Standard Article

Unobserved Heterogeneity

Part 2. Marketing Research

  1. Neeraj Arora

Published Online: 15 DEC 2010

DOI: 10.1002/9781444316568.wiem02073

Wiley International Encyclopedia of Marketing

Wiley International Encyclopedia of Marketing

How to Cite

Arora, N. 2010. Unobserved Heterogeneity. Wiley International Encyclopedia of Marketing. 2.

Author Information

  1. University of Wisconsin, Madison, WI, USA

Publication History

  1. Published Online: 15 DEC 2010

Abstract

In marketing, it is important to recognize differences between individuals, or heterogeneity. When observables such as consumer demographics are used to investigate any patterns in the data via a model, we call it observed heterogeneity. The remaining unaccounted heterogeneity in model parameters is called unobserved heterogeneity. A limitation of a model that ignores heterogeneity is that it is highly restrictive – marketers like to segment the market based on interpersonal differences in price sensitivity, product feature (e.g., low fat, high fuel efficiency) sensitivity, and brand preference. Ability to accurately characterize unobserved heterogeneity requires multiple observations for each unit of analysis. Common contexts involve syndicated data available for most packaged goods and conjoint analysis (ratings and choice-based) data, where the very purpose of the experiment is to uncover the heterogeneity in relative importance of product features across the population of interest. Recent innovations in computing power and Bayesian simulation-based methods of model estimation have significantly simplified our ability to estimate models with unobserved heterogeneity.

Keywords:

  • demographics;
  • logit;
  • bayesian;
  • panel data;
  • conjoint