Abstract: Extensive research demonstrates that reference products can systematically influence consumer preference and choice. However, consumers' utilization of different reference products is usually not observable. In their decision-making process, consumers may use different reference products and the utilization of reference products can be quite distinct under different contexts. In this paper, four discrete choice models are proposed to incorporate three types of external reference products (i.e., the most preferred product, the least preferred product, and the middle product in the current choice set) and a potential internal reference product, respectively. A choice-based conjoint experiment is conducted to collect consumer choices for smartphones, which are used to calibrate the models. The parameters of different models are estimated with maximum likelihood estimation method (MLE). The research results indicate that the reference effect or comparison effect does exist in consumer choice and that consumers are more likely to use the most preferred product (in the current choice set) as their reference product. When the most preferred product serves as the reference product, there is significant comparison effect for non-price attributes while the comparison effect for price is not significant.

Key words: reference product, conjoint analysis, discrete choice model, comparison effect, smart phones