Conferences

In recent years, researchers using international survey data have become more concerned with the cross-cultural comparability of their measurements. Before attributes can be validly compared across cultural groups, it has to be shown that they are measured in a sufficiently equivalent way. By now, there is considerable agreement on the concrete operationalization of (various levels of) measurement equivalence and on analytical tools to test for equivalence (i.c. the multi-group CFA framework).

There is less clarity, however, on what one should do when confronted with measurement scales that are not cross-culturally equivalent—something that occurs not infrequently in practice. In the literature, a number of strategies for dealing with measurement inequivalence have been put forward. A very promising proposal was made by Poortinga (1989), who suggests to give a substantive interpretation to measurement inequivalence. Inequivalence is then treated as a useful piece of information on cross-cultural differences. Yet, to the best of our knowledge, there have never been elaborated concrete guidelines on how researchers could proceed in a systematic way to interpret inequivalence. Probably because of this lacuna, this potentially very useful strategy is applied only very rarely.

This paper makes an attempt to fill this gap. We show how a relatively new statistical tool, namely multilevel structural equation modelling, can be used to test whether deviations from measurement equivalence can be explained by relevant context variables. Finding such context effects can provide insight on why particular items function differently across cultures. Applications on ISSP data are presented.

References:

Poortinga, Y.H. (1989). Equivalence of cross-cultural data: an overview of basic issues. International Journal of Psychology, 24, 737-756.

Key words: measurement equivalence, multilevel SEM, cross-cultural research