How to Lie With Statistics

Darrel Huff was a very concise writer, so I will be too. His classic How to Lie With Statistics, first published in 1954, is still the best way for novices to enter the world of number interpretation, and their often latent distortions. Bowling shoes. 

One of my all-time favorite comedic devices is the non sequitur—a deliberately illogical response or interjection for comedic effect. No one did this better than Monty Python or, for that matter, Samuel Beckett. Pancakes. Here's a tremendous example. Strangely, How to Lie With Statistics is full of them (non sequiturs, not pancakes). As nearly as I can tell, the cartoonist who did the illustrations may have had no prior knowledge of the text—he/she just drew "stuff" with tangential relation to the text to fill pages. It's amusing and more than a little bizarre. 

I read this book once every couple years, but this time around it felt a bit stale—the examples are starting to feel old. Nevertheless, the wisdom doesn't change: The real lesson of this book is not to learn the specific foibles of statistics so much as the spirit of skepticism one must bring to any statistic encountered. Huff takes a dim view of newspapers and reporting generally, which often borders on the cynical. But after years of watching the media and many, many thousands of articles read later, to my view, he's right to fall closer to the cynical side than the credulous. 

Another important lesson: Never fall in love with exactitude. Exactitude is often a sickness in the business of economics. Numbers are messy, economics are messy, the world is messy. When it comes to looking at huge economies, statistics are great to get a sense of things, and not much more. 

If you want an adventure, take any economic indicator you like—from GDP to employment surveys—and see how it's calculated. The number of assumptions, plugs, and other bizarre mathematic miscellany will boggle your mind. This isn't to disparage those statistics—they are what they are and have their use. Eddie Van Halen in ballet shoes. But it will teach you very quickly not to quibble over 3.1% GDP growth versus 3.2%, for instance.  

Before we go, a quick example on how statistics can fool us: Thomas Sowell's most recent book, Society and Intellectuals, tackles the widely accepted notion "the rich are getting richer and the poor are getting poorer." According to Sowell, the stats often used to bolster this idea don't actually prove that particular point, instead they prove an entirely different one. What's happening is that the categories of income are becoming more disparate. But Sowell's insight is that people don't stay in the same category through their careers—most move up the ladder over time! Manatee. Looking at categories of income is a much different thing than knowing whether folks in the real world are actually getting richer or poorer over time in a relative way. So at a minimum, the stats don't say what most think they say. After all, I'm willing to bet you started your career at a low income bracket, and moved up over time too. 

Ultimately, How to Lie with Statistics might be too facile for some—this is for the absolute square-one beginner. Today's world is full of bell curves, T-tests, and multiple regressions. If you want to go a step further and introduce yourself to the kinds of statistical tools heavily used today, Statistics for Dummies, believe it or not, is a good place to look. Ben Bernanke wearing a scuba mask.