Revisiting the decoy effect : replication and extension of Ariely and Wallsten (1995) and Connolly, Reb, and Kausel (2013)

Abstract

The decoy effect refers to the phenomenon whereby an inferior, unpreferable option (referred to as the decoy) reverses people’s preferences and increases the choice share of a targeted option. Because of its potential practical applicability, the effect has attracted extensive attention from researchers and practitioners since it was first demonstrated by Huber et al. (1982). Nonetheless, in recent years, the replicability and robustness of the effect have been questioned (Frederick et al., 2014; Yang & Lynn, 2014). Given the recent concerns on the overall replicability of findings in behavioral and psychological sciences (e.g., Ioannidis, 2005) and calls for more direct replication studies (e.g., Zwaan et al., 2018), the first aim of the research reported in this thesis was to directly replicate two impactful studies from the decoy effect literature: Experiment 1 from Ariely and Wallsten (1995) and Experiment 1 from Connolly et al. (2013). Given the large number of successful demonstrations of the decoy effect in the past, it was hypothesized that the effect would be replicated. In addition, Connolly et al. (2013) found evidence that regret salience could reduce the decoy effect. Therefore, the second aim of the current research was to replicate this reduction effect. Finally, in an extension to the replications, the current research also aimed to test a novel hypothesis that low decision reversibility (i.e., relative difficulty to modify or undo the outcome of a decision) leads to the reduction of the decoy effect. No evidence for decoy effect was found in the replication of Experiment 1 from Ariely and Wallsten (1995) (Study 1 in this thesis). The replication effects were not in the expected direction, and their sizes were trivial. I found evidence for the decoy effect in the replication of Experiment 1 from Connolly et al. (2013) (Study 2 in this thesis). Nonetheless, the replication effect was much smaller than that observed in the original study. I found insufficient evidence in support of the idea that regret salience reduces the decoy effect as well as the idea that low decision reversibility does the same. The failure to replicate the decoy effect in Study 1 may be attributed to the uncommon and convoluted design of the study that made it difficult for participants to notice the relationship between the decoys and their targets. While the decoy effect was successfully replicated in Study 2, the reduced effect size suggests that the original study, and possibly many others in the literature, are not powered enough, which might be a reason behind the insufficient evidence obtained for the hypotheses regarding regret salience and decision reversibility. Future studies on the decoy effect should continue exploring its boundary conditions (e.g., whether the effect is specific to certain study designs and what these designs are like), use larger samples to ensure sufficient statistical power for detecting the effect, and utilize stronger and validated manipulations when investigating its moderators.