The Identical Attribute Value Assignment Problem in Conjoint Experiment

Abstract

In this paper, we report a novel problem in conjoint experiment with the assignment of identical attribute values in a task. Similar to the relationship between intention-to-treat estimate (ITT) and the complier average causal effect (CACE) or local average treatment effect (LATE), AMCE is calculated as AMCE from heterogeneous attribute assignments (AMCE) multiplied by the share of heterogeneous attribute assignments. In othere words, AMCE is computed not directly from the full sample but indirectly from a "complier" subpopulation who receive heterogeneous attributes. By construction, AMCE considers those who receive identical attributes as (one-sided) noncomliers.