| There are differences in diffusion profiles across vintages or generations of optical scanners that have been neglected by a previous study by Levin, Levin and Meisel (1987). The authors examine the diffusion of this innovation, assuming that it is a single innovation. In contrast, I focus on the diffusion of different vintages of optical scanners for the time period June 1974 through March 1985, taking into account the timing decisions of potential adopters and, consequently, find that the nature of diffusion of the vintages is quite different in the initial phase relative to the later phase. Thus, I account for changes in the optical scanner technology and incorporate them in the estimations.; A contribution of this research is to examine the diffusion of vintages using the full range of estimation techniques within the survival/duration framework. In the estimations, I incorporate linkage variables to control for subsequent adoption or update decisions of grocery companies, and to examine whether adoptions of earlier vintages by the company or its rivals could have influenced their adoption decisions. I find that the nature of diffusion in the earlier phase of the technology exhibits the following characteristics: the incentive to adopt this innovation increases with the number of competitors but decreases with prior adoptions by rivals. This result supports a product differentiation effect. In addition, I find that companies having large average square feet per check out, operating in concentrated markets, with high populations, and high levels of real effective buyer incomes diffuse the innovation faster. The nature of diffusion in the later phase is different compared to the earlier phase. I find that the number of competitors is not statistically significant, and that markets that are less concentrated, and have relatively lower levels of real effective buyer incomes diffuse the innovation faster. Besides, the incentive to adopt this innovation increases with increases in the number of companies that adopted the first vintage, but decreases with increases in the number of companies that adopted the second vintage. All the results discussed above are obtained using the semi-parametric and parametric approaches and remain unchanged and are consistent with a priori expectations. Finally, I examine the sensitivity of the results to different assumptions of perfect foresight and find that the results remain unaffected. |
