Attraction and Compromise Effects in Choice-Based Conjoint Analysis: No-Choice Options as a Remedy: An Abstract

Abstract

The composition of the choice set strongly influences consumer choice in a way that is inconsistent with the concept of stable preferences (Prelec et al. 1997). During the last decades, researchers have identified a variety of corresponding phenomena, commonly referred to as context effects (Lichters et al. 2015). While context effects in consumer decision-making have several striking implications for marketing research and practice, they also cast the validity of choice-based conjoint (CBC) analysis into doubt (Drolet et al. 2000). However, CBC notably differs in certain aspects from consumers’ actual purchase decisions (Meißner et al. 2016). Specifically, CBC sets out to measure consumers’ preferences, relying on hypothetical and repeated choices, often presented in a forced-choice format. These aspects have already been shown to moderate context effects’ emergence (e.g., Ahn et al. 2015; Dhar and Simonson 2003; Lichters et al. 2016).

This research addresses the question whether the compromise effect (CE) and the attraction effect (AE) as the most common representatives of context effects truly occur in CBC studies. In a series of three online experiments, we systematically vary the choice set composition of holdout tasks in order to provoke the CE and the AE in both between- and within-subjects designs. By doing so, our research is the first that includes a direct measurement of consumers’ inclination to engage in context effects in the course of a CBC study. As another between-subjects factor, we experimentally manipulate the CBC choice format. Our results reveal significant CEs and AEs in forced-choice CBC, whereas both effects do not occur in free-choice CBC, which allows subjects to opt for a “no choice” instead of a product option.

Our results offer confidence for practitioners that typically rely on free-choice CBC, while, likewise, they reaffirm the importance of behavioral estimation models in forced-choice CBC that seek to incorporate context effects (e.g., Chorus, 2014; Rooderkerk et al. 2011). Furthermore, our results are relevant for researchers that seek to develop more efficient choice designs, as advancements allow inhibiting the AE (Sawtooth Software 2017), while other design strategies foster the CE (Huber and Zwerina 1996). Finally, our results point out potential difficulties within recent methodological advancements, which strongly rely on forced-choice tasks (e.g., Johnson and Orme 2007; Schlereth and Skiera 2017).

References Available Upon Request