Do polygenic score shifts reflect actual differences in traits, and are they due to natural selection?