So there does seem to be truth to if you are black you are more likely to be on food stamps

Sure but in context that should not matter at all if the question is what demographic is the appropriate choice to represent the typical US welfare recipient. In this case total numbers are more relevant than relative proportion, you'd want the median person.