Responding Weighting in HB

Keywords

preference heterogeneity, utility shrinkage, market simulation

Abstract

CBC/HB is often used to analyze CBC studies. A recent Sawtooth Software survey indicates that among those who used CBC in the previous year, 69% of users employed CBC/HB for their final models (Sawtooth Solutions, Fall 2007). Often these CBC studies cover different market segments that may have heterogeneous preferences. These different preferences could be related to differing demographics such as country of residence, gender, or whether the respondent has bought specific products in the past. One of the properties of CBC/HB is that individual respondent’s utilities are shrunk toward the average utility of the population for each attribute level. This has been a concern for samples that are not representative. An over-sampled group (customers for instance) could have a disproportionally large impact on the utilities of the under-sampled group. The effect of this is that products appealing to the larger group in a market simulator will have their shares overstated while products appealing to the smaller group would have their shares understated. Since the bias is inherent in the utilities, weighting the market simulator will not completely correct for the misrepresentation of the shares. The purposes of this paper are to examine the severity of the bias that comes from non-representative samples and to propose a possible solution. Because the problem manifests itself as bias in the utilities, the paper will use a simulation study to examine the results.

Original Publication Citation

Howell, John R. (2007), "Respondent Weighting in HB", Proceedings of the Sawtooth Software Conference, Sequim, WA.

Document Type

Conference Paper

Publication Date

2007

Publisher

Sawtooth Software Technical Papers

Language

English

College

Marriott School of Business

Department

Marketing

University Standing at Time of Publication

Assistant Professor

Link to Full Text

Share

COinS