Darrell Huff spent most of his life as a journalist and writer, yet he gained worldwide fame as an expert in statistics. This recognition came thanks to his most successful book, How to Lie with Statistics, written back in 1954 but still relevant to this day. At least, publishers continue to successfully reprint it time and again.
The entire tone of the book can be summarized with its very first epigraph, which quotes Benjamin Disraeli: “There are three kinds of lies: lies, damned lies, and statistics.”
No, the author does not try to portray statistics as a terrible or useless science. Instead, in a light and ironic manner, he explains what many people perceive as infernal mathematical gibberish. He also demonstrates how numbers can be used to manipulate opinions and the perception of information.
Formally, the book is divided into thematic chapters where the author delves deeper into specific aspects: what a random sample is, the types of averages, how perception can be manipulated using the “right” graphical representation of results. He discusses what is often left unsaid in conclusions, how errors can significantly alter the picture, or how cause and effect are substituted (or a connection is found where none exists). All of this is described using real-life examples, which, of course, make his points much clearer than if it were purely academic narration. But I say “formally” because, in reality, the real-life examples are not artificial tests, and often, in the same “statistical study,” multiple issues are uncovered. The author mentions these here and there, sometimes repeating himself. And as he himself says, statistics is “not just a science but also an art.” For this reason, the same data can often be presented with different nuances, depending on the result the statistician or their client wants to achieve. Here, Darrell Huff even compares statisticians to copywriters:
The statistician, whose findings are used in commercial promotion, is no more inclined to use a technique that will make the product look bad than a copywriter is to describe it as ‘flimsy and cheap’ when he can say ‘lightweight and economical.’
Moreover, when I say that the book has not lost its relevance, I am being slightly disingenuous—but only slightly. The tricks still exist and are even used today. However, they are no longer the “revelation of the century” (naturally, given that this book has been published for 70 years!). Many of the described techniques are now the ABCs for any self-respecting analyst. At the same time, truly professional analysts often consider it beneath their dignity to use such low-grade methods in their work (for example, some of these “tricks” are described as fraudulent by Vasiliy Sabirov in his book The Numbers Game).
The book is relatively short, with chapters resembling standalone articles (sometimes slightly repetitive, as I have mentioned before). Nevertheless, it is an easy read, contains a touch of humor, and is undoubtedly useful for anyone wanting to better understand “statistical conclusions” and critically evaluate them rather than blindly trust every “clever analyst.”
That said, I had expected a more structured approach and a deeper exploration of the subject, but the book turned out to be more of an entertaining pamphlet than a full-fledged popular science work. This mismatch between my expectations and reality likely resulted in my final average rating.
My rating: 3/5
[…] on his descriptions, it seems like all of statistics is pointless. You’d be better off reading How to Lie with Statistics. While that book was written decades ago, it’s far more engaging and provides much more practical […]