A functioning democracy needs good quality economic and social statistics. Politicians may have an easier life in a post fact world but we, the citizens, shouldn't let them get away with it. An important part of holding our elected representatives to account is the existence of a well functioning organization, not under direct government control, tasked with the production of high quality statistical information. That is what ONS is supposed to be.
It's been common knowledge that things at ONS have not been going so well. The move out of London to Newport resulted in a significant loss of expertise that has been difficult to replace and the Bean report confirmed what everyone had long suspected. Yesterday I saw a stunning illustration of just how bad things have gotten.
For reasons that I won't go into I wanted to know the mean and standard deviation of full time earnings in the UK. Let's leave aside the fact that the standard deviation of an earnings distribution is not an especially sensible thing to be inquiring about. I doubted that I would easily be able to find the information, but as always I cheerfully Googled and was surprised to find that I was not the only one with this strange interest. In March 2014 a certain Mr Travers made a FOI request to ONS asking for the mean and standard deviation of annual salaries in the UK. Even better ONS provided him with the information and you can see their response on their website.
Great, I thought. That was easy. Then I started looking at the numbers. They come from the Annual Survey of Hours and Earnings (ASHE). In 2012-13 the mean earnings of a full-time employee in the UK was about £33,000 which looks plausible. In fact from other sources I know that the median in that year was around £27,000. Then I looked at the estimate of the standard deviation. Oh dear. ONS expect Mr Travers to believe that the standard deviation is £133. That implies a coefficient of variation of about 0.004 or a Gini of roughly 0.002 (the current Gini for income in the UK is about 0.34). If you take this seriously you reach the conclusion that UK is the most equal society in recorded history.
How can it be that even in its emaciated state our national statistical agency employs people to communicate with the public that have no feeling for the correct orders of magnitude of the numbers they deal with? OK, maybe it was delegated to the dimmest of the interns or maybe it was a Friday night rush job, but my guess is that nothing normally gets out of ONS before it is signed off by a more senior supervisor who you would expect to know better.
It's easy to make mistakes with numbers, I know because I make them all the time (but mostly spot them before I embarrass myself) but ONS should do better. They owe it to us and we have a right to expect it. In a world where populist politicians get off on denigrating experts making mistakes like this is another small step towards the new age of darkness.
Posts
2 comments:
I think I can shed some light on this, having gained some insight into the weird and wonderful world of government statistics.
Apart from the relocation to Newport, which is almost inevitably devastating for such an institution, it seems to me that statistical offices have a preference for technical skills over substantive knowledge about specific issues as the former are more easily transferrable between departments and projects. Sometimes very embarrasing things happen when very technically competent people try to estimate things they don't know the first thing about.
Secondly, I suspect that the ONS budget has been squeezed since the onset of the Great Recession, leading to work intensification and leaving people with less time to read up on things and (probably more importantly) less time to double check their work and get colleagues to review it.
Not that it excuses anything. I imagine that in terms of resources the ONS is a paradise in comparison to Statistics Iceland, which runs on a skeleton crew.
My own experience as a former frequent user was that post move things just got a damn sight harder to find, Eurostat hard. There was almost a wilfully sadistic aspect to the frequent redesigns.
Post a Comment