The confounding effects of consumer heterogeneity on model-based inference of attribute non-attendance.

Several empirical studies conclude that a majority of economic agents ignore some of observed product attributes when choosing among discrete alternatives. Many of these ndings are based on latent class logit with partially constrained support points wherein the share of each point is interpreted as the probability of ignoring particular attribute(s). We note that because the logit kernel is mixed over these points to approximate unmodeled interpersonal taste variation during the estimation stage, the interpretation of estimated shares is necessarily ambiguous. Using simulated examples, we explain why common forms of unobserved consumer heterogeneity can be confounded with attribute non-attendance.