Measurement Invariance Issues in International Management Research

Abstract

In this dissertation, the question was addressed as to how meaningful (and reliable) cross-country comparisons are when such comparisons are based on (estimated) factor mean scores at country-level. A comparison of factor mean scores across countries is very common in international management research. International management research comprises a very wide spectrum of areas of research (e.g. globalisation of the marketing mix, organisational behaviour, cross-cultural communication, etc.). A discussion as to how international comparisons are made within all of these areas would not be feasible. For this reason, the focus in this dissertation was on two particular areas of international management research: global consumer research and global research in international Human Resource (HR) management. These areas were selected as they are of particular importance to multinational companies. The main methodological objective of this dissertation was to investigate the extent to which violations of the principle of measurement invariance across groups (e.g. nations or cultures) lead to wrong conclusions regarding factor mean comparisons across groups. To date, there is no consensus in the methodological literature as to what level of measurement invariance across groups is required before such cross-group comparisons at factor-level are meaningful. Some authors, for instance William Meredith (1993), firmly stated that factor loadings and indicator intercepts of all indicators need to be invariant across groups (i.e. ‘tau-invariance’ across groups). Others have advocated less stringent conditions, such as the invariance of factor loadings across groups (i.e. ‘metric invariance’ across groups) (e.g. Duane Alwin and David Jackson, 1981) or the invariance of only a subset of all factor loadings across groups (i.e. ‘partial metric invariance’) (e.g. Barbara Byrne et al., 1989). References to many papers were provided in Section 3.1 of Chapter 3. Metric (or partial metric) invariance across groups does not require the invariance of indicator intercepts across groups.