I don't see the author claiming that. You evaluate someone as an individual when you're specifically talking about an individual. In this case it's about groups so those are the results being discussed.

Group level analysis is also useful as a backdrop before diving into specifics even when discussing a specific individual and has been shown to have predictive power at every level.