How to Lie with Statistics is a 1954 nonfiction book written by journalist Darrell Huff to give the general reader ways to recognize deceptive statistics. Huff, supported by illustrations by Irving Geis, uses a combination of fabricated and actual examples of statistics that have errors, are misused, or can otherwise mislead an audience. Although Huff is a journalist with no background in the field, How to Lie with Statistics remains a best-selling text translated into multiple languages.

This study guide refers to the 2010 Norton eBook edition of the book.

Content Warning: The source material and this guide include discussions of suicide and systemic racism.

Summary

How to Lie with Statistics is divided into 10 chapters and an introduction. In each chapter, Huff examines a different element of statistics and explains how each can deceive the viewer. Throughout the book, Huff uses a mixture of humorous writing and statistical examples to get his points across to the reader. Each chapter also features cartoons and graphs by the illustrator, Irving Geis. In the Introduction, Huff builds the book’s central argument that the reader should view any statistic skeptically: It may use incomplete data, irrelevant connections, or misleading visuals. As a result, Huff notes that it is dangerous to see any statistic as unbiased.

Chapters 1 through 3 focus on problems with collecting and analyzing statistical data. If a study’s basis has issues, Huff argues, its results cannot be reliable. Chapter 1 deals with the different ways biases appear in samples. For example, sampling can overlook or misrepresent parts of a population, leading to results that don’t reflect reality. In Chapter 2, Huff mentions the term “average” when discussing statistics. There are three ways to measure an average in statistics: mean, median, and mode. Huff says the difference between averages can be minimal but can also vary enough that using the wrong one can mislead readers, especially if the type used is not stated. Chapter 3 focuses on the little problems in statistics that the reader may not notice. He discusses sample groups that are too small, which can create inaccurate assumptions. He also begins to touch on issues of misleading graphs, which he focuses on in later chapters. While Chapter 3 warns against missing information, Chapter 4 points out the problems in results with differences that are too small to be important. Huff makes special note of examples in which the amount of statistical error is either ignored or left out of the results.

In the next section of chapters, Huff shifts to the representation of results in a visual form. Chapter 5 discusses the different ways general graphs manipulate statistics, while Chapter 6 discusses bar charts and visual representations of information. While Huff says in Chapter 5 that graphs are less intimidating to the general reader than raw numbers, they can also be manipulated without falsification. Besides leaving out numbers, which he pointed out in Chapter 3, manipulation can also be achieved by removing parts of the graph or changing its proportions. Chapter 6 focuses on bar charts and the related “pictorial graph,” which uses a picture to represent the differences between amounts of a given thing. Huff emphasizes the problem of the width of the bar or the picture changing out of proportion to the height, which makes differences appear larger than they are.

The following two chapters deal with problems caused not by the data but by creating connections that don’t exist. Chapter 7 focuses on the “semiattached figure.” In it, Huff talks about the ways that unrelated ideas can connect. Information is left out so the leap from one idea to another appears more logical, or an issue at the data source may cause distortion in the resulting statistic. Chapter 8 regards the “post hoc” fallacy; Huff explains that correlation does not equal causation. Even if the statistical data lacks the other issues Huff discussed in the book, an apparent cause-and-effect relationship could be due to overlooked factors, rather than to the specific data being analyzed.

Huff uses Chapter 9 to discuss statistical manipulation in general, combining the information covered in the previous chapters. He states that such manipulation is only sometimes due to the statisticians. Instead, it comes from others who use the data, such as journalists or advertisers. Data published with the public in mind is presented in ways that exaggerate and sensationalize the results, instead of minimizing them. These results yield a better story or a desired result at the cost of accuracy. Huff emphasizes that because statistics are created by people with biases, they can never be totally objective.

Finally, Chapter 10 is a practical guide for the reader on how to look for statistical manipulations in their own lives. These can include bias, missing information, irrelevant conclusions, or failing to hold up to the common-sense test. While Huff concludes that readers can’t test every statistic they encounter, they should prepare to scrutinize false or exaggerated data to be better informed about reality.