The conventional adoption pathway for a new technology may be depicted by the logistic frequency distribution and its corresponding logistic curve shown in Figures 1a and 1b respectively (Davies, 1979; Sahal, 1981; Stoneman, 1983; Mahajan et al., 1990). If N is the fixed population of potential adopters of a new technology, then the number of new adopters in period t may be expressed as
where parameter measures the speed of diffusion. For constant , the absolute increase in adopters at any point in time, , depends on the product of the proportion that has already adopted, nt/N, and the number of remaining potential adopters, N-n. Equation 1 may be solved for the frequency distribution of adoption over time as:
where is the constant of integration, that positions the distribution curve on the time axis. Equation 2 is the cumulative density function of the logistic frequency distribution and for constant , it gives a bell-shaped frequency distribution for numbers adopting over time (Figure 1a). Equation 2 also gives sigmoid (S-shaped) logistic curve (Figure 1b), which is symmetric around the inflection point occurring at time -(/) corresponding to 50% adoption, and approaches zero and N asymptotically, as t tends to minus and plus infinity respectively. However, any unimodal frequency distribution will have a sigmoid cumulative density function but may or may not be symmetric depending on, for example, whether the population is homogenous or heterogeneous, and how quickly the new technology is modified or become obsolete and replaced by newer technology (Sharif and Kabir, 1976; Mahajan et al., 1990; Davies, 1979; Sahal, 1981; Chatterjee and Eliashberg, 1989).
In the model described above, at a point in time a population is divided into two groups, adopters and potential adopters. Rogers (1983) identified five stages in a typical technology adoption-decision process and categorized adopters, according to time of adoption, as innovators, early adopters, early majority, late majority and laggards (Figure 1a). Innovators are described as respectable local opinion leaders; the early majorities are deliberate and willing followers, while late adopters often needed peer pressure or influence to adopt. The laggards are skeptical about the new, so cling to the past and adopt at the tail end.
Models of this nature implicitly assume that the entire population eventually adopts the innovation and that, once adopted, the innovation is never rejected (Thirtle and Ruttan, 1987). In some models a population is divided into adopters, rejecters, disapprovers, and the remainder who are as yet uncommitted (Sharif and Kabir, 1976). However, the implicit assumption here is that once rejected or disapproved, the technology is never adopted again. In reality, neither `never rejected' nor `for ever rejected' is a realistic assumption for most agricultural technology adoption process, particularly at the early stage of adoption.
Most agricultural innovations evolve as they diffuse. An innovation may be changed or modified by a user in the process of its adoption and diffusion. Therefore, potential adopters may play an important role in the process of technology generation by being involved in the generation process rather than being merely passive recipients of an innovation once it has been generated (Rogers, 1983). Incorporation of farmers as participants and their perceptions and preferences as important elements in the technology generation process are considered essential for generation of appropriate technology (Ashby et al., 1989; Asfaw Negassa et al., 1991).
When farmers are not involved in the technology generation process, awareness and knowledge about a new technology precedes any adoption decision. Several authors have emphasized the importance of information gathering and updating information through learning-by-doing in the adoption process. There may be a lag between the time when farmers first hear about an innovation and the time they adopt it (Kislev and Shchori-Bachrach, 1973; Lindner et al., 1979; Stoneman, 1981; Rogers, 1983; Bhattacharya et al., 1986; Oren and Schwartz, 1988; Tsur et. al., 1990; Feder and Umali, 1993; Fisher et al., 1996). However, empirical verification of the linkage between learning and adoption and what factors influence such linkage is rare. Saha et al. (1994) have developed and tested a model in which producers' knowledge about a new technology (Phase I) determine the decision to adopt (Phase II) which in turn determine the intensity of adoption (Phase III).
Learning about and adoption of a technology may actually involve more complex processes (Figure 2). Any adoption decision is preceded by a period of awareness and learning. Initially only limited amount of information may be available or only a limited amount of available information may be digested. The information includes knowledge about how the technology functions and where and how to get access to it. The optimal level of information is reached when information acquired over a period of time reaches a threshold level at which a decision on adoption can be made. Following Saha et al. (1994), a producer's optimal information level may be considered as the outcome of an underlying utility maximization problem:
i* i (S) (3)
where i* denotes the optimum level of information and S is a vector of related producer characteristics. A producer is considered to know about the new technology if
i* (S) > io (4)
where io is the threshold level of information at which a decision about adoption can be made.
On the basis of knowledge at a point in time, a perception or belief about the technology is developed and a decision to adopt or reject or defer decision may be taken. The subsequent decisions may follow two pathways (Figure 2). In the first pathway, a decision to adopt is followed by a decision about the intensity or extent of adoption (in practice, these two decisions may be initially taken simultaneously). New knowledge and experience is gathered from learning-by-doing as well as observing other adopters, and a decision is made to increase intensity and/or modify the technology,1 or to discontinue the use of the technology. After acquiring more knowledge, a decision to re-adopt or defer adoption is taken and the process continues until a more stable decision is taken.
In the second pathway, the initial perception or belief is modified on the basis of new knowledge and/or observed performance of adopters, and a new decision about adoption is taken. A decision to adopt takes the farmer along pathway 1 (Figure 2). A decision to reject or defer decision will keep the farmer within the second pathway whereby a new decision is taken after acquiring more knowledge.
Thus, the "innovation assessment lag", defined as the time required between initial awareness and actual use of a technology, may vary depending on the farmer's access to knowledge, ability to decode that knowledge and formulate decision (Lindner et al., 1979; Fisher et al., 1996). The lag is very short for innovators and very long for laggards.
The possibilities of permanent discontinuation or temporary discontinuation and re-adoption imply that a distinction need to be made between "the number of new adopters" (Equation 1) and "the number of net new adopters" in period t; the latter being defined as
where nnt = nt - not + nrt is net new adopter in period t, nt is the number of new adopters in period t, not is the number dropped out in period t and nrt is the number re-adopted in period t. It is obvious that the frequency distribution of net new adopters, nnt, over time is likely to give a bell-shaped curve only if not = nrt. If not > nrt, i.e. number of drop-outs is greater than the number of re-adopters, the density function may not be bell-shaped but the shape of the logistic curve may be bell-shaped rather than S-shaped, i.e. as t tends to infinity, nnt tends to zero.
Equations 1 and 5 have completely different implications about the time frame and volume of potential impact of a new technology. They also have important practical implication for farmers and extension agencies. Compared to equation 1, the situation under equation 5 implies a much longer period will elapse before a majority of the potential adopters will adopt and use the technology in a sustained manner. It is therefore necessary to understand the possible pathways for adoption of a new technology and the associated factors, and take corrective measures, e.g. take more positive steps for diffusion of information for increasing awareness, remove supply constraints, to facilitate rapid adoption.
The adoption pathway described above is tested with vertisol technology in Ethiopia.
1 Technical progress consists of infrequent major innovations coupled with a steady accretions of innumerable minor improvements and modifications done by users, particularly innovators and early adopters (Rosenberg, 1982).