Lying with Statistics


Recently, I picked up a copy of How to Lie With Statistics, which is likely the best selling stats book ever written. First published in 1954, the book was written not by a statistician, but by a service journalist named Darrell Huff, who labored to show all the ways that statistics could be (un)intentionally abused by advertisers, politicians, journalists, and companies. The book is a superficial introduction to statistics that is mostly interesting as a historical document. The examples Huff uses to illustrate statistical hijinks are delightfully anachronistic and serve as a fascinating record of the origins of our modern data-driven culture. 

For a math book, How to Lie with Statistics has a remarkable legacy. It’s received glowing recommendationsfrom luminaries like Bill Gates and it’s spawned an entire subgenre of popular stats books like How Charts Lie or How to Lie with Maps. It’s not hard to understand the book’s remarkable influence. Its premise is that basic statistical literacy is a bulwark against the tyranny of data, which can be twisted in all sorts of ways to endow the most tenuous of claims with numerical authority. Huff realized that it’s hard to argue with numbers; he wrote this book to help readers cultivate a statistical bullshit detector. How to Lie with Statistics isn’t actually a manual on deception; it is, ostensibly, a safeguard against it. Which makes it all the more surprising that Huff ultimately became a willing participant in one of the most infamous examples of statistical deceit in the 20th century. 

Illustration from How to Lie With Statistics.

A decade after publishing How to Lie with Statistics, Huff was approached by Ed Jacobs, a New York attorney who served as outside counsel to the tobacco industry’s most notorious propaganda arm, the Council for Tobacco Research Special Projects program. The CTR was run by the Tobacco Institute and is perhaps best known for funnelling money from cigarette companies to scientists. These researchers were either tasked with demonstrating the benefits of smoking or disproving the link between cigarettes and negative health effects like lung cancer. And Huff was to become another one of their pawns. 

The history of the tobacco industry’s cynical use of the scientific method is well documented, but its fabrications have entrenched themselves in the popular imagination. For example, we’re all familiar with the “I need a smoke” mentality, where cigarettes are seen as a way to alleviate tension in a stressful situation. The idea that cigarettes are stress relievers is a tobacco industry fabrication that was legitimized by bankrolling Hans Seyle, an endocrinologist who is best known for his pioneering work on the science of stress. Seyle was an eminent scientist who was nominated for the Nobel prize 17 times before he was bought by the tobacco industry. He spent most of his late career working to establish a link between cigarettes and stress relief when none existed, but despite all the evidence to the contrary, the myth he cultivated persists. 

The mid-century tobacco industry had some of the deepest pockets around, and it proved time and time again that everyone has a price. In Huff’s case, his price was about $9,000, the equivalent of $70,000 today. This was his payment for writing How to Lie with Smoking Statistics, a follow-on to his best-selling book that aimed to show how statistics had been twisted to smear the tobacco industry. The book was rejected by a couple of publishers before it was picked up by Macmillan, which planned to publish more than 100,000 copies of the tobacco industry propaganda. In the interim, Huff testified to the Senate on the warped statistics used to undermine the tobacco industry and worked hand in hand with Jacobs to write a popular stats book that would suit the industry’s needs. 

Illustration from How to Lie with Statistics.

Huff’s publishing deal with Macmillan collapsed in 1968 after the tobacco industry was rocked by a scandal that revealed that several allegedly independent pro-smoking articles that had appeared in national magazines were sponsored by the tobacco lobby. To avoid a similar controversy with Huff’s book, the Tobacco Institute recommended placing a disclaimer at the beginning acknowledging its involvement. Tobacco industry lawyers were also concerned that if Huff had made any errors in his book and it was promoted by tobacco companies, the FTC could charge those companies with misleading advertising. Although Jacobs and the lawyers working with Huff had confidence in the book, the Tobacco Institute apparently did not have confidence in the lawyers. The project was dropped and the book was never published. 

The exact details surrounding the demise of How to Lie with Smoking Statistics are unclear. The vast collection of industry documents gathered through lawsuits in the past 50 years don’t make any reference to it after 1968. The little we do know about it almost entirely due to Alex Reinhart, a data scientist at Carnegie Mellon, who detailed the saga in a 2014 paper for Significance. Reinhart also hosts a draft of How to Lie with Smoking Statistics on his personal website, which is a master class in the ways statistics can be wielded to support just about any conclusion.

The whole thing is beautifully ironic. Here’s the man who wrote the book on how to avoid getting duped by misleading statistics—in fact, many of the examples in the book use the tobacco industry as an exampleand then used those same techniques to undermine the credibility of the tobacco industry’s critics. It’s as if Huff’s dance with Big Tobacco was meant to underscore the lessons of his best seller. But that would be giving him far too much credit. Instead, the entire project comes off as deeply cynical, an acknowledgment that even if you know how statistics can be used to fool you, the deck is stacked against you and you’ll be fooled anyway. In retrospect Huff’s best seller is much like the Surgeon General’s message on a pack of smokes. Sure, he willingly used statistics to defend a product that is responsible for 1 out of 5 American deaths, but you can’t say he didn’t warn you.