CFE 2019: Start Registration
View Submission - CMStatistics
B1818
Title: E-consumers' attribute non-attendance switching behavior: Effect of providing information on attributes Authors:  Michel Meulders - KU Leuven (Belgium) [presenting]
Leonard Maaya - KU Leuven (Belgium)
Martina Vandebroek - K.U. Leuven (Belgium)
Abstract: Choice experiments are used to investigate how product attributes affect product preference. A choice experiment consists of multiple sets of designed product alternatives that are described by a combination of attribute levels. Respondents are asked to choose the most preferred alternative from each choice set. Based on random utility theory, standard multinomial logit choice models can be used to estimate the utility of attribute levels. Standard choice models assume that respondents examine all attributes in the same fully compensatory manner. However, research has indicated that respondents may ignore part of the attributes if the choice task is too complex. Accounting for such Attribute Non-Attendance (ANA) is important as failure to account for it may lead to biased preference estimates. We develop a dynamic mixture latent Markov model to model the dynamics in attribute non-attendance behavior due to providing information about key-attributes in the course of the choice experiment. The model is illustrated with an application on e-consumers' preferences for webshops. The results indicate that providing information about attributes leads to an increase in the attendance probability of those attributes. Moreover, a dynamic ANA model fits the data better than a model that assumes a change in the preference parameters.