Race from a sociopolitical identity standpoint is about phenotype and not genotype. The phenotype includes things very disparate from genetics including dress, posture, speech pattern, occupation, etc. As an example, people often have disparate representations of African blacks as they do American blacks.
Medically, things are much more complicated. Racial classifications are useful for study of disease, but they include much more than genetic determinants and almost as a rule are multiply determined with the various determinants being largely unknown.
This study, though, mostly just states how evolution works. Traits are determined, generally, by a whole variety of genes, which is a great thing because a genetically diverse population already has genetic mechanisms to adapt to significant environmental changes. But within a diverse multiply determined pool are a subset with the right largely arbitrary combinations of genes such that a single mutation can introduce a very significantly different phenotype. And humans are tribal, so even peoples that look very much the same but breed in separate pools can have very different genetics underlying their sameness. So instead of a random distribution of combinations of 8 genes (2^8 = 64 possibilities), you might have 10 tribes that represent all 8 of those genes but in different combinations that are very much alike within a tribe. This allows single mutations or interbreeding (increased in times of distress) to produce with much higher than random frequency offspring with very different traits that might be adaptive.
Without this setup, we'd either lack the genetic diversity to adapt to a significant environmental change, or we'd have the diversity but not be able to make the phenotype common enough to adapt through selective breeding, or our populations would be very genetically unstable.
