Dianthus Medical Blog Archive

More dodgy statistics

Back in April, I wrote that perhaps I should start a "dodgy statistic of the month competition". OK, May and June have been and gone with no more nominations, so I'll make up for it by nominating 3 for July.

The first nomination goes to Vince Cable. This surprises me, as he is one of the few politicians we have who more often than not actually talks sense. Nonetheless, earlier this month he gave a speech in which he gave the statistic that "graduates, on average, could expect to earn £100,000 more in their life-time than non-graduates".

Now, I don't know whether that figure is realistic or not, although clearly it's reasonable to expect graduates to earn more than non-graduates. However, it does strike me that it must be very hard to measure just how much that difference is. To obtain such a figure reliably would have to  involve following up graduates for decades after graduation. Perhaps that has been done in the past, but the profile of university graduates has changed beyond all recognition in recent years, now that a much higher proportion of the population goes to university than in times gone by. Any such figures, if they exist, almost certainly are not generalisable to today's graduates.

But there is a further problem: reporting this statistic in the context of arguing in favour of a graduate tax (subsequently abandoned) only really makes sense if you believe the relationship is causal. It almost certainly isn't. Sure, graduates earn more money than non-graduates, but is that because of their degrees or is it just that they are generally higher-calibre individuals who would earn more money anyway? We are not told where the statistic came from, but I doubt it would represent a genuine causal effect of a university education.

My second nomination goes to the BBC and Ipsos Mori, who today have published the results of a poll claiming to show that the majority of Londoners support the 2012 Olympics being held here.

I would be surprised if that were true. Personally, I think it was a crazy idea to spend all those billions on hosting the Olympics, and pretty much everyone I know agrees with me. Of course, my circle of friends is certainly not representative of the general population, so I acknowledge that it's just possible that my experience is unusual, that the BBC poll could be right, and that Londoners really do support the games.

However, I note that no details of the methods are published with the results. How was the sample selected? What was the precise wording of the questions? Did any preliminary questions help to frame the questions in a way that could have biased the results? Without knowing those things, the results are meaningless. My guess is that there was something seriously dodgy about them. Why would they not publish the methods if they had nothing to hide?

But despite that stiff competition, the winner of July's "dodgy statistic of the month competition" goes to Experian. On the home page of their website, they tell us that "1 in 8 people are victims of identity theft". That sounds alarming, doesn't it? Perhaps you might even think that they want us to be alarmed, although I can't think why. The statistic links to a service that supposedly helps to protect you against identity theft for only £4.99 per month, but surely that couldn't have anything to do with it, could it?

There is no information on the Experian website to say where this figure comes from. I did email Experian's PR department to ask them, and at first it looked like they might send some data justifying their statistics, as I had a phone message from them saying they had tried to reply but been blocked by our spam filter (which they may well have been, as it can be rather aggressive). So I unblocked them, but my subsequent phone call and email asking them to resend the data have still gone unanswered. If they do send the data at some stage in the future I'll post an update.

So, where might we get statistics on identity theft? Well, CIFAS seems to be, as far as I can tell, a bona-fide organisation working in this area. According to their statistics,  there were 77,600 cases of identity fraud in 2008 (the most recent year currently available). Given that the adult population of the UK (and let's be charitable and assume that Experian really only meant adults) is about 48 million, that means that your risk of identity theft in a given year is about 1 in 600.

So, given the total lack of justification for the statistic of 1 in 8, and the almost a hundred-fold difference with my own (admittedly extremely crude) calculation, Experian can consider themselves a worthy winner.

← Plagiarism hurts An economic analysis of which journal to choose for your publication →