How to Lie with Statistics

In How to Lie With Statistics, Darrell Huff makes an impressive successful attempt to teach the reader about all of the logical fallacies, misrepresentations and bias associated with reported statistics. The book is a short read and does not require that the reader know a great deal about math and experimental design before tackling the book. In about one hundred forty pages, Huff and Geis, the book’s illustrator, use a range of real life examples to show the reader problems with sampling, bias, averages, post hoc fallacies, and intentional and unintentional statistical manipulation. In the introduction Huff explicitly states that the book is meant to teach the reader how to make false claims using data gathered in a various number of ways. He makes an analogy to crooks and their methods of crime to be useful to anyone in need of self defense. In this case, the reader can learn to not fall victim to misinterpretation of statistical claims. I really enjoyed the first chapter as Huff opens by reporting the built-in-bias associated with income statistics of Yale graduates. It quickly becomes apparent that most statistics involving income are biased towards a specific group. This group usually consists of men and women whose addresses are known, and are willing to report an accurate figure to represent their annual income. As he continues to provide examples, the reader can begin to understand how to think scientifically about the accuracy of such claims. It occurred to me when reading this book that most people take reported statistics too seriously and without careful judgment. I learned that a lot of times statistics can be manipulated to be in favor of a certain perspective. I also learned that many times the data required to make an accurate statistic is often times unavailable to the public and in some cases nearly impossible to achieve. One example of this is the psychiatrist’s perspective on the mental health of the general public. It’s humorous to know that even a trained and educated medical doctor can make an assumption about the general public using a sample that is not representative of the overall population. Huff points out that when the psychiatrist thinks everyone is crazy, he is using a sample of patients whom are already seeking counsel. Although I was ultimately very satisfied with the book and Huff’s accurate portrayal of logical fallacies associated with statistics, I found that a few of the concepts were a bit redundant. Huff continuously talks about the difference between mean, median, and mode, and while he does a great job providing such examples, I felt that the first few examples were enough for the reader to get the point. I guess I must consider that the book was written many years ago and that people in the fifties were not as educated on how to be a skeptic. For me, some of these concepts were already learned in high school and college in such classes as, well, statistics! I felt that some of the points, such as averages, could have been made clear using his two or three best examples so as not to bore the reader. The examples that Huff uses were perfect in demonstrating issues such as post hoc fallacies. One example that I particularly liked was the one involving smokers and college grades. According to the text, one study showed that smokers tend to have lower grades in college because the data reported that those who smoke cigarettes generally have lower grades. Although the data clearly showed a correlation between the two variables, as I learned in statistics, correlation does not show causation. Huff very well points out that just because these two instances occur together, does not mean that one necessarily caused the other. In many cases such as this one, there could be hidden variables, truncated data or too small of a sample size. The illustrations in the book are similar to what one might find in a newspaper in the comic section. They are simple, funny and to the point. Geis uses Huff’s concepts to portray a highly entertaining depiction of people just acting

In Darrell Huff’s How to Lie with Statistics the main objective is to portray the idea that even though statistics pose to be legitimate, they are devious and misrepresentative of information. Although How to Lie with Statistics is only slightly over one hundred pages it efficiently offers information on how to recognize misrepresentations in statistics. How to Lie with Statistics illustrates different errors in statistics that are common in books, magazines, newspapers, and advertising. Huff describes statistics as “employed to sensationalize, inflate, confuse, and oversimplify” (Huff 10). How to Lie with Statistics portrays intentional and unintentional claims that can lead to inaccurate conclusions. Throughout the book Huff touches upon a variety of statistical graphs or charts, samples with built-in-bias, post hoc fallacies, the one-dimensional picture, and many more. Huff not only explains these different aspects of statistics in detail, but also reveals the importance of being able to differentiate between a legitimate fact and an imprecise statistic. How to Lie with Statistics is organized, straight to the point, and a great book to consider when wanting to learn about statistics. Even though How to Lie with Statistics was written well over fifty years ago the detailed examples offer clean cut explanations of what is trying to be communicated. One of the strongest aspects of this book is the detailed illustrations. When Huff talks about the one-dimensional picture he not only explains what a one-dimensional picture is, but has illustrations to offer further understanding. The artist for How to Lie with Statistics, Irving Geis, illustrates most of the examples with entertaining figures, graphs, and cartoons. Since a great deal of statistics is portrayed through illustrations, such as graphs, it is important that books about statistics include such illustrations. Imagine trying to understand the depiction of a graph through a verbal explanation. I believe it would be more difficult than simply looking at a graph. In an example pertaining to the one-dimensional picture Geis illustrates the comparison between the average weekly wage of construction workers in the United States and Rotundia. The bar graph illustrated to the left portrays the wage rate per week of United States workers at sixty dollars and the Rotundia bar graph showing a wage rate of thirty dollars. The bar graph is honest and straight to the point. However, the one-dimensional picture transforms this graph into a picture of a Rotundia money bag that is very small and a United States moneybag that is very large (twice as high and twice as wide). The one-dimensional picture takes this one step further by portraying the idea that Americans are far better workmen than Rotundians. The one-dimensional picture misrepresents an honest fact by involving a personal opinion. The illustrations in the book make the explanations far easier to understand. Instead of having to explain what the picture looks like, it is more effective to draw the situation. A great example of how Huff portrays information about statistics is when he talks about post hoc fallacies. When Huff explains certain examples they are very clear and straight to the point. In chapter eight Huff starts off by saying, “Somebody once went through a good deal of trouble to find out if cigarette smokers make lower college grades than non-smokers. It turned out they did…The road to good grades, it would appear, lies in giving up smoking; and to carry the conclusion one reasonable step further, smoking makes dull minds” (Huff 89). Although this experiment was done properly, it is statistically incorrect for one reason and one reason alone—the post hoc fallacy. This fallacy states that if B follows A, then A has caused B. This is a very common statistical error because even though something like smoking may cause students to get bad grades, it doesn’t mean that in order to get good grades all students who smoke have to quit. A lot of times this is an unintentional error in

Recommended by various sources, and rightfully so. It's not a long read, it's relatively entertaining, and it covers the fundamental ways in which statistics are misrepresented: manipulation of sample size, manipulation of averaging method (mean / median / mode), implied authority, etc. It would be especially good for people who haven't had a statistics class. However, it desperately needs an updated edition. The examples are based upon social and economic values of 1954 (back when US $25K a year was an impressive salary, comparable to about $200K today!), making them difficult for contemporary readers. Recommended mostly because I don't know of a current book that would be better.