Social groups are roughly evaluatively equivalent but perceivers exaggerate differences favoring their own group

Evidence

With social groups, it is likely that actual group differences along these valenced dimensions (as portrayed in figure 8.1) tend to cancel one another out across multiple attribute dimensions. That is, social groups on average are probably more or less evaluatively equivalent. Consider two social groups, A and B. On some valenced attribute dimensions A is likely to be higher than B, while on others the group difference would be reversed. Nevertheless, social perceivers may be biased in reporting these valenced group differences, such that they may tend to see systematically larger group differences on dimensions that favor one group over the other (say A over B) and smaller group differences on dimensions that favor the other group (say B over A). We use the term valence inaccuracy to refer to the systematic tendency to overestimate the evaluative difference between groups across multiple valenced attribute dimensions. — John F. Dovidio, Peter Glick, Laurie A. Rudman, On the Nature of Prejudice, 862

Categorization

• DEFINE: Defines ‘valence inaccuracy’: a novel third kind of stereotype inaccuracy distinct from the traditional mean/dispersion exaggeration forms.
• QUESTION: Is there empirical support that actual group differences cancel out across valenced dimensions, or is this purely an assumption the authors are asserting?
• corroborated: 3 independent passages support this claim — John F. Dovidio, Peter Glick, Laurie A. Rudman, On the Nature of Prejudice: Fifty Years…, 862