Contributions

  • Ainslie, Andrew - Contributor
  • Harvard Business School - Contributor

Publication

2008 - Harvard Business School, Boston, Massachusetts

Language

English

Word Count

6,500 words, Guess

Page Count

26 pages

Identifiers

Description

The purpose of this paper is to show that allowing for taste heterogeneity does not address the similarity critique of discrete-choice models. Although IIA may technically be broken in aggregate, the mixed logit model allows neither a given individual nor the population as a whole to behave with perfect substitution when facing perfect substitutes. Thus, the mixed logit model implies that individuals behave inconsistently across choice sets. Estimating the mixed logit on data in which individuals do behave consistently can result in biased parameter estimates, with the individuals' tastes for desirable attributes being systemically undervalued.

Subjects

Reader Reviews

No reviews yet for this book.

Be the first to share your thoughts!