Professor Christian Welzel is Chair in Political Culture Research, Center for the Study of Democracy at Leuphana University, and Vice-President of World Values Survey Association.

In the special panel on measurement invariance and validity in cross-cultural research, he will give a speech based on the recent work: Welzel, C., L. Brunkert, R. Inglehart & S. Kruse (2019). “Measurement Equivalence? A Tale of False Obsessions and a Cure.” World Values Research 11 (3): 54-84.

Abstract:

During the last decade, Multi-Group Confirmatory Factor Analysis (MGCFA) has risen to the unchallenged gold standard of establishing measurement equivalence in cross-cultural research. A key premise of MGCFA is that multi-item constructs are incomparable between countries when their dimensional cohesions within countries are dissimilar. We argue that this comparability logic is lopsided because, in fact, dimensional cohesions within countries are incomparable between countries. The reason is a simple, albeit largely unknown, principle: how strongly coherent a construct can appear within countries is a function of the country means’ scale positions. Indeed, as country means approach the extreme ends of a closed scale, dimensional cohesion statistics deteriorate and make the respective construct appear less coherent. But this impression is deceptive because the correlation calculus that underlies every dimensional analysis operates within narrower margins as country means turn more extreme. Hence, seeming variability in dimensional cohesions is an altogether inconclusive indication under greatly varying country means. Because of that, such variability proves irrelevant to measurement equivalence properly understood—which is present when similar overall scores on a construct map in similar fashion on other variables of interest. We exemplify the latter point using a most prominent victim of MGCFA-based incomparability verdicts: the Emancipative Values Index (EVI). Our insights re-enforce Welzel and Inglehart’s conclusion that the comparability of multi-item constructs should be judged by their predictive powers across countries, instead of their dimensional cohesions within countries, even more so in recognition of the fact that these two criteria operate against each other.