Good Books: How to Lie with Statistics, Darrell Huff

I think I was in fifth grade when I discovered a copy of this book on my school library shelves. I was enthralled and How To Lie With Statistics became one of the most influential books I have ever read. It taught me to be a skeptic.

Darrell Huff had been an editor for several mainstream magazines before he returned to free lance writing. In 1954 he published How to Lie with Statistics which was the single biggest success he ever had in his writing life. I saw the book a year or two after it was first published and it was already into its 8th (as I recall) printing. I believe it had fifty or more reprintings — more than a half million copies sold — before it was re-packaged by Norton in 1993 into the volume that is still selling today. It has been translated into many languages — the Chinese version was published 2003 — and called the most widely-read book on statistics ever published, which might be true.

The first four chapters introduce concepts familiar to most statistics students — biased or inadequate sampling, distorted use of the average as opposed to the mean, and so on. Later, Huff writes about “correlation not meaning causation” and the post hoc fallacy. All of this is common street knowledge now, less so in the 1950s. So statistics-smart people sometimes pooh-pooh this book — unless they teach an introductory course and then they are hoping that their students will get as much out of a semester as Huff presents in less than 140 pages. The concepts are illustrated with examples taken from advertising or promotional material and by sharp caricatures by Irving Geis.

So we learn that an organization wishing to appear progressive may hype its average salary, but in a company where the pay rates range from $2000 to $45000 a year, the average doesn’t mean much. Note how Huff/Geis illustrates the concepts of average/median/mode:

huff1

That item brings another criticism that I have seen often about this book, that it is too old and people can no longer relate to the examples from the 1940s and ’50s used by Huff. After all, no one expects only $2000 a year wages any more. These critics, obviously, are very limited thinkers.

The chapters that excited me as a kid were the ones about using graphics to lie. For instance:

huff2

Same data, different charts.

Besides charts, Huff also pointed out how pictures were used to distort fact. One of the figures below is twice the other, but the illustration for the larger has been doubled in height, which means that it is four times larger than the smaller.

huff3

There was more of this, a lot of information in a short book, but the last chapter has a valuable set of tests to apply to claims and propositions:

1. Ask “Who Says So?”
2. How Does He Know?
3. What’s Missing?
4. Did Somebody Change the Subject?
(…[W]atch for a switch somewhere between the raw figures and the conclusion.”)
5. Does It Make Sense?

I probably apply these rules to stuff I read ten or twenty times a day. Huff taught me to be skeptical. But there’s a dark side to that.

In 1965, Huff appeared before a US Congressional Committe investigating the 1964 Surgeon-General’s Report linking smoking to cancer. Huff was being paid by the tobacco lobby and he pointed out to the Committee, in entertaining fashion, that correlation does not equal causation and that post hoc does not mean propter hoc. Afterwards Huff picked up some money to write a book, How to Lie about Smoking Statistics, from the tobacco companies but it was never published. Some suggest that this was because Huff sabotaged his own manuscript or that he recognized that the evidence was against him, but I doubt it. In the last analysis, Huff was a hack, a freelance writer trying to make a buck — a Mad Man, if that is meaningful for you.

So here is the thing: skepticism is fine, but you have to remember that Holocaust deniers, climate change skeptics, and tobacco apologists all use Huff’s methods to bolster lies. The point is not to be skeptical so much, as it is to assess the evidence as well as you can. You have to be skeptical about everything, even your own analysis.

 

About these ads

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s