Distinguishing taste variation from error structure in discrete choice data

We propose as a practice that researchers investigate other sources of heterogeneity besides taste variation in the search for parsimonious recommendations about new product development and marketing program design. Some evidence exists that regularities in choice processes may be more common than previously thought (Louviere and Swait, 1996, Louviere et al., 1999; see also Stigler and Becker, 1977). In fact, it may be that taste homogeneity is more prevalent than expected, if we recognize other sources of heterogeneity properly. In this paper we show how discrete choice models confound taste heterogeneity and differences in error structure. We then illustrate the use of the Tree Extreme Value (TEV) model to investigate taste homogeneity in three trans-oceanic air travel markets, while controlling for error structure heterogeneity. We conclude that partial taste homogeneity exists across the markets, despite accentuated cultural differences; in addition, two of the routes exhibit a much higher degree of taste homogeneity, despite a significant difference in trip length.