Vol.2 No.1 2009

30/88

Research paper : Modeling the social acceptance of industrial technologies (M. Matsumoto et al.)−27−Synthesiology - English edition Vol.2 No.1 (2009) consumers select the most economically rational technologies and products. There are finer models of consumer preference and decision-making, and conjoint analysis is often used for fine modeling[7][8]. Since this approach allows fine analysis of effects on consumer preference brought on by changes in price or product performances, it enables analysis of effect of the factor changes on diffusion. However, since this model basically has no temporal dimension, it is unable to analyze temporal transition of diffusion, particularly for long-term.3.3 Learning curve modelThe learning curve model is used to analyze the cost reduction of industrial products[9][10]. New products tend to decrease in cost with mass production. Learning curve model describes this trend. Figure 2 shows the transition of production volume and price of solar cell. From past data, there is empirical rule “when the cumulative production doubles, the production cost and time required for production decreases by certain percentage.” The observed percentage of reduction is 15~30 % in semiconductor industry and 5~20 % in machine assemblies[10].3.4 Characteristics of existing modelsThe characteristics of the above models are summarized in Table 1.4 Model formulation 4.1 Process of model determination Three existing models were described in the previous chapter. Long period of trial and error was necessary in figuring out how to utilize (or not utilize) and to integrate the models. The research goal was clarified at this point, and the model was created based on whether it was persuasive or not. There were three types of persuasiveness selected for this research: (1) persuasiveness of result, (2) persuasiveness of logic, and (3) persuasiveness of analogy. (1) is the persuasiveness gained from the match between the result of modeling and the reality, and it is naturally the most convincing item. However, in many cases, the actual result of a projection is not available (for example, we do not know the “diffusion 20 years from now”), and this standard cannot be applied. However, it can be used as counter-evidence when the model fails to explain the state of diffusion, or the non-adequacy of the model. (2) is persuasion through the adequacies of assumption and logic of the model, and (3) is persuasion through referencing the similar cases in the real world.Based on these points, Fig. 3 shows the process of determining the model for this research. Initially, we thought understanding consumer preference was primary concern, and tried to build the diffusion model based on consumer preference model. However, we were unable to draw the diffusion curve that matched reality (failure of (1)). We attempted to create something that resembled reality through numerous revisions, but were unable to obtain sufficient level of persuasion in the logic of the revisions (failure of (2)). At this point we reconsidered the model. The key of reconsideration was whether the subject of diffusion analysis was long-term (several decades) or short-term (few years). We realized that consumer preference model was effective in short-term while Bass model was good for long-term. This point was not indicated in existing literature. Confirming that our study was for long-term, we set Bass model as the foundation of our model. In the Bass model, it was possible to reference the diffusion coefficients of similar products (section 3.1). For example, it could be seen that 40 to 50 years would be required for diffusion of hybrid cars in reference to past automobile products (such as automatic transmission cars), and similar number of years would be necessary for diffusion of energy-saving appliances in reference to similar appliances. This is (3) persuasiveness of analogy.Next, to enable sensitivity analysis that was difficult to accomplish using the Bass model, we attempted to incorporate the consumer preference model. The integration method will be described in the following section. We determined that it was most convincing among considered integrated models (superior in (2)). However, we do believe that there is room for more discussion for this integrated model.4.2 Formulation of modelIn this research, the original equation (1) for the Bass model was modified, and the model was formulated as follows: Xt=( p+r・nt)・(1−nt)・ N・− (4)Ht/H0 is multiplied to equation (1). Ht and H0 are values Production (kW/year)System price (1000 yen/kW)Production and price of PV400,000200,000250,000300,000350,000100,000150,000050,00019791986198519841983198219811980199019891988198719991998199719961995199419931992199120002003200220011600060008000100001200014000200040000ProductionSystem priceModel for transition of cost reduction of industrial products.Learning curve modelMicro-model for looking at diffusion by consumer preference.Appropriate for short-term diffusion transition analysis and sensitivity analysis.Long-term diffusion analysis is difficult.Consumer preference modelMacro-model where diffusion transition is seen as a whole.Appropriate for long-term diffusion transition analysis.Sensitivity analysis is difficult.Bass modelModelCharacteristicFig. 2 Production and price of photo voltaic (PV) in Japan.Table 1 Characteristics of current models.HtH0

※このページを正しく表示するにはFlashPlayer9以上が必要です