Library:How to Lie with Statistics
| How to Lie with Statistics | |
|---|---|
| Author | Darrell Huff |
| First published | 1954 |
| ISBN | 978-0-393-31072-6 |
| Role | Analysis |
| Domains | Science & Environment |
| Tactics | Statistical & Evidentiary Distortion |
| Specimen: Huff later took payments from the tobacco industry and drafted a companion book, How to Lie with Smoking Statistics, to cast doubt on the link between smoking and cancer. | |
How to Lie with Statistics is a 1954 book by Darrell Huff.
Summary
Darrell Huff, a freelance writer and magazine editor rather than a statistician, wrote this short, witty book in 1954, with cartoon illustrations by Irving Geis. It became one of the best-selling statistics books ever published and for decades was assigned in introductory courses as the standard warning about how numbers mislead. It earns a place here twice over: as a classic analysis of statistical deception, and as a specimen of it.
The catalogue of tricks
Huff organizes the book as a tour of techniques. The sample with a built-in bias shows how surveys reach only the people likely to give a particular answer. The well-chosen average explains how mean, median and mode can describe the same data in very different ways, letting a speaker pick whichever flatters the point. Other chapters cover the little figures that are not there, meaning missing sample sizes and margins of error; differences too small to matter presented as significant; the gee-whiz graph, whose truncated axis turns a modest change into a dramatic one; the one-dimensional picture, in which a pictogram scaled in two dimensions exaggerates a one-dimensional difference; and the semi-attached figure, which proves one thing and implies another.
Correlation and the reader's defence
A chapter titled after the "post hoc" fallacy explains why correlation does not establish causation. The book closes with five questions for talking back to a statistic: who says so, how do they know, what is missing, did somebody change the subject, and does it make sense. Those questions remain a usable checklist.
Why it is a specimen
The book's own history became a lesson in the deception it describes. Researchers later documented that in the 1960s Huff was paid by the tobacco industry. He testified before Congress in 1965 to cast doubt on the statistical link between smoking and disease, and worked on a manuscript, never published, applying his techniques to undermine that evidence. The author of the classic warning against misleading statistics lent his skills to one of the most consequential statistical deceptions of the century.
Relevance to Bullshitopedia
The book remains a clear introduction to the Statistical & Evidentiary Distortion tactic. Its author's later work is a reminder that credentials and reputation are not evidence, and that skeptical techniques can be turned to manufacture doubt as easily as to expose it. See Merchants of Doubt and Doubt Is Their Product for the wider campaign.