Too much, too little? A CBC approach accounting for screening from both sides

Consumers are often assumed to use a two-stage decision process, screening out products in the first step and choosing among the remaining alternatives in the second step. When analyzing data from discrete choice studies, a compensatory decision strategy is usually presumed. Gilbride and Allenby (2004) introduced a method to model a decision process in a choice-based conjoint analysis combining the compensatory assumption with the two-stage decision process. Respondents first screen out alternatives that do not meet minimum requirements for attributes, followed by a choice between the remaining alternatives using the compensatory rule. In this paper, we extend their approach by considering not only screening with a minimum threshold but also with a maximum value for every attribute. We compare this extension to the original method by Gilbride and Allenby (2004) and a single-step compensatory model. We do so on the basis of one simulation scenario as well as three empirical conjoint datasets. The results indicate that two-sided screening is applied especially to prices. Both the original and extended models exhibit nearly identical performance. However, they outperform the one-step choice model that ignores screening in terms of fit and predictive validity. • Investigates two-sided screening behavior in a two-stage decision process. • Not only screening with a minimum but also with a maximum threshold is considered. • Alternatives with an unacceptable feature are rejected in the first step. • Alternatives with prices that are too low are rejected in the first step.