Methodological issues in using structural equation models for testing differential item functioning

In cross-cultural studies, groups often differ in various characteristics (e.g., demographics, socioeconomic status, language, culture, etc.) and these characteristics may not be relevant to the goals of a particular study. Even when reasonable precautions have been taken to prepare a test or survey that is equivalent across cultural groups, it is possible that the attribute being measured has different conceptual meanings in different groups (de Beuckelaer, Lievens, & Swinnen, 2007) or that some items have different importance for one group more than another (Cheung & Rensvold, 1999). In such cases, observed group differences may represent measurement artifacts related to the instrument rather than true differences on a relevant construct. This disparity between observed and true group differences, in turn, adversely affects the comparability of their scores (Byrne & Stewart, 2006; de Beuckelaeret al., 2007; Raju, Laffitte, & Byrne, 2002; Vandenberg & Lance, 2000; van de Vijver & Poortinga, 1992). Thus, researchers have highlighted the importance of measurement equivalence as a prerequisite for meaningful group comparisons (Drasgow, 1984; Little, 1997; Reise, Widaman, & Pugh, 1993). Accordingly, standards established by both the American Psychological Association (APA) and the International Test Commission (ITC) have emphasized evaluation of measurement equivalence for fair use of a scale (1999).