In a recent piece in the New York Times titled The Wrong Way to Teach Math, author Andrew Hacker made some excellent points about statistical literacy in the U.S.. We especially love this passage where he wrote: "What’s needed is a facility for sensing symptoms of bias, questionable samples and dubious sources of data."
That said, we wonder if Hacker could have gone even further in some of his analysis, which would have helped further strengthen his points. For example, he wrote: "...women in Nebraska are averaging 2.2 children, while Vermont’s ratio is 1.6. Any theories?” In our opinion, this seems like the perfect opportunity to talk about what an average is—and isn’t. Why? Because the "theories" someone would have about this statistic depend (at least in part) on understanding that an average often hides variation in the data.
In another section, Hacker writes: "Fertility rates for white and black women in 1989 stood at 60.5 and 84.8 per thousand, a discernible difference. By 2014, they were 59.5 and 64.5, a much smaller gap. There’s a story here about how black women are reconfiguring their lives."
Is there a story here? Maybe, maybe not. Perhaps the change is in how people are identified (or self-identify) as black vs. white. A statistician might look at sampling error to determine how accurate the per thousand numbers are. And is the story about black women or black families or some other group? What other factors are involved in these changing fertility rates?
Don't get me wrong—we love seeing content like this in the New York Times, because it sparks a much-needed conversation about statistical literacy. We simply wish Hacker would have gone further. We appreciate his vision—we just wanted more discussion about the skills needed to achieve it, such as being able to distinguish correlation from causation, understand sampling, and identify cherry-picked data.