Aren't these proportions of the voting population, not the actual population?

Seems like a bit of a jump to assume they're distributed the same way and start making conclusions based on that.