Constructing Explicit Prejudice: Evidence From Large Sample Datasets