I was thinking about the power of statistics in relation to one piece of data that you may have seen in the media recently. It goes as follows: 7% of U.S. adults believe that chocolate milk comes from brown cows. This is staggering. While reading the article that cited these data I wondered about the size of the study that this number had been extracted from, whether it skewed toward a specific group of respondents, and whether the respondents had simply misunderstood the question. I found it so ridiculous that I even wondered whether people knowingly chose the wrong answer simply because it was funny. Could it be that 7% of U.S. adults have such a sense of humor?
Cows and chocolate milk aside, statistics are everywhere. Love them or hate them, they are a fantastically powerful tool that allows us to uncover trends in past events as well as to look at the future and predict the likelihood that something may happen. We use them effectively for weather forecasts, emergency preparedness, disease predictions, genetics, medical studies, political campaigns, insurance, stock markets, and more. But of course, when used predictively, the degree of uncertainty can be huge—improbable events happen all the time—and so it is this particular application that is mostly to blame for earning statistics a bad reputation. During political campaigns, for example, statistics are used to predict the chances of a candidate winning the election. And they can—and often do—get it wrong. The most recent election cycle in the U.S. or the Brexit vote in the U.K. are good examples of that.
Like any science, statistics are not free from error or misuse. There can be error (oversimplification, failure to identify dependencies, and more) in the application of statistical models, leading to flawed conclusions, and taken to the extreme, direct manipulation of data to favor unwarranted conclusions. Sadly, both instances of misuse are common. Mark Twain popularized the expression, “There are three kinds of lies: lies, damned lies, and statistics.”
Of course, in science statistics are crucial to our day-to-day. I recommend reading a very interesting piece in the Guardian about one of Britain’s most eminent statisticians and the use of statistics in science. David Spiegelhalter, the president of the U.K.’s Royal Statistical Society, believes that “a sloppy attitude towards statistics has led to exaggerated and unjustified claims becoming commonplace in science.”
His basic premise is that “exaggerations” and questionable practices such as cherry-picking data threaten public trust in science. He draws parallels to the rise of fake news.
He adds that the pressure to publish scientific results is partly to blame for this, as it leads researchers to misuse statistics to make their findings appear more impactful. Another factor could be the greater lack of transparency due to the increased complexity and sophistication of statistical tools and models. Interestingly, Spiegelhalter is not concerned with Twain’s “lies, damned lies” (that is, false data or fabrications) because there are accepted methods to detect and correct them. His concern is instead with how scientific claims based on statistics are “filtered” from the source researcher to a university press office and then the media. As a result of this filtering, he concludes that the public “would be right to treat [the scientific claim or result] with suspicion.”
I fundamentally disagree with this last statement, but we may be arguing semantics. I don’t think that the word “suspicion” is appropriate in this context. The public should approach information—and not just scientific data—critically but not with suspicion. This is the equivalent to assuming innocence until proven guilty. Whether the people are prepared or in a position to do that is another question.
Regardless, the comparison with fake news is valid, and the potential decrease in trust in expertise is worrying and something we should be ready to combat.
Views expressed on this page are those of the author and not necessarily those of ACS.