On the Comparability of Constructs in Cross-Cultural Research

Abstract

A critique of Cheung and Rensvold’s article describing the use of mean and covariance structures (MACS) techniques for assessing bias in cross-cultural research is offered. In particular, the author highlights the critical distinction between item bias and construct bias. Although the author’s view is in fundamental agreement with Cheung and Rensvold’s primary argument that MACS analyses are a very useful tool to determine comparability of constructs across cultural settings, The author disagrees with their conclusion that measurement invariance indicates a lack of construct bias. The author argues that cross-group equivalence of all reliable measurement parameters only indicates a lack of differential item bias and does not indicate lack of construct bias. Finally, two related and important issues in MACS analyses are discussed, namely, (a) the choice of rationale for significance testing and (b) the effects of different identification methods on estimating latent mean levels.