Incorporating reference products into modeling consumer choice decision: A mixtures-of-experts model

Abstract Understanding the process of consumer decision making is important for many decision support systems. Consumers evaluate different alternatives and then come to a decision. Prior research suggests that consumer evaluations leading to choice are comparative in nature and can be affected by other alternatives or reference products. This study proposes a mixtures-of-experts model framework to examine the role of different reference products in consumer choice of multi-attribute products. While multiple external and internal reference points have been proposed, previous studies have very rarely investigated more than one reference point in the same model. Using data from a choice-based conjoint experiment, our empirical model enables us to identify which product consumers tend to use as the reference product by incorporating four different reference products and includes consumer characteristics to examine how consumers differ in their utilization of different reference products. The results show that our model outperforms other reference-dependent models in prior literature. In our empirical context of smartphone choices, the most commonly used reference product is the most preferred product in the choice set, while the least preferred product and the average product are rarely used. We also examine the role of consumer characteristics such as gender, product familiarity, and product interest in utilizing reference products. This paper provides insights into the unobserved comparison process in consumer choice, which can be applied to decision support systems such as recommendation engines.