Personality Differences Between the United States and the Netherlands: The Influence of Violations of Measurement Invariance

摘要

Cross-cultural comparisons of personality have yielded inconsistent results, which might be partly due to poor model fit and disregard of differential item functioning (DIF). Here, the brief Dutch form of the Multidimensional Personality Questionnaire (MPQ-BF-NL) was tested for cross-cultural measurement invariance (MI), using multiple group confirmatory factor analysis (MGCFA) for categorical outcomes. Representative Dutch and U.S. samples (both Ns = 1,055) were used for model exploration and student samples (both Ns = 410) for cross-validation. The MPQ-BF-NL appeared partially strict invariant. In all, 19% of the items contained DIF when scales were treated separately, while in the full model, allowing for 150 cross-loadings, 40% of the items contained DIF. The majority of DIF was observed in thresholds (i.e., difficulty parameters). Not accounting for DIF would yield invalid inferences for 4 out of 11 primary scales. Higher corrected factor scores were evident in the U.S. sample for Social Closeness and Stress Reaction, whereas lower for Achievement and Aggression, respectively, suggesting that the U.S. society, relative to the Dutch, fosters more community and less agency.