Modeling simplifying information processing strategies in conjoint experiments

Conjoint experiments are usually based on the assumption that respondents consider all attributes varied in the experiment when providing preference evaluations or choosing between choice alternatives. Recently, some research has examined the validity of this assumption by empirically analyzing the impact of the number of attributes on estimated utilities. It suggests that respondents not only build up a mental representation of the decision problem in reality, but also when they provide value judgments in a conjoint experiment. It implies that explicit modeling of this process of mental representation and information processing may improve the validity of conjoint estimates. This paper puts forward such a modeling approach, based on principles of bounded rationality. The approach uses attribute thresholds to construct individuals' preference structures from which heterogeneous decision heuristics can be exactly inferred. Decisions are modeled as a two-layer process with an individual selecting a heuristic first and then applying the chosen heuristic for decision making. The whole process is modeled with a latent class structure and the choice of heuristic is assumed to be influenced by mental effort, risk perception, and expected outcome. An application of the approach is carried out using data about people's choice of a new transit system. The results show the ability of the proposed approach to estimate different decision heuristics and information search patterns in different stages of the decision.