Among various methodologies for demand forecasting of new products, the random-coefficient discrete-choice model using stated preference data is considered to be effective because it reflects heterogeneity in consumer preference and enables the design of experiments in the absence of revealedpreference data. Based on estimates drawn from consumer preference data by structural hierarchical Bayesian logit models, this study develops the overall, strategic, demand-side management for new products by combining market share simulation and a rigorous clustering methodology, the Gaussian mixture model. It then applies the process to the empirical case of electronic payment instruments.
Rethinking the Role of Stated Preference Data in Travel Demand Forecasting.
