THE DATA DETECTIVE

“Understanding causation is tough even with good statistics, but hopeless without them.” British economist and historian Harford crunches the numbers and finds us gullible—but corrigible.

Regarding the 1954 bestseller How To Lie With Statistics, Harford finds it overly cynical: “What does it say about statistics—and about us—that the most successful book on the subject is, from cover to cover, a warning about misinformation?” It’s a pregnant question, because of course people lie with numbers; indeed, the very author of that earlier book was pressed into service by the tobacco industry to prove that there was no link between smoking and cancer. Thankfully, notes Harford, British medical researchers worked out the smoking numbers for themselves, with the result that “doctors became the first identifiable social group in the UK to give up smoking in large numbers.” (This was years before the Surgeon General’s report in the U.S.) As the author rightly notes, there’s no reason to mistrust numbers, but we must interrogate them better, adopting the hopeful, forward vision of those researchers in the place of industry flunkies. True, there are all kinds of ways in which even the best prepared of us can overlook the truth that the numbers reveal. There’s a wishful-thinking bias called “motivated reasoning,” as well as the “powerful illusion” of “naive realism” by which we project our thinking onto the world and wonder why it doesn’t behave according to our rules. Harford also delves into “premature enumeration,” where we look at the numbers without quite understanding the questions we’re asking of them. Defending the process of polling and decrying the totalitarian habit of certain politicians to invent their own numbers, Harford lays out 10 commandments of examining statistics—e.g., ask tough questions about the data, be aware that something may be missing—along with a worthy “golden rule” : “Be curious.”

An entertainment for numerate readers and a user-friendly introduction to statistics for beginners.