A unified conceptualization of the attraction effect

Abstract

This research focuses on the attraction effect, a phenomenon in which the share of the focal alternative (called “target”), relative to a second alternative (called “competitor”), increases when a third alternative (termed “decoy”) is introduced, such that the target dominates the decoy but the competitor does not. On the one hand, research has demonstrated the existence of the attraction effect in a wide variety of situations; on the other hand, more recent research has questioned the existence of the effect under some circumstances. The purpose of this research is to provide a unified conceptual framework using the reference dependence and loss aversion principles of prospect theory to examine the circumstances under which the attraction effect benefits the target, the competitor, or neither. With its unifying conceptualization of the attraction effect, the research contributes to the literature by reconciling some contradictory findings and viewpoints and offering directions for further exploration.

Appendices

Appendix 1. A Selective and thematic review of the attraction effect literature

We briefly highlight the role of four sets of factors in influencing the attraction effect: (1) the nature of the choice set and the information/data provided to the study participants in experiments, (2) the type and extent of information processing done by the participants and the associated resources involved in the choice task, (3) customer-specific factors, and (4) external contextual factors.

Nature of information about the alternatives

A central theme of the recent debate on the existence of the attraction effect is the way the information about the alternatives is made available to the study participants. Yang and Lynn (2014) and Frederick et al. (2014) postulate that the attraction effect occurs only in a stylized and numerical representation of the alternatives and not when the alternatives are presented in a qualitative/pictorial manner. However, Choplin and Hummel (2005) and Lee et al. (2008) demonstrate the attraction effect with pictorial cues (line and circle size variations in the former and pictures of people in the latter). More generally, although research has argued that the attraction effect occurs when information presented is less meaningful (i.e., when participants are not able to interpret the stimuli in the context of the choice task and are unable to compare the alternatives in the choice set along each attribute; Ratneshwar et al. 1987), Malaviya and Sivakumar (2002) show that the effect of such less meaningful information is further enhanced if study participants are asked to justify their choice. Sen (1998) explores numerical versus verbal presentation of information and shows that category knowledge attenuates the attraction effect in the former and increases it in the latter. Celedon et al. (2013) show that the inclusion of brand names (i.e., more meaningful information) reduces the attraction effect. Simonson (2014) argues that the attraction effect manifests when one of the attributes is price (a numerical attribute).

Another point of contention between the different research camps on the attraction effect is the notion of the customer specifically realizing the asymmetric domination relationship based on the stimuli. Simonson (2014) argues that if a study participant does not realize that the decoy is an asymmetrically dominated one, the evidence for the target not benefiting from the decoy (e.g., Frederick et al. 2014; Yang and Lynn 2014) is not adequate to refute the attraction effect.

Type and extent of information processing by participants

Another stylized conclusion from the literature is that the attraction effect is a result of people using heuristics in complex decision-making tasks (Simonson 1989) through peripheral rather than central processing of information. This also dovetails with the previous point regarding less meaningful information being presented: such occasions are apt for customer use of heuristics and intuition because of a lack of relevant information, which in turn increases the attraction effect (Mao and Oppewal 2012). Malaviya and Sivakumar (2002) argue that increased allocation of resources need not always result in a reduction of the attraction effect but that this effect is contingent on the nature of information available: increased resource allocation in the presence of less meaningful information increases the attraction effect, while additional resource allocation to process more meaningful information decreases the effect. Thus, evidence shows that the extent to which the customer allocates resources for processing is not unequivocal.

Customer-specific factors

The involvement level of the participant influences the extent of the attraction effect by determining the level of resources contributed to the choice task. However, this link is not monotonic. Mishra et al. (1993) find that product category knowledge diminishes the attraction effect, as does the customer’s task involvement while Malaviya and Sivakumar (1998) show that the attraction effect increases for consumers who have a moderate (vs. very high or very low) level of product category knowledge and for customers who assign more (vs. equal) importance to one product attribute over the other. In addition, customer learning and feedback reduce the attraction effect (Ahn et al. 2015). More specifically, past experience with the choice set as well as repeated choice situations can mitigate the attraction effect (Sivakumar and Cherian 1995. Chatterjee et al. (2011) show that context effects can occur for both promotion- and prevention-focused customers. Relatedly, Mourali et al. (2007) show that products associated with a promotion focus are more likely to exhibit the attraction effect.

In summary, research findings show that customer learning mitigates the attraction effect. Similarly, increased experience and customer expertise diminish the magnitude of the effect.

External contextual factors

Research has shown that asking study participants to justify their choice can increase the attraction effect (Simonson 1989); this is because choices under asymmetrically dominated alternatives are difficult and time consuming. However, Malaviya and Sivakumar (2002) show that justification increases the attraction effect only in the presence of less meaningful information about the alternatives and decreases the attraction effect when more meaningful information is presented. Malkoc et al. (2013) find that the attraction effect exists only for the attributes customers perceive as positive or desirable (vs. negative).

The nature of consequences of the decision is another important external factor that can affect the attraction effect. In particular, if the customer faces real economic consequences for the choice decision (i.e., the economic reward received depends on the choice the customer makes) versus a hypothetical choice, the attraction effect diminishes in magnitude (Lichters et al. 2016b). The issue of economic consequences is clearly related to the nature of the experiments, the amount of reward the participants receive, and how the award amount is determined.

Appendix 2. Decoy effects with more complex outcomes

Regarding the remaining cells in Table 1, the total effect of the two attributes is not straightforward to assess but some conclusions can be drawn regarding which alternative is likely to be favored. For each cell, we can combine the respective effects of the reference dynamics of the two attributes from Table 1 (i.e., the relevant proposition from P2, P3, P4, P5, and P6 for attribute 1 and the relevant propositions from P7, P8, P9, P10, and P11 for attribute 2) to advance some conclusions. For example, when the reference level for the target-superior attribute is between the competitor and the decoy, and the reference level for the competitor-superior attribute is above that of the competitor, predictions from P3 and P7 must be combined. As discussed previously, according to P3, the introduction of the decoy favors the target if the reference level is in the middle of the range or lower; according P7, the introduction of the decoy favors the competitor. Combining these two predictions, we can conclude that the introduction of the decoy clearly favors the competitor whenever the customer puts more importance weight on the competitor-superior attribute (this is because the effect predicted in P7 will be stronger than the effect predicted in P3). When the customer puts more importance weight on the target-superior attribute, the conclusion is not straightforward. The only situation in which the introduction of the decoy favors the target is when the reference level for the target-superior attribute is in the lower part of the range and when the customer puts more weight on the target-superior attribute (i.e., when the effect predicted in P3 becomes stronger than the effect predicted in P7). In other instances, the introduction of the decoy favors the competitor.

Similar logic can be used for explaining the outcome for the other remaining cells. The predictions are summarized in Table 1.

Appendix 3. The attraction effect for relatively inferior decoys

In addition to investigating asymmetrically dominated decoys, researchers have examined decoys that are not asymmetrically dominated by the target but are relatively inferior in terms of tradeoffs (e.g., Lichters et al. 2015; Malkoc et al. 2013; Ratneshwar et al. 1987; Simonson 1989). Such decoys are located below the indifference curve (the locus that provides the attribute combinations that are equally preferable). Therefore, the definition of a “relatively inferior decoy” depends on the shape of the indifference curve. For example, the region in which the decoy is considered as relatively inferior will be larger for a convex indifference curve but smaller for a concave indifference curve. Regardless, our conceptualization can still be applied to reference-level dynamics and the attraction effect due to the introduction of such inferior decoys (see Fig. 12). We briefly discuss the effect of such decoys to show the utility of our approach.

Fig. 12

Attraction effect for relatively inferior decoy.

Note that the reference dynamics specifically related to attribute 2 are all scenarios that have been covered already for the asymmetrically dominated decoy, even in the zones with the relatively inferior decoy (the new location of the decoy is within the purview of the reference-level dynamics we have already considered: there is no change in the zones of attribute 2 for the new decoy). That is, the first column in Table 1 covers all the possible reference-level dynamics related to attribute 2 for the relatively inferior decoy as well. Regarding attribute 1, while some results (see the first row in Table 1) are applicable, others are not. When the variable reference level for attribute 1 is below C or the same as D for the relatively inferior decoy, the results for the corresponding conditions (columns) from Table 1 apply. When the variable reference level for attribute 1 is between C and T, the result for the variable reference level for attribute 1 between C and D discussed for the asymmetrically dominated decoy applies (in the case of the asymmetrically dominated decoy, the reference level for attribute 1 being less than that of D implies that it is also less than that of T). When the reference level for attribute 1 is above D, the result for the condition of the variable reference level above T for the asymmetrically dominated decoy applies (because the attribute value of D is more than that of the attribute value of T). That is, predictions in five of the columns in Table 1 apply for the relatively inferior decoy as well. One condition that needs to be examined and which is unique to the relatively inferior decoy is when the reference-level for attribute 1 is between T and D.

When the reference-level for attribute 1 is between T and D (see Fig. 13), the attribute 1 for both T and C are coded as losses. The introduction of the decoy increases the reference level (reference level moves to the right). In this case, both the target and the competitor experience loss increases. Because the loss increase occurs closer to the origin for the target and farther from the origin for the competitor, the negative effect is stronger for the target than for the competitor. Therefore, the introduction of the decoy favors the competitor because of the reference-level dynamics for attribute 1.

Fig. 13

Reference dynamics when reference level for attribute 1 is between T and D (relatively inferior decoy). a Reference levels (reference level without decoy: ^________; reference level with decoy: - - - -) b Utility changes due to reference-level dynamics

We can combine this result with the scenarios from the reference level for attribute 2 to arrive at the net result of introducing a relatively inferior decoy. We can use the same logic used for assessing the effect of introducing an asymmetrically dominated decoy to combine the effects and derive propositions for the introduction of a relatively inferior decoy as well.

Table 2 provides a summary of the effects in the context of a relatively inferior decoy. We highlight some propositions as new propositions because the rest of the propositions are based on the propositions developed for the asymmetrically dominated decoy.

Table 2 Reference-level scenarios and the attraction effect (relatively inferior decoy)