How To Lie With Statistics

I recently concluded How To Lie With Statistics (Darrel Huff, 145 pages), and I enjoyed it quite a lot. For a book introduced in 1954, I think its contents remain very actual.

There are some interesting aspects of going through a book that was written 7 decades ago. One of them is the used language, which presents a few nuances not popular today anymore, but which I find interesting to observe. The second aspect is that several numerical examples used show just how much, and how fast, some things have changed. For example, the US population size was of 154 million people. While today the figure climbed to over 330 million. It is fascinating to reflect on how much infrastructure has got to come in place to support such a dramatic increase in population!

On the other hand, what does not seem to have changed much is all the biases and manipulations different actors make when presenting mathematical and statistical values to the public. Typically to make their numbers “look better”, for whatever purpose it may be.

The book goes over several cases in which this happens, in situations for example where the adoption of mean or median values are chosen only to highlight the results that better favors what the presenter wants to convey. Even cases in which figures and numbers in the same picture are out of proportion (e.g., doubling a value is illustrated by doubling an associated image; which actually implies multiplying the area by 4!). Such cases are depressingly common still today, even after all these decades.

I hope the general public has gotten more educated since the book first came out. Even with all inequalities in this world, the tendencies are towards much higher literacy over time. And this book is a reminder of how much critical assessment of the sources is so important not only for fact-checking, but for statistical and numerical data-checking as well.

No comments: