THE introduction of high-yielding technologies during the 1960s and the socioeconomic impact of these innovations in developing countries have been subjects of considerable interest in both theoretical and empirical research. However, while many papers provide detailed descriptions of the experiences of different regions and countries and propose various heuristic arguments to explain observed behavior (which differs among regions), rigorous analysis is needed to determine in precise terms the conditions under which these arguments are valid and to specify meaningful relationships for econometric estimation. For example, in an extensive survey of the developing-country, technology-adoption literature, Schutjer and Van Der Veen (1977, p. 28) conclude that, with respect to" size of holding as an economic constraint, there appears to be no consistent pattern of land size acting as a constraint to technology adoption" and" quantitative information regarding the effects of economic factors on interfarm variation in the profitability of new techniques and inputs is lacking." This paper develops a model that explains land-use allocation and technology adoption taking into account the interfarm variation of land holdings and the role of land holdings (wealth) in determining risk preferences. The model specifically recognizes differential uncertainties in both traditional and modern technological processes and the marginal effects of modern inputs on risk as well as fixed costs attached to adoption. As a by-product, the results also suggest a structural econometric specification that can serve as a basis for empirical examination of the relationship of farm size and adoption.

Most formal studies of adoption thus far have not produced a framework that lends itself easily to econometric estimation nor which considers the stochastic relationship of production under old and new technologies. 1 Because of the intractability of handling related random variables with the Sandmo (1971) approach, most rigorous analyses thus far have abstractly assumed that only the new technology produces stochastic yields [eg, Feder