I think the rule of thumb is that articles containing statistics are guilty until proven innocent. There are just so many ways to get a wrong answer that fits a convenient narrative, and so few right answers.