Working Papers

Adaptive Designs for Likert-Type Data: An Approach for Implementing Marketing Surveys

Jan 1, 1990

Reprinted with permission from the Journal of Marketing Research, Vol. XXVII (August 1990), 304-321. Copyright © 1990 by the American Marketing Association.

Describes a Latent Trait Theory (LTT)-based approach to implementing marketing surveys, termed adaptive survey designs, in which the survey is adapted to each respondent.

Type of Report
Provides general background on graded-response LTT models and compares them to common-factor models; develops the notion of adaptive survey designs for Likert-type data; presents a simulation study to demonstrate the efficiency of adaptive designs.

To assist marketing researchers in understanding the theory, methods, and advantages of adaptive survey designs.

Presents a graded-response LTT model and uses it to describe a novel approach for implementing efficient marketing surveys. Discusses the key decisions involved in designing an adaptive survey. Provides results from a simulation of adaptive survey design used to measure consumer discontent. Suggests specific guidelines for researchers and practitioners in using adaptive systems.

Marketing researchers and practitioners who are interested in using computerized interviewing methods to obtain better quality survey data with greater efficiency.

Main Points
We describe a novel approach for conducting surveys in which the questionnaire is tailored to individual respondents. An adaptive design is best implemented by a computerized facility (e.g., telephone interview- and computer-based questionnaire and response-entry facility) and involves two basic steps performed iteratively:

  1. estimate the respondent’s attitude level after the response to every given item, and
  2. select the next item to administer to the respondent so that the item provides the most information around the respondent’s estimated attitude level.

Using a computer program, ADLMAX, we simulate an adaptive design for measuring consumer discontent with data collected from 235 consumers. When the simulation study is evaluated by multiple criteria, it reveals that an adaptive design is significantly more efficient (i.e., comparable information is obtained with fewer items) in comparison to conventional survey methods.

Marketing researchers and practitioners should pay greater attention to adaptive survey designs because of several reasons. First, the notion of item information function, which describes the amount of information available at each level of the underlying trait from each and every item on the questionnaire, represents a major advance in our understanding of survey items and opens new ways to think about questionnaire designs and obtaining measurements.

Second, initial studies in marketing combined with earlier research in educational psychology appear to suggest that the promised advantages of adaptive designs can be realized. Specifically, evidence at this point indicates that efficient marketing surveys can be practically implemented.

Third, and perhaps most importantly, such efficiencies in implementing marketing surveys are likely to yield several direct and indirect payoffs. Administration as well as data handling and manipulation costs are all expected to decline. These direct cost savings are likely to be substantial (e.g., between 33 and 50 percent).

Marketing researchers and practitioners must realize that adaptive designs are technologically intensive and theoretically complex. Consequently, advances in adaptive design depend on a sense of cooperation between practitioners (because they provide technologically intensive equipment) and academic researchers (because of their expertise on the underlying measurement theory). Interestingly, at a time when the gap between marketing practitioners and researchers appears to be widening, adaptive designs offer an opportunity to bring them closer by capitalizing on their strengths. Adaptive designs offer a promise of lasting impact on marketing research practices in 1990 (i.e., in data collection scale development, measurement precision) by combining the elegance of measurement theory (graded-response LTT model) with the technological advancements of 1980 (CAQ/CRT interviewing technology). We hope the marketing profession can enjoy the benefits from this mix of theoretical elegance and technology.

About the Authors
Jagdip Singh is Assistant Professor at Case Western Reserve University. Roy Howell is Professor of Marketing at Texas Tech University. Gary Rhoads is Assistant Professor at Idaho State University.

This resource is not available online.

By using you agree to our use of cookies as identifiers and for other features of the site as described in our Privacy Policy.