Predicting Japanese dairy consumption behavior using qualitative survey data

Three qualitative attribute models were applied to data from a 1997 Japanese consumer survey conducted for milk, domestic cheese, and imported cheese consumption. The three models estimated were neural networks, Quantification Theory Type II and probit models. The neural network model had the best predictive performance of three models. Important factors affecting changes in consumption of these three products were identified. Major positive factors affecting changes in fluid milk consumption included health and taste preferences by consumers. Interestingly, both health and taste concerns were also major negative factors impacting decreases in milk consumption as well. The most important positive factors affecting changes in domestic cheese consumption were price, taste, and health reasons, while health and taste concerns were also important negative factors. The single most important factor impacting changes in imported cheese consumption was taste, which means that the Japanese have acquired a new taste for imported cheese. © 1999 John Wiley & Sons, Inc.