On Statistical Testing for Intransitivity in Multiattribute Consumer Preference Surveys

In recent years various structures have been proposed for estimating intensity of consumer preference in consumer surveys. Empirical studies have shown intensity measures to be reliable and cost effective by providing more information per question. These structures use consumers' responses to determine whether the relevant preference function to be estimated is interval, ratio or ordinal in nature. This paper shows that the statistical tests to determine the nature of the preference function in one of the more popular structures often introduce bias in subsequent analysis. An alternative approach is indicated for use in some situations, which avoids some of the complexities involved in the more commonly used tests.