Should there be some sort of exam where journalists have to display even a rudimentary grasp of statistics before they are allowed to go around journalling all over the place?Â Â I have just been reading this piece from the Telegraph about people taking sick days.It seems to be suggesting that there is some significance to the fact that people with particular names are more likely to take sickies.Â It gives the top ten list of men who are likely to call in sick:
- and so on…
Now I am too lazy to do any research or anything for something as trivial as this, but I would not be surprised to find that Andy, Steve, Paul, John, Dave, and Christopher are amongst the most common names in the workforce so the fact that they call in sick the most (from the sample used) is not really something especially noteworthy.
Wouldn’t you need to find the percentage of people in the workforce with a specific name and then find the percentage of sick calls made by people with that same name to be able to draw any sort of conclusion?Â Â Â You can’t just ignore the general population and draw conclusions.Â It would be like seeing how many dentists died last year that were men and how many that were women and then concluding that male dentists are more likely to die because there were more of them, without taking into account the gender imbalance in the dental profession.
This is just a piece of fluff, regurgitating a press release from Viva entertainment channel, based on a survey, as so much news now seems to be, and nothing to worry about ((Unless you are an Andrew who has not had a day off sick in three years who is being unfairly maligned here)) but it is a worry that the same journalists who produce this sort of thing might also be producing headlines about more significant matters, relying on skewed data from special interest groups without giving it, or even being equipped to give it, any sort of validation or basic sanity check.
And this is from one of the serious newspapers. Imagine how much worse it is in the Daily Mail!
For a near-perfect example of the same fallacy being applied to more serious matters I can recommend Don Paskini’s thorough fisking of an ipsos-mori report on local authorities.Â To summarise it enormously, a report found that although Labour only control about 10% of councils, various lists of the top-performing councils had over 20% of them (up to 35% for some lists) being Labour.Â The summary of the report, which is what a lazy and/or innumerate journalist would print as a headline, was that Tory-controlled councils are better because there are more of them in the top ten, or twenty or whatever.
That is a bit like saying that the British are better long-distance runners than Kenyans because there are more British runners in the London Marathon.
This week’s book recommendation: A Mathematician Reads the Newspaper by John Allen Paulos