Being a statistics junkie, a couple of people recommended to me the book How to live dangerously by Warwick Carins. Normally, I would read it, enjoy, and move on. But this book has prompted a mini-review (several years late, but who cares…), because it commits several statistical crimes.

One is that Cairns plays fast and loose with surveys. Surveys here, surveys there. No mention of how many people asked, by which method, or the sources. We can all cherry pick surveys to prove any point we like. A health warning is needed.

Second, Cairns is too casual to dismiss what we don’t know, and uses little data to back up the main thrust of the argument (which I broadly agree with), peppering his prose with “probably”s and “these days”. Example:

In 1970, eight out of ten elementary schoolchildren used to walk to school. In 2007, less than one out of ten did – and they were probably the ones who lived across the road, or whose dads were the school caretakers. Most children these days are driven to school in cars, even if they live just round the corner.

Really?

Thirdly, and far worse, it actually uses statistics to deceive, rather than prove a point. The worst offence is comparing the data on child abduction and murder with death from fires.

It is clear that the media make more of the former than the latter – a child killed in a fire is a tragedy that is maybe mentioned in the local news, while an abduction and murder will make national headlines quite often.

But Cairns breaks down the stats by pointing out that in any one year, only 100 or so US children are abducted by strangers, and of those 46 are killed. He then extrapolates that to say that the average child has a 0.00007 per cent chance of this fate, which equates to it taking 1.4m years for a stranger to murder your child if you left him or her unguarded on the street.

Obviously the idea of living for 1.4m years is nonsense, and a cunning way of pointing out our ridiculous fear of this event. But then he points out the relative danger of keeping a child indoors and the risk of fire, to show how foolish we are at stopping children going out.

Not citing which country (I assume the US again) he says “one child dies of [fire in the home] every ten days.”

So he sums up our fears thus (from p46):

So, they go out, and face the 1-in-1.4 million chance of being abducted and murdered. Or they stay in, where one child gets burned to death every ten days.

This is the worst statistical argument I have ever come across. Comparing a 1-in-1.4m chance (which is not the same as 1-in-1.4m years anyway) with one-in-10 days sounds like a logical slam dunk – why on earth would we care about the million chance when every 10 days a child dies in a fire? Except that these are far more similar stats than the way they are presented. Actually, using Cairns’ data, one child is abducted and then murdered every 8 days, compared to a death every 10 days in a fire. Or, put it another way, there are 46 abductions and murders every year in the US compared to roughly 37 fire deaths.

Either Cairns is being appallingly deceptive, or incredibly sloppy and can’t understand the stats himself. Either is hard to forgive in a book that tries to cut through the froth and present our fears and risk in a rational way.

Overall – for a book that cites statistics and tries to uncover our irrational fears, it is sloppy, prejudiced and patronising. It is poorly sourced, and although entertaining, lacks rigour. This is an important topic. It’s a shame that it is treated so badly.