I am planning to go watch Tim Harford, the Undercover Economist, speak at the RSA in a few weeks. If you haven’t come across Tim, amongst his other talents, he encourages readers and listeners to think a bit deeper about the statistics we come across on a day to day basis. Which got me thinking about some of the facts we’ve come across recently.
Anna, our resident statistics expert, sent an email round the office pointing us to the thought provoking Not-so-governmental Guide to the Unemployment Rate. Not just a clever way to present data but also some pretty scary statistics – if the US unemployment rate was calculated in the same way as it was 80 years ago, it would currently be 17.5%.
Mike, the head of our Research and Innovation Team, adds his own word of caution: “It’s a good reminder for all of us to be cautious about unemployment stats. There are multiple classifications – the ILO one, which is based on people fit, available and seeking work, and the claimants-based one which is all about those who receive job-seeker allowance benefits (JSA). In addition there is a large mass of ‘economically inactive’ – those out of jobs who ‘are not available for work’. Those guys make up significant proportion of inner-London boroughs but don’t feature in unemployment stats”.
And that gets me back to Tim Harford again. He hosts a show on Radio 4 called ‘More or Less’. In a recent episode he examined the concept of an average wage. A stat I commonly reel out to shock people is that the average wage in Tower Hamlets, my home borough, is £71,838. That’s the same borough where, in 2007, two thirds of children lived in income-deprived families, making it the most deprived borough in England for this indicator. What’s going on?
Well it’s all about how you report the statistic. The mean average wage in the UK is £31,323. That seems quite high, and it’s because wages are not normally distributed. Unlike height and weight, where there are an approximately equal number of short and tall people and a lot in the middle, for income there are a large number of people earning a small amount of money, and a small number of people earning a large amount of money. In fact, in London 75% of full time employees earn less than the mean wage. So in Tower Hamlets the City workers that live in affluent Wapping and Canary Wharf cause that mean average to soar.
To counteract this, in 2004 the Office of National Statistics switched instead to reporting the median. Across the UK, amongst full time workers, this is £25,123 – a £6,200 difference from that median figure. The median can be a tricky concept, but it essentially means that is if you lined up everyone working full time in the UK, the person in the middle of the line would be earning £25,123.
Why does all of this matter? Well firstly it means that we, as researchers, have to be very careful about how we report statistics. How many times do you hear someone state that something is “half the average wage” or “three times the average income”. Who knows to what they are referring? Not everyone uses, or in fact understands, the most recent ONS terminology. And secondly, because it genuinely influences the core arguments in a debate. Take the sector you work in as an example. Do you think people in the private sector or public sector earn more? If you look at the mean it’s the private sector workers, look at the median and it’s actually those in the public sector. Suddenly a debate on public versus private sector working habits and benefits has a whole new angle.
Read more here
on Apr 23rd, 2009 at 8:17 am
Interesting stuff. There are always a plethora of ways to conduct research and then again to report the findings. It can decrease trust if the reported stats are always the ones that paint the rosiest picture. But then how many of us are equipped with the knowledge and skills to truly understand the nuances of the evidence?
on Apr 28th, 2009 at 9:14 am
This is an excellent point! Not many people have knowledge of statistics extending beyond the popular notion of the mean (or the ‘average’). For that reason, the mean is often overused as a way to summarise the data, even if other options (mode, median, the very distribution of the data) offer a much more interesting and telling picture. Ruth has shown that if we use a wrong statistic, we may even completely miss the point! I always say: use the mean if the data is normally distributed (fits under the bell curve), and question the usage of the mean if it doesn’t!