Review: How to Lie with Statistics

Surveys are all about getting numeric data, and then reporting on it. What points do you want to make with those numbers? I’ll assume that you are aiming to be honest, respecting your data and your audience.

One fun way of learning about the tricks that people play with data, or perhaps the mistakes that they make inadvertently, is to read my March pick for survey book of the month: Darrell Huff (1954) How to lie with statistics. It’s published in the US by W. W. Norton & Company with illustrations by Irving Geis; in the UK by Penguin with illustrations by Calman.

Huff aims to teach us “how to look a phoney statistic in the eye and face it down; and no less important, how to recognise sound and usable data in [the] wilderness of fraud”.

Some relief from harder books on statistics

All through March, I have been trying to write about statistics. March is a long month, and statistics meant it felt longer. I spent a lot of time trying to get my head around a variety of volumes from my statistical bookshelf – a bookshelf that I have, not because I’m particularly enthusiastic about the subject, but from my continuing quest to find a book on the topic that makes sense to me without being patronising or trivial. When I find one, you’ll hear it here first. Meanwhile, I needed some respite, and turned to Huff.

Writing about statistics can be difficult

I opened it and found this: “I have a great subject [statistics] to write upon, but feel keenly my literary incapacity to make it easily intelligible without sacrificing accuracy and thoroughness” – Sir Francis Galton.

Too right, Sir Francis.

A statistics book with a humorous touch

I bought my first copy of How to Lie with Statistics when I was 16, and it has nearly disintegrated. So I was delighted when my brother, who owns a second-hand bookshop, gave me a nice little hardback one. It says on the front ‘9th edition, “Wildly funny” – Evg. News’.

Well, I have to be honest and say if you want something that really is wildly funny, and you don’t object to some fruity language, then get 5 Very Good Reasons to Punch a Dolphin in the Mouth (And Other Useful Guides) by Matthew Inman (The Oatmeal).

Darrell Huff might be better described as ‘witty’ – or perhaps “wildly funny compared to other books on statistics”.

A book from the 1950s that is still relevant

You may have noticed that this book was first published in 1954. Can I hear you say “that’s from before I was born”?

You are right, it is a golden oldie. And to enjoy it, you’ll have to cut it some slack. For example, at that time, an average salary in a neighborhood might have been $15,000 (a high one to aspire to). Or perhaps $3,500 (a more typical figure for the time).

Today, those might be monthly figures rather than annual. But the fun part is the way that he shows that both amounts can be true of the same neighborhood, depending on whether you calculate the mean or choose the median as your average.

He does that sort of thing all through the book, showing how a little manipulation one way or another can put the same data in good or bad light – or simply a misleading one.

What will you find in Huff?

Here are the 10 chapters, with my notes on what’s in them:

1. The Sample with the Built-in Bias: on sampling and how a bad sample undermines your results.
2. The Well Chosen Average: wherein you find out about mean, median and mode – as in average salary, above.
3. The Little Figures That Are Not There: how to select your results to prove almost anything.
4. Much Ado About Practically Nothing: about margins of error.
5. The Gee-Whiz Graph: my favourite chapter, showing you how to critique a line graph in four short pages.
6. The One-Dimensional Picture: on the distortions you can achieve by using drawings for example, of a moneybag, rather than bars in your bar charts.
7. The Semi-attached Figure: looking for gaps between the figures and the conclusions that appear to be drawn from them.
8. Post Hoc Rides Again: statisticians always love to remind us that ” correlation is not causality”. Huff does the same in this chapter but fun examples.
9. How to Statiscalcute: a relatively long chapter where we learn how to choose the base for calculating a percentage so as to prove almost anything.
10. How to Talk Back to a Statistic: probably the most valuable chapter for its five questions. “Not all the statistical information that you may come upon can be tested with the sureness of chemical analysis or of what goes on in an assayer’s laboratory. But you can prod the stuff with five simple questions, and by finding the answers avoid learning a remarkable lot that isn’t so.”

A short book about the basics, such as means

If you’re familiar with mean, median, mode, standard error, the difference between a percentage and a percentage point and the importance of always including a scale in your graphs – then there won’t be all that much that is new in this book, although you may enjoy it anyway for the choice of examples and lightness of touch.

“It’s all a little like the tale of the roadside merchant who was asked to explain how he could sell rabbit sandwiches so cheap. “Well,” he said, “I have to put in some horse meat too. But I mix ’em fifty- fifty: one horse, one rabbit.”