Put simply, when you come up with a class/category taxonomy, you're drawing your own boundaries inside a continuum of variation, and where you draw them is something you're choosing based on the purposes you are pursuing and the assumptions you have made, not mapping out something that's intrinsic to the external world independently of you. "Category X includes object Y" is a property inherent to the definition of category X; it is not a quality inherent to object Y.
And when the things you're classifying are other people, and you're trying to develop a taxonomy that models their behavior, you have a situation where the entities you're analyzing are self-aware beings whose behavior is driven by the analytical models and taxonomies in their minds, not the ones in yours.
People who propose class theories as a usable model for understanding human intentions and behavior lump other people together according to their own criteria, even though those people can -- and often do -- make their own decisions according to completely different criteria. People do not see strangers as having common interests or values simply because a third set of strangers have reduced them to data points that fall into the same bar on a narrowly tailored histogram.
When put to the test, these class theories consistently fail to predict actual human behavior, and people with normative attachements to the theories then to hand-wave away these discrepancies: "false consciousness", "voting against their interests", etc.; these notions are layers of epicycles meant to sustain a speculative theory in the face of falsifying evidence.
Much better to do away with a priori class theories, and just analyze what actual people are actually saying and doing.