Lies, damn lies and statistics

So the general election campaign is hotting up. With less than a year to go, expect to be bombarded with statistics – particularly those on the NHS – which prove, and then, according to political opponents disprove whatever it is that is claimed on that day at that time.

So far we have seven key battlegrounds, according to one newspaper. Four seats to be won by UKIP, according to another. One opinion poll says we, the public, have a low opinion of the potential Prime Ministers post May 2015 – 25 per cent for David Cameron and 19 per cent for Ed Miliband. I could go on. The fact is that statistics, as we all know, can be used for or against any argument.

Unfortunately when, as a journalist, people start to use statistics to back up their claims they expect these “facts” to be swallowed whole and don’t like it when they are challenged. Nor do they like it when they are, in their opinion, taken out of context.

For example “Our customer satisfaction survey shows 90 per cent approval ratings.” Forget the fact that this sentence means absolutely nothing to anyone but the marketing team. Now 90 per cent approval ratings. So one in ten of your customers think your service/product is rubbish?

Shock, horror, much spluttering and accusations of typical media types always looking for trouble generally follow this question. But is the question a fair one? I would say yes it is. If you are going to start making claims based on numbers then make sure those figures are watertight and actually mean something. Work out the answer to the one in ten before you do the media interview. It might not come but have an answer none the less.

The other thing to remember about statistics is avoid getting hung up on them. For scientific/academic/technical reports it matters whether the figure is 71 per cent or around 70 per cent. But when it comes to writing a press release or sending out anything to the media please don’t get too attached to the number. We will round it up anyway and then put the detail later on in the report.

I would go even further and say don’t just round up the figures but say them out loud. Would you really say 75 per cent or would you say three-quarters? Is 46 per cent around a half? Yes I would say it is. I realise this heresy means I won’t be invited to the statisticians' Christmas party but as one in three of the turkey dinners is likely to be overcooked, I probably won’t know more than a quarter of the attendees and there is half the chance I will have to work the following day I can live without it.