Application a Multi-Generation Diffusion Model to Technology Products－A Case of DRAM

• Main Authors: Chia-Hsun Chang, 張嘉訓
• Other Authors: Ming-Yeu Wang, Ph. D.
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/80100405620322463643

碩士 === 真理大學 === 管理科學研究所 === 90 === Newer technologies are continually replacing older ones and newer ones are more efficiently. DRAM (Dynamic Random Access Memory) is an obvious technological substitution instance in semiconductor industries. 4K-capacity DRAM was introduced in early 1970’s and DRAM capacity increased four times for average three years. Nowadays, the DRAM has reached 512M-capacity and replaced ten generations during 30 years.Research on the modeling of the diffusion of innovations has results in a body of literature since the publication of the Bass model in 1969. The diffusion model of adoption and substitution for multi-generation products didn’t be introduced until Norton and Bass introduced in 1987. So far, there are only a few researches on the multi-generation diffusion model not only in domestic area but also in overseas (Norton and Bass(1992)、Speece and MacLachlan(1992，1995)、Mahajan and Muller (1996)、Islam and Meade (1997)). Norton and Bass proposed multi-generation diffusion model in 1987 but didn’t consider price variable. Besides, the model proposed by Speece and MacLachlan (1992) included price and market growth rate variables. Therefore, we attempt to investigate the adoption and substitution effects of seven-generation DRAM products by applying the multi-generation model developed by Norton and Bass (1987).The data of global DRAM shipment employed in this study included seven generations: 4K, 16K, 64K, 256K, 1M, 4M, 16M. We collect 25 observation and price data from 1974 to 1998.The model parameters were estimated using the nonlinear three-stage least squares procedure SYSNLIN, available as part of the Statistical Analysis System (SAS). Furthermore, in order to examine the model fitness and forecasting properties, we will use the procedure SIMNLIN to simulation and forecast based on the parameters we estimated.From this article, we found that, First, when applying a multi-generation diffusion model on more than two-generation diffusion products, the results of model fitness and prediction are quite well even if applying to a rapidly change products such as DRAM. Second, the model proposed by Norton and Bass (1987) didn’t include price variable. In our article, we consider variables of price and market growth rate refer to the research of Speece and MacLachlan(1992，1995) and we find that the fitness and prediction of modified model are greater than Norton and Bass model.