This article claims to do three things. First, develop a formula for measuring size of research benefits which is generally applicable to all types of supply shift. Second, use this formula to assess the possible error involved in previous studies employing alternative equations. Third, initiate discussion on variables which might influence the type of supply shift. The article concludes that uncritical application of previously developed formulas without regard to the type of supply shift can result in substantial bias in estimates of research benefits. The implication is that calculation of rates of return on agricultural investment may also be severely biased.

Some hypotheses about the timing of farmers becoming aware of an innovation and the subsequent decision to use that innovation are derived from a recentlydeveloped, decision-theoretic model of the adoption process. They are tested using empirical evidence on the time taken by early adopters of trace element fertilisers in S.A. to discover and decide lo use this innovation. The central role of information search in the adoption process is emphasised and it is postulated that various distance measures provide a useful measure of information availability and reliability. The results of the empirical analysis are consistent with the hypothesised relationships. Another finding is the importance of distinguishing between early adopters who are genuinely innovative, and those potential later adopters who adopt early because they happen, by chance, to operate a farm in close proximity to another early adopter. IntroductionThe subject of this paper is the factors affecting inter-firm variability in the timing of the discovery, and subsequent decision to first use, an 'off-the-shelf agricultural process innovation (the adoption decision). In particular, attention is paid to the effect of distance to information source on the timing of these decisions. Jones (1967) and Rogers and Shoemaker (1971) have reviewed the considerable number of empirical studies that have been conducted in an attempt to 'explain' individual variability in the timing of the adoption decision. Most have been able to account for only 50 per cent or less of the variance in the dependent variable, and they have often reached conflicting conclusions about the importance and/or direction of the effect of various explanatory variables. In part, these poor results can be attributed to omitted variables, especially those capturing locational considerations. They also reflect the lack of a satisfactory conceptual framework on which to base empirical analysis of the adoption decision. For instance, notwithstanding frequent suggestions in the literature that the adoption process comprises several stages, Rogers and Shoemaker (1971, p. 350) cite many hundreds of attempts to account empirically for variation in timing of individual adoption decisions, but only six which attempt to relate it to measured duration of one or more of the above stages. A simpler approach to defining adoption stages is likely to help remedy this situation.A theoretical model of the adoption process, in which a central role is assigned to information search as the prime determinant of duration of * Presently at the University of Minnesota.

No abstract

The central proposition in our paper on supply shifts and research benefits was that the size of GARB is sensitive to the nature of the shift in the supply curve induced by adoption of a process innovation. We also argued that because the available formulas for calculating GARB typically presumed a particular type of shift, estimation error would result from the application of such formulas if the actual shift did not correspond to that presumed. Despite the claim by Rose that "U made a fundamental error which invalidated the bulk of their findings, " none of the arguments made in either comment have persuaded us to reject either of the above propositions. All of the comment by Wise and Fell (WF) and the first part of that by Rose relate to the calculations we carried out to numerically illustrate these propositions, and to provide some idea of their quantitative importance. The second part of the comment by Rose is a restatement of arguments contained in a paper by Mishan (1968), and in our view is a general attack on the notion of producer surplus rather than a specific criticism of our paper.While we cannot accept all of the detailed criticisms made in the comments, we do accept the main thrust of the argument by both authors that the computational procedures underlying the results in our illustrative tables were in error; and that as a result we overstated both the sensitivity of GARB to the nature of the supply shift and the magnitude of the possible error which would result from using a formula based on a particular type of shift. We also have some reservations to be outlined below about the computational procedures ad vocated by WF and by Rose, and hence about the tables of results they derive using these procedures, but would agree that either approach is to be preferred to our earlier proposals.Nevertheless, we note that the illustrative tables calculated by WF and by Rose still support our two main contentions, namely, that the size of GARB is sensitive to the type of supply shift, and that formulas which presume a particular type of shift can lead to substantial errors. According to Fell and Wise, "the exact method still shows benefit estimates varying by a factor of three, depending on the shift R. K. Lindner is a visiting research fellow, University of Sussex, and F. G. Jarrett is a professor of economics, University of Adelaide.Peter Wagstaff helped clarify the authors' thoughts on a number of points, but any errors are the authors. postulated." Similarly, it can be seen from table 1 of Rose that a formula which presumed a pivotal shift when the actual shift was convergent would underestimate GARB by a factor of two or more. Furthermore, there is no reason why the particular cases of a pivotal and convergent shift which we chose originally for illustrative convenience should be treated as limiting cases. It is by no means inconceivable that particular innovations might only affect marginal output, and have no effect at all on the cost of producing the bulk of intramarginal output. In such a case, th...