Completely Unacceptable Levels in Conjoint Analysis: A Cautionary Note.

An experiment is reported on the extent to which respondents adhere to the implications of choosing the "completely unacceptable" level in hybrid conjoint (and related) applications. The findings indicate that the form of the instructions matters, but that respondents often ignore the implications of previous responses when responding to full-profile options containing unacceptable attribute levels. The authors discuss the impact of this inconsistency on internal predictive validity in both empirical and theoretical terms. [ABSTRACT FROM AUTHOR]