Academic Insights: An Application of Multiple-Group Causal Models in Assessing Cross-Cultural Measurement Equivalence

Abstract

In cross-national marketing research, the question whether measurement accuracy, reliability, and validity are achieved across samples traditionally has hampered research efforts. Measurement equivalence across samples has particularly perplexed researchers in their efforts to evaluate responses regarding latent variables. Recently, multiple-group structural equations modeling has been suggested as a reliable method for determining measurement equivalence. To date, however, the use of this approach has remained limited. In this study, the authors apply multiple-group structural equations modeling to assess measurement equivalence in three distinct constructs across U.S. and Korean samples on data derived from cross-cultural advertising research. The authors propose an extension of this method using the measurement error covariance matrices, which will enable researchers to evaluate measurement reliability across samples better and to attempt to disentangle cultural differences in instrument usage from measurement-related differences. The authors evaluate the results and discuss the findings, as well as outline the implications and limitations of this method for further research.

Data equivalence in cross-cultural international business research: assessment and guidelines

“…Despite the importance of measurement equivalence, and the heightened attention paid to it in the literature (e.g., Craig & Douglas, 2000;Mullen, 1995;Myers, Calantone, Page, & Taylor, 2000;Steenkamp & Baumgartner, 1998), the results indicate that all three aspects of measurement equivalence are rarely fully reported or established. In part, this may reflect difficulty in the implementation of the procedures to establish calibration, translation, or metric equivalence.…”

“…Post data collection, metric equivalence (inclusive of scoring consistency and scalar equivalence) can be assessed in a number of ways, including multi-group structural equation models (SEM), profile analysis, optimal scaling, and regression analysis, and by comparing the standard deviations and means of the subjects' responses over a large number of items across cultures (Bollen, 1989;Mullen, 1995;Myers et al, 2000;Salzberger, Sinkovics, & Schlegelmilch, 2001). Scoring consistency can be checked by comparing reliabilities between groups, or by examining factor loadings and measurement error variances.…”

“…To establish scalar equivalence, pooled analysis (i.e., deculturing data by standardizing the responses to each observable variable within each sample separately, and removing scaling factors from the measurements) and adjustment factors for differences in reliability are typically used (Davis et al, 1981;Durvasula, Andrews, Lysonski, & Netemeyer, 1993;Singh, 1995). For example, the extant literature suggests alternating least-squares optimal scaling, which is a general extension of principal component analysis for use with non-metric or mixed metric data (Mullen, 1995;Myers et al, 2000;Salzberger et al, 2001), or using multiple respondents to assess consistency of results (e.g., Calantone, Schmidt, & Song, 1996;Davis et al, 1981). However, the most prevalent approach to address scalar equivalence has become hierarchical diagnoses of invariance of measurement using Data equivalence in IB research multiple group structural equation modeling (e.g., Mullen, 1995;Myers et al, 2000;Steenkamp & Baumgartner, 1998, 2001).…”

“…For example, the extant literature suggests alternating least-squares optimal scaling, which is a general extension of principal component analysis for use with non-metric or mixed metric data (Mullen, 1995;Myers et al, 2000;Salzberger et al, 2001), or using multiple respondents to assess consistency of results (e.g., Calantone, Schmidt, & Song, 1996;Davis et al, 1981). However, the most prevalent approach to address scalar equivalence has become hierarchical diagnoses of invariance of measurement using Data equivalence in IB research multiple group structural equation modeling (e.g., Mullen, 1995;Myers et al, 2000;Steenkamp & Baumgartner, 1998, 2001). For example, Steenkamp and Baumgartner (1998) provide detailed procedures for testing full and partial measurement invariance that offer a more comprehensive diagnosis than other techniques and simultaneous test of samples.…”

The Dissimilar Significance of Functional and Experiential Beliefs when Marketing Brands in Cross-Cultural Settings

“…Finally, the multi-group analysis in SEM was conducted to define the level of measurement invariance by comparing the fit of a baseline model (unconstrained) with other four constrained models, including: i) equal covariance, ii) equal lambdas, iii) equal covariance and lambdas; and, iv) equal covariance, lambdas, and error variances (Mullen, 1995;Steenkamp and Baumgartner, 1998;Myers et al, 2000). The results of Chi-square and degree-of-freedom and other fit indices (e.g.…”