International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
Left, right or random? Disentangling position bias from status quo effects
Kennet Christian Uggeldahl, Danny Campbell, Søren Bøye Olsen

Last modified: 28 March 2017

Abstract


The way in which individuals perceive different goods has repeatedly been shown to be affected by the physical ordering of goods in a given choice situation (see Campbell and Erdem (2015) for a recent review). Horizontally, products positioned in the center of the shelf are often perceived as more popular (Valenzuela and Raghubir, 2009), and persons in a central position perceived as being more important (Raghubir and Valenzuela, 2006). Within stated preference and Discrete Choice Experiments (DCE) a left-right bias has been reported by a few studies (Scarpa et al. 2011, Matthews et al. 2016), where alternatives positioned to the left are chosen more frequently compared to alternatives positioned to the right, irrespectively of the attribute levels. This has been explained by primacy effects, where items examined first are subjected to deeper cognitive processing, and establish a baseline against which the other alternatives are compared (Krosnick, 1999). Thus, in cultures where individuals read from left to right, this would entail a bias towards the leftmost alternative. Vertically, where metaphors such as “on top of things” and “hitting rock bottom” are widely used in daily life, a position bias where products and items located at the top are perceived as better compared to products place lower, has been well documented in the marketing and psychology literature (Meier et al. 2007, Chandon et al. 2009). The top-bottom bias has also been found to affect choices in a best-worst scaling study by Campbell and Erdem (2015), where trust in different institutions not only depended on the institution itself, but also on its position in the choice task.

With the growing evidence of both vertical and horizontal position biases from neighboring strands of literature, it is somewhat surprising that position bias has received relatively limited attention in the DCE literature. In an unlabeled DCE, horizontal position bias is usually accounted for by specifying alternative specific constants for n-1 alternatives. However, for the results from DCEs to be consistent with welfare economics, one of the alternatives in the choice set has to be the no-buy or status quo (SQ) alternative. If the position of this alternative is not varied throughout the choice sets, the resulting alternative specific constants will confound potential status quo effects with the position bias. Randomizing the location of the SQ alternative allows for the independent estimation of a SQ effect as well as any position bias, however, this is not normally done. In a recent study where the position of the alternatives is randomized, Matthews et al. (2016) find a left to right bias but no SQ effect. However, as their experimental setup does not include a treatment where the position of the alternatives is not randomized, it is not possible to properly test the convergent validity, and the question regarding the actual impact of the randomization is thus left unanswered. To the authors’ knowledge, such a comparison does not appear in the DCE literature, despite the obvious implications for the design of stated preference experiments.

Apart from affecting the size of the SQ effect, we hypothesize that randomizing the location of the SQ alternative will encourage respondents to make more informed choices, as the opportunity for using simplifying decision heuristics assisted by the position of the alternatives is removed. A respondent using a heuristic, such as always choosing the SQ or cheapest alternative, would have to actively find this alternative in every choice set, rather than using position as a cue, and in the process have to attend to information in other alternatives. On the other hand, it might also make the choice task more complex and thus confuse respondents, increasing the noise in the data.

This paper focuses on the effects caused by varying the position of the SQ alternative in a DCE. We do this with three treatments: one where the position of the SQ alternative is to the left in a three alternative choice set, one where the SQ alternative is to the right, and one where the position of the SQ is randomized. The data is based on a DCE concerning preferences for different riparian buffer strip (i.e., the interface between land and a river or stream) policies in Denmark, and consists of roughly 260 respondents in each of the three treatments. We present and compare estimates based on the three treatment samples, as well as investigate the effect that the position of the SQ alternative has by looking at the differences in scale and consideration sets between treatments. To do so, we jointly estimate an independent availability logit with scale adjusted latent classes and random parameters. The results indicate that randomizing the position of the SQ alternative has the effect of ‘nudging’ respondents away from using simplifying decision heuristics, such as only choosing the SQ alternative or any of the two experimentally created alternatives. Furthermore, we find evidence that randomizing the position of the SQ alternative reduces the SQ effect, which leads us to believe that some of what has previously been labelled as a status quo effect may in fact be a result of position bias.

References:

Campbell, D., & Erdem, S. (2015). “Position bias in best-worst scaling surveys: a case study on trust in institutions.” American Journal of Agricultural Economics, aau112.

Chandon, P., Hutchinson, J. W., Bradlow, E. T., & Young, S. H. (2009). “Does in-store marketing work? Effects of the number and position of shelf facings on brand attention and evaluation at the point of purchase.” Journal of Marketing, 73(6), 1-17.

Krosnick, J. A. (1999). “Survey research.” Annual review of psychology, 50(1), 537-567.

Matthews, Y., Scarpa, R., & Marsh, D. (2016). “Using virtual environments to improve the realism of choice experiments: A case study about coastal erosion management.” Journal of Environmental Economics and Management

Meier, B. P., Hauser, D. J., Robinson, M. D., Friesen, C. K., & Schjeldahl, K. (2007). “What's" up" with God? Vertical space as a representation of the divine.” Journal of personality and social psychology, 93(5), 699.

Raghubir, P., & Valenzuela, A. (2006). “Center-of-inattention: Position biases in decision-making.” Organizational Behavior and Human Decision Processes, 99(1), 66-80.

Scarpa, R., Notaro, S., Louviere, J., & Raffaelli, R. (2011). "Exploring scale effects of best/worst rank ordered choice data to estimate benefits of tourism in alpine grazing commons.” American Journal of Agricultural Economics, aaq174.

Valenzuela, A., & Raghubir, P. (2009). “Position-based beliefs: The center-stage effect.” Journal of Consumer Psychology, 19(2), 185-196.


Conference registration is required in order to view papers.