Executive Summary
Accurate measurement lies at the heart of marketing research. Although the measurement of marketing constructs has greatly improved in recent years, several topics have received insufficient attention in marketing academia and practice. First, in many areas of research, it is necessary to draw a clear distinction between the trait and the state aspects of a construct. The measurement model the authors propose enables a sophisticated investigation of the stable and transient components of a construct and avoids the problems associated with assessments of stability through test–retest correlations.
Second, the extended paradigm for scale development explicitly accounts for various sources of measurement error and enables a much more detailed diagnosis of the quality of individual items and the measurement scale as a whole. Furthermore, it effectively accounts for the biasing effects of random and systematic measurement error, which can be extremely damaging to the validity of the substantive conclusions derived from the research.
Third, the procedure incorporates the item and scale means into the measurement analysis, makes assessments of measurement invariance an explicit component of the scale validation process, and ultimately leads to cross-sectional and longitudinal comparisons of means (and relationships between constructs) that are methodologically justified.
The authors present an extended application of the model to the constructs of brand loyalty and deal proneness to illustrate the benefits of the proposed procedure. They analyze commercial data collected by a global market research agency based on its consumer panel in the Netherlands. The authors find that approximately 25% of the total variance is systematic (nonrandom) measurement error, most of it stable over time. Moreover, scale reliability using the conventional procedure (e.g., Cronbach’s a) is substantially overestimated. This result is probably not unique to the current context, because the sources of systematic error will also be present in other scales. In addition, the authors find that ignoring systematic measurement error has a substantial biasing effect on the results because the correlation between the substantive constructs of brand loyalty and deal proneness is underestimated by a factor of almost 3 ( .25 versus –.70). In general, the direction of the bias depends on the sign of the true correlation. Measurement errors are typically positively correlated. If the true correlation between two constructs is negative, it will be underestimated, as in the current illustration. If the true correlation is positive, it will be overestimated.
The proposed methodology requires data that are more effortful to obtain than those used in conventional procedures. That is, large samples of respondents (typically 500 plus) must be surveyed at least three times. Although the requirement of three repeated measurements may seem daunting, the benefits can be substantial, as shown in the current application. Furthermore, the authors believe that there are at least two opportunities for collecting such data.
First, consumer panels organized by market research agencies and universities (particularly Internet panels) are natural candidates for repeated surveys. They make the data collection less burdensome, guarantee a reasonable response rate, and, as an added bonus, make the results more generalizable because of the use of a heterogeneous participant pool. Second, participants in a subject pool or students enrolled in a class can be surveyed repeatedly throughout the semester or even during the course of the year. For example, the data used in the current application consist of nine items, and it should not take more than ten minutes to complete the questionnaire on a given occasion.
Biography
Hans Baumgartner (PhD, Stanford University) is Professor of Marketing, Charles & Lillian Binder Faculty Fellow, and Director of PhD Programs in the Smeal College of Business at Pennsylvania State University. His research interests are in the areas of consumer behavior and research methodology. He has published articles on these topics in Journal of Consumer Research, Journal of Marketing Research, Marketing Science, Journal of Marketing, Journal of Consumer Psychology, International Journal of Research in Marketing, Organizational Behavior and Human Decision Processes,and Journal of Economic Literature. He is a past associate editor of Journal of Consumer Research and currently serves on the editorial boards of Journal of Consumer Research, Journal of Consumer Psychology, International Journal of Research in Marketing, Journal für Betriebswirtschaft, and Marketing: Journal of Research and Management.
Jan-Benedict E.M. Steenkamp is C. Knox Massey Professor of Marketing and Marketing Area Chair in the Kenan-Flagler Business School at the University of North Carolina, Chapel Hill. He has published in Academy of Management Journal, International Journal of Research in Marketing, Journal of Consumer Research, Journal of Marketing, Journal of Marketing Research, and Marketing Science. He serves on the editorial boards of International Journal of Research in Marketing, Journal of Consumer Research, Journal of Marketing, Journal of Marketing Research, and Marketing Science and has been editor of International Journal of Research in Marketing.
J Marketing Research, Volume 43, Number 3, August 2006
View Table of Contents.