Cognitive processes in preference reversals

Abstract

A preference reversal occurs when revealed preference orderings for the same alternatives differ across response modes. Previous studies have relied primarily on input-output data to infer the presence of particular cognitive mechanisms hypothesized to produce reversals. To obtain more detailed measures of such cognitive mechanisms we used a computer-based experimental methodology to conduct two process tracing experiments, of pricing-choice (Experiment 1) and pricing-rating (Experiment 2) reversals. Results show that in choice (relative to judgment), subjects take much less time, use different patterns of information search, often use strategies that directly compare alternatives (rather than consider them sequentially as in judgments), and allocate their attention differently. In pricing and rating subjects frequently use a strategy of anchoring and adjustment in which the starting point is heavily influenced by those gamble elements most compatible with the response scale. Further, we find that several of these process measures, especially starting points and adjustments, and differential attention to probabilities and payoffs, are strongly related to the frequency of reversals. The degree to which starting points are influenced by the gamble elements that are most compatible with the response scale is particularly important. Results of a third experiment further explore the role of anchors, by showing that a starting point manipulation significantly reduced the rate of pricing-rating reversals. Finally, the results of the three studies are integrated, and implications for theory and future research are discussed.