The Pitfalls of Poor Psychometric Properties: A Rejoinder to Hofstede's Reply to Us

Hofstede (this issue) has taken exception to our conclusions (Spector, Cooper, Sparks, et al., 2001) that the poor internal consistency reliabilities of his Values Survey Module 1994 (VSM94) should be cause for concern. He suggests that (1) individual level psychometric properties for a scale are irrelevant when one uses it to make inferences to a group (or country) level, (2) that our samples were poorly matched on demographics (and were not very representative) and this prevented our finding internal consistency at the country (ecological) level, and (3) the VSM94 must be reliable because there is so much evidence for validity. We will address each of these points, explaining our case for concern. INDIVIDUAL LEVEL PSYCHOMETRIC PROPERTIES DON’T MATTER Hofstede argues that the VSM94 was designed to assess values at the aggregate (country) and not the individual person level. Therefore, the psychometric properties at the individual level are irrelevant. We first wish to point out that we reported the internal consistencies at the aggregate level, and results were not appreciably different from those at the individual level. Furthermore, Hofstede is absolutely correct that one must be careful about the ecological fallacy in mixing levels when This is the post-printed version of an article. The final published version is available at Applied Psychology: An International Review 51:1 (2002); doi: 10.1111/1464-0597.00085 ISSN 0269-994X (Print) / 1464-0597 (Online) Copyright © International Association for Applied Psychology, 2002. Published online: 08 Apr 2002