*How to Lie With Statistics* by Darrell Huff—A Review

New York: Norton & Company, 1954. Reprint 1993; 142 pgs.

Three score and three years ago (in 1954) a seasoned journalist, Darrel Huff, who’d made a name for himself writing "How To" books, produced a cleverly titled and cutely illustrated book about statistics. This "how to" book soon became an endearing fixture on the shelves of critical thinking students for years to come. *How To Lie With Statistics*, by the 1970’s and 1980’s, was a standard text for college statisticians and may be the best selling statistics text of all time. Needless to say, this book has intrigued us here at The Best Schools because we are deeply interested in the best books especially as they may serve the best schools and the best education by bolstering the best classrooms be they public, private, or homeschool classrooms.

Statistics, unfortunately, gets a bad rap sometimes. Measuring things, quantifying reality, and organizing data, statistics are harmless little helpers, depending on how they are used. But the problem is, as Darrell Huff points out, that statistics can’t defend themselves when someone wants to use them in some biased, distorted, or even malicious way. In this sense, statistics are treated like a pliable mass of data that are neither true, nor false, but are subject entirely to the creative fancy of pollsters and reporters. Ideally, a wealth of statistics would work like concrete, forming factual foundations undergirding solid theories and smart practices. Good fair statistics work just like that. But, all too often, people use statistics more like wet cement that could be poured into most any space without ever quite drying. For these reasons, as Darrel Huff explains, we have good reason to weigh the statistics we hear with a heavy dose of reasonable doubt. It’s no wonder we have the oft-quoted, albeit cynical allegation, “there are three kinds of lies: lies, damned lies, and statistics.”

Darrell Huff's How to Lie with Statistics (illustrated by Irving Geis) is a breezy little study of that statistical “pliability.” Huff’s treatment of statistics is overall positive, affirming the dignity and value of good statistical analysis. But he arrives at this point by way of playful and even cynical examples illustrating mishandled stats. This brief text weighs in at a mere 142 pages—including cartoonish illustrations every third or fourth page. Originally published in 1954, the book is a classic, but admittedly it can feel quite dated when, for example, he uses examples based on income or population levels from 63 years ago. Nevertheless, the book is easy to read, easy to understand, practical, and quite charming as far as schoolbooks go.

The book is divided into an introduction and ten chapters. Each chapter deals with a particular trick in statistics, for example “built-in biases” that can distort polling data (Chapter 1); or selecting a “well-chosen average” according to one's agenda (chapter 2). Other chapters cover misrepresentative samples (chapter 3); misinterpreted margins of error (chapter 4); distorted graphs (chapter 5); misleading illustrations (chapter 6); semi-attached figures (i.e., the data is barely related to the conclusion; chapter 7); the post-hoc fallacy (chapter 8); and outright deception and miscalculations (chapter 9). The final section, chapter 10, is devoted to critically analyzing statistics or, what Huff calls, “talking back to a statistic.”

The real charm of this book is in its light-hearted tone, replete with jokes, humorous anecdotes, and comical illustrations, all couched in a “devil's advocate” framework. The author, Huff, could have titled the book more conservatively, “*Understanding and Correcting Statistical Errors*,” or even, “*How to Avoid Being Duped by Bad Stats*.” But instead, Huff presents his case from the vantage point of a modern sophist coaching people on how to best deceive people. A pragmatist in the spirit of predecessors like Nicollo Machiavelli, Saul Alinski, or Saul Goodman, Huff speaks without any preachy objections against manipulation or deception. This is essentially a manual on “How to Lie…” The angle-of-approach is not just (morbidly) interesting, but it also functions in cutting straight to the point without dawdling too deeply in meta-level commentary, or moralizing about his material. That said, it's still clear throughout the book, and especially in chapter 10 that the author ultimately intends to help students avoid being duped, helping readers cut straight through the agenda-driven distortions, push against manipulative data, and correct against fallacies. He might be playing the devil's advocate but he's only playing.

This book is not without its drawbacks however. It's brief and it's not very deep. Huff apparently intended Lie with Statistics to be accessible to both young and old without any background in statistics. And while that makes the book marketable to a wide audience, it also means the text doesn't cover much ground in the actual field of statistics. Huff's book is more practical than theoretical. Readers should not expect this book to serve, by itself, as an introduction into statistics. The book would make a great little addition into many classrooms, but it's just too limited to be a stand-alone statistics text.

As noted, Huff's book is also pretty old, and by old I mean, it's “real-world” examples and illustrations are all at least 63 years out of date. Anyone using this book in a conventional or homeschool setting will either need to spend some time explaining the more antiquated of Huff's illustrations, or find a way to update some of his illustrations. For example, on page 11, Huff treats a yearly income of $25,111 as the salary of a wealthy individual. Of course, by today's standard, that income borders on poverty. And he speaks on page 28 of a $3,500 yearly salary like it's a livable U.S. income when, by today's standards, it's not even close.

Most of Huff's illustrations, however, aren't damaged by time. For example, he points out how (in his day) the number of auto accidents was higher during clear daylight hours than during late-evening fog. But before we can jump to the conclusion that dark, foggy conditions are safer than daylight, he points out that fog can still be a dangerous road condition when we remember that the number of cars on the road is the more relevant third cause (pg. 77-79). That said, the age of the book isn't a serious problem. Antiquated figures are easily recast in contemporary terms. And most of the examples and illustrations are fairly timeless and universal.

Another subtle problem emerges in Huff's citations of Alfred Kinsey (pgs. 22-23; 45-46, and 95). Readers aren't likely to see the problem unless they know some backstory. Huff approvingly cites the famous research of Alfred Kinsey. Kinsey, if you didn't know, was a scientist whose now famous studies into human sexuality would play a major role, academically, in fueling the sexual revolution. Huff does not mention sex or anything from Kinsey's work that would be inappropriate for younger audiences. He does however defend Kinsey as having statistically supported his theories about “normal” relational practices and behaviors. Now Huff could not have known at the time that Kinsey's core studies into human sexuality have since been widely discredited for many of the same sampling errors that Huff criticizes in this book. For example, Kinsey drew a disproportionate amount of his survey data from prison populations, even though the subject matter was about romantic relations, and he presented his findings as if they represent the general population. Obviously, male prisoners will find their romantic options restricted if they aren't allowed near women. That factor taints the sample, so Kinsey wasn't justified in applying his results to behavior in the general population. Huff, however, has an excuse. He was a contemporary of Kinsey and most of the exposés refuting Kinsey's work wouldn't publish until many years later. Unfortunately, some sources today commit the same oversight, perhaps more concerned about Kinsey's influence than his accuracy. Naturally, most readers will be oblivious to the Kinsey problem. And in sum, this quibble is immaterial to the overall value of the book. For most readers, the references to “Kinsey” will amount to little more than a generic example of a scientist dealing in statistics.

Having covered the summary and critique of this book, a practical question remains: How can educators use this book? How to Lie with Statistics is a helpful, though sparse, introductory level supplement for use in classes on statistics, mathematics, critical thinking, and logic. It would also be valuable for marketing and advertising students who want to understand some of the (implicit) ethics involved in selling a product. Teachers and students could easily spend a brief time, maybe 5-10 minutes, discussing what a given chapter is about, interact with some examples, and then move on to a different textbook or lesson material. Its brevity means it can work in tandem with longer and more systematic books without “overloading” students. This book is not an academic textbook, or a workbook, but each chapter is discrete so that it could stand alone, and the subject matter easily lends itself to sample problems. The teacher will have to create these problems though; the book doesn't provide them. And because of it's brevity, students could also engage this text in independent study, in a research paper, in a book report, or a statistics project. It's the kind of book that doesn't need a lot of outside instruction. It “puts the cookies on the bottom shelf” so to speak.

Overall, Darrell Huff's writing and Irving Geis's illustrations combine for an endearing little book that's surprisingly interesting for a practical text. It lends itself to several classroom settings and uses. And it's well worth a read even beyond the classroom. Readers may have to translate a few of the outdated illustrations into modern terms, and there's a tertiary objection to one of the references. Nevertheless, this book has earned its reputation as a fun, informative primer on some of the common slippery little tricks in statistics.