The old saw that there are “lies, damned lies and statistics” is attributed to various figures, but was already considered proverbial in 1890 – perhaps as an adaptation of the old lawyers’ joke that there were three kinds of liars: “the liar simple, the damned liar and the expert witness”.
Even though we have been well warned for more than a century, people still use statistics dishonestly all the time – as when, for example, it emerged that the number of Covid-19 tests the British government claimed were being performed each day included tens of thousands of test kits that had merely been sent out in the post, as well as multiple tests performed on the same individuals.
There remain strong incentives for officials to lie in this way, however, because we assume – especially in an age that bovinely worships “data” – that numbers presented in the media are objective and disinterested: just neutral measures of the world. But as Sanne Blauw, a Dutch economist and journalist, insists in this sharp (and funny) little book, they aren’t. Every decision about what to measure and how to measure it, indeed, bakes in social and moral assumptions – which is why, as she notes, “machine learning” systems trained on masses of data routinely turn out to be racist or otherwise incompetent.
Blauw illustrates this theme through selected case studies, beginning with a triumphant use of statistics at the time of the field’s infancy, when the mathematically trained nurse Florence Nightingale sent painstakingly compiled charts of the mortality of British soldiers in Crimea to the authorities. She “was one of the first people in the world to use graphs to effect change”, Blauw remarks, and it was perhaps the very success of such early victories that attracted the subsequent wave of unscrupulous quantifiers.
The author goes on to debunk what her chapter title perfectly characterises (in Suzanne Heukensfeldt Jansen’s translation) as “The Dumb Discussion about IQ and Skin Colour”, addresses how problems of sampling populations can lead to odd conclusions (as, she argues, in the famous Kinsey reports into American sexual behaviour), and offers a punchy, sardonic history of the deliberate misuse of correlations, graphs and other techniques by the tobacco lobby and global-warming deniers.
In such a brief survey, some of her topics inevitably beg for deeper treatment: the discussion of p-hacking (a way of massaging numbers to make something look significant) in scientific research, for example, is explained at more satisfying length in George Zaidan’s excellent recent book, Ingredients. But the digestibility of Blauw’s offering is also a public virtue in itself, if it encourages more people to read it and immunise themselves against the virality of numerical disinformation.