An Experimental Assessment of Reference-Dependent Models of Context Effects
27 Pages Posted: 8 Aug 2019 Last revised: 2 Dec 2019
Date Written: November 2, 2019
In a choice experiment with 1866 Amazon Mechanical Turk workers in the domain of cars, we test some of the predictions of a family of multi-attribute choice models that incorporate a menu-dependent reference point whose value in every attribute is the minimum value of that attribute among the available options. In particular, these models predict that in two of the scenarios we consider in this paper, one should expect a significant attraction effect when an asymmetrically dominated option is added to the choice set, due to a change in the reference point. However, in the third scenario, these models predict no change in the choice probabilities of the existing options after adding the asymmetrically dominated option, and thus, no attraction effect, as the reference point remains intact. These models also make a number of predictions regarding the presence or the absence of another context effect called the compromise effect, which are tested in this paper. This amounts to a total of 8 predictions. We find support for 7 of these predictions, suggesting the aforementioned reference point has a significant impact on preferences and choices.
Keywords: Context Effects, Reference-Dependence, Attraction Effect, Compromise Effect, Multi-Attribute Decision Making
JEL Classification: D01, D03, D12
Suggested Citation: Suggested Citation