Good measurement instruments have to meet certain necessary conditions, especially if they are to be used in cross-cultural research. In this study, we examine whether the German and Japanese translations of Shalom Schwartz's PVQ-57 fulfill two of these conditions. The survey measures 19 human values by 57 indicators. First, we examine whether these human values are reliably measured in a student survey which was carried out in autumn/winter 2018 at German and Japanese universities. Second, we investigate whether the instruments allow the comparison of latent value means across the German and Japanese student samples. Such a comparison is meaningful only if scalar equivalence, a high form of measurement equivalence, is reached. Results suggest that at least partial scalar invariance can be established across the samples. We discuss the findings in the broader perspective of value research and outline topics for further investigation.
Anzeige
Bitte loggen Sie sich ein, um Zugang zu Ihrer Lizenz zu erhalten.
Bold letters in italics indicate that values are, technically speaking, vectors. We never refer to a set of several values so that it is unnecessary to distinguish between matrices and vectors.
If the latent variable has unit variance \({(\sigma }_{v}^{2}=1\)), then under the model assumptions the explained variance is simply the squared sum of the factor loadings. Cronbach’s α is calculated under the simplifying assumption that all loadings are equal (\({\lambda }_{1}+\dots +{\lambda }_{J}=\lambda\)) so that the explained variance becomes \((J{\lambda )}^{2}\).
The German version of V17 [show her/his abilities] and V48 [achievements are recognized]—clearly indicate an extrinsically motivated orientation. The third indicator, V32 [be successful], is ambiguous in this respect: we may strive for success because it gives us an inner feeling of satisfaction or because our success is socially recognized.
If the equality constraint is relaxed in the PSE models, the estimates of the freed parameters of the groups may largely differ. Other approaches—the Bayesian approximation approach or alignment optimization approach—may produce smaller differences. For a comparison of these approaches in a Monto Carlo simulation, see [7]. We cannot say, however, that one of the approaches is superior to the others. Ultimately, the theoretically most plausible model should be selected.