Correlation is not causation. And I wouldn't ask you to explain why equally poor zips don't have the same murder rate, since it's obvious you've never considered controlling the data that way.