The value of consideration data in a discrete choice experiment

In stated preference surveys, data regarding the considered alternatives is sometimes collected prior to preferred alternative. When chosen alternative not in consideration set, inconsistent with choice data. Several modeling approaches have been used such situations. Some researchers ignore and assume all are considered. Others only use consistent delete observations. The most intricate methods a latent set formation approach process. We extend model incorporate but allow for inconsistencies data, individual-level heterogeneity compare recovery of mean population parameters our existing through simulation. results show that if there similar effect attributes both phase phase, mixed logit outperformed by two-stage models. contrast, when sufficiently different models can recover better than model. Furthermore, we conclude having barely improves compared uses Finally, illustrate using empirical about preferences mobile phones.