Saturday, June 13, 2015

How to Lie with Statistics

This is a book that two of my favorite people report that they give as their holiday gift of choice. Bill Gates recommends it on his summer reading list. After finishing How to Lie with Statistics, I can see why.

Just on the craft part of writing, the book is a great example for how to communicate important but dry concepts in an engaging, entertaining way.

"Permitting statistical treatment and the hypnotic presence of numbers and decimal points to befog casual relationships is little better than superstition. And it is often more seriously misleading. It is rather like the conviction among the people of the New Hebrides that body lice produce good health. Observation over the centuries taught them that people in good health usually had lice and sick people very often did not. The observation itself was accurate and sound, as observations made informally over the years surprisingly often are. Not so much can be said for the conclusion to which these primitive people came to from their evidence: Lice makes a man healthy. Everybody should have them"

Huff's book is written in a tongue in cheek way as if the reader wants to use stats for a tool to fool other people, of course plenty of people do do this. It can be hard to tell the difference. The current age has so much data, but far less quality analysis and consequently we have very little information relative to data we are awash in.

"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.'"

Anyone who has tried, say, Kona coffee and Kona coffee blend can relate to this. Statistical models are often wielded to distract from important points. As Sherlock Holmes said "there is nothing more deceptive than an obvious fact."

As to investing, we need look no further than dividend yield to see a great example of misleading stats. Once a yield gets too high, say triple the current S&P yield, you are generally in "sucker yield" territory. The company that says its going to payout 8% dividends today should not be taken at face value, and in fact it should be looked at as a negative, because once you look at the quality metrics you are likely to find the company will have a hard time delivering on that number.

The last chapter should be required reading for any citizen, frankly, as a self-defense mechanism in the so-called information age. Its a set of rules for how to talk back to a statistic, always ask:

  • Who says so?
  • How does he know?
  • Did somebody change the subject?
  • Does it make sense?
I might add to this list - how can you test this to see if the trend holds or not? The reason I see this book as required reading is that people with agendas are wont to throw out stats to prove their point, if you do not ask these questions you miss important points. 

I will give Mr. Huff the last word:

"Encephalitis cases reported in the central valley of California in 1952 were triple the figure for the worst previous year. Many alarmed residents shipped their children away. But when the reckoning was in, there had been no great increase in deaths from sleeping sickness. What had happened was that state and federal health people had come in in great numbers to tackle a long-time problem: as a result of their efforts a great many low-grade cases were recorded that in other years would have been overlooked, possibly not even recognized."

1 comment:

  1. Yes, this is an excellent book. Well worth a read. A similar one which has a more UK-focused set of examples is: The Tiger That Isn't by Andrew Dilnott and Michael Blastland. Another must read in my opinion, and not just if you're in the UK!

    I would really check it out!