namely that the two modes significantly differ, is considerably more unlikely. Such an
outcome is only possible when economic rationality in the selection of R&D projects is severely
violated. Assuming only a moderate level of selection inefficiency, the two modes should roughly fall
together.
The next step is to link the ex ante distribution with the ex post distribution. The two should
be roughly identical if the following assumptions hold: (1) the differences between expected and
actual rates of return of R&D projects are only stochastic and not systematic; and (2) the outcome
distributions of the ERRs are more or less symmetric. This latter assumption clearly does not hold
when the success of an R&D project is considered discrete - it is either a full success or a complete
failure. Although a common metaphor, it is more realistic to assume that the success of R&D
(especially biological R&D) is relative rather than discrete and, hence, creates a continuous statistical
distribution. Given the various parameters that enter an ERR calculation, each with its own
probability distribution, the overall probability distribution of the ERR of an R&D project or program
is not something that can be calculated easily. This is particularly true when the relationship between
parameters and benefits is nonlinear. In such an instance a Monte Carlo simulation can be used to
estimate measures of the central tendency (e.g., mean or mode) and dispersion (e.g., variance and
coefficient of variation) of the outcome distribution (Sprow 1967). To my knowledge, only a few
studies (Greig 1979; Anderson 1991) have actually used this approach in an agricultural R&D setting.
They reported probability distributions that are only slightly skewed.
Based on a rather strict, neoclassical interpretation of the R&D priority-setting process (i.e.,
that R&D investments are based on full information about profit opportunities and rational priority
setting), the following postulate is proposed:
(3) Assuming that the evaluated R&D projects are selected at random and their ex post rates of
return differ only stochastically and not systematically from the expected rates of return, the