I have no objection to calling "almost all criminals vote Democratic" a "claim" rather than a "statistic," although I'm not sure what it changes. I still doubt the claim, and, even if there is no pre-defined numeric threshold at which one can say precisely that one more criminal voting Democratic would mean "almost all" do, and one less would mean the contrary, then there are certainly quantitative data that could be judged to refute it—for example, if fewer than half do—or that could be judged to be evidence for it—or example, if 90% do. I think it is reasonable to ask for a reference to some confirmatory evidence for such a statement.