Accuracy, Consensus, In-Group Bias, and Cultural Frame Shifting in the Context of National Character Stereotypes

1. The instructions and rating scale, in which participants were directly asked to compare Finns and Swedish-Finns, were essential for being able to investigate the expected compensation effect (CitationYzerbyt et al., 2005). The expected negative correlation between warmth and competence judgments could appear to conflict with the firmly established halo-effect (CitationRosenberg, Nelson, & Vivekananthan, 1968; CitationThorndike, 1920). According to the halo-effect a general good vs. bad feeling will color the judgment of the separate qualities of the target. Resolving this conflict, CitationJudd et al. (2005) showed that when participants judged a single target, either described as high or low on warmth or competence, a clear halo effect emerged: those described as high on one dimension were judged to be high also on the other dimension. The compensatory effect between warmth and competence judgments only appeared when participants were instructed to compare two targets with each other—this method was therefore employed in the present research.

2. ICCs are quite similar to Pearson's correlation coefficients, but there is one important difference. Whereas Pearson's rs are only sensitive to the shapes of the profiles, ICCs are sensitive to both the shapes of the profiles and to differences in elevation. This has been argued to make ICCs more appropriate for calculating profile similarity. Empirically verifying this assertion, CitationMcCrae (2008) compared different measures of profile agreement and found double-entry ICC to be the best. However, CitationFurr (2010) could not replicate these findings, and argued that the Pearson's rs could instead be used—being responsive only to the shape of the profile, the Pearson's r is clearer to interpret. We therefore ran all analyses using both ICCs and Pearson's rs. In general, the correlations between profiles tended to be somewhat stronger when using the Pearson's rs, suggesting the occurrence of mean-level differences between the profiles that were captured only by the ICCs. However, as none of our conclusions changed regardless of which method we used, except in one set of analyses for which we will report both ICCs and Pearson's rs, we report only the ICCs. We prefer the ICCs because they should improve comparability with previous research on national stereotypes (e.g., CitationTerraciano et al., 2005; CitationTerraciano & McCrae, 2007, 2008; CitationRealo et al., 2009). Regarding the interpretation of ICCs, similarly to Pearson's rs, the square of the ICC is an estimate of the amount of variance of one variable that another variable can account for

3. The desirability ratings were gathered in 2006–2007 from three Finnish language gymnasiums (located in Raahe, Lumo and Helsinki) different to those from which the stereotype ratings were gathered. The procedure was similar to the one used to gather stereotype ratings: the questionnaires were administered on-line in various classes, participation was voluntary, and as an incentive, participants were given feedback on their personal values profiles. Altogether 281 (95 men, gender data missing for one class consisting of 26 students) students out of the around 290 who were approached decided to take part in the research. The mean age of the participants was 16.9 years (SD = 0.9). For desirability ratings, each of the sixty S5 items was rated with the following instructions: “Descriptions of people are often evaluative. Some characteristics are judged as socially desirable, and others as undesirable. Indicate for each item whether the characteristic described in the item is socially desirable, undesirable, or neutral”. The ratings were made on scale from –3 (wholly undesirable) to 3 (wholly desirable).