Monday, January 28, 2013

Hidden Dangers in the Unemployment Figures

It was recently announced that unemployment had fallen by 37,000, this certainly led to a lot of boasting from Conservative types  but it may not be as much of a cause for celebration as they first think. The key to this likes in the way the statistics are put together.

The first point to consider statistics are not month on month statistics, but are rolling statistics that compare the most recent 3 month period with the previous 3 month period. As an example, the most recent statistics relate to unemployment from the months of September, October and November 2012 and compare those figures with the months of June,July and August 2012. So, over the period Sep-Nov 2012 unemployment was around 37,000 lower than the period Jun-Aug 2012*. This diagram should help to explain things visually.

A key point to consider is that this means that the figures overlap, the fall of 37,000 in November includes two of the same months as the fall of 82,000 in October. The large fall of 82,000 in October is not necessarily a sign that October was a good month, it may simply be that a sign that the month of July where a fall of only 7,000 was in fact an underestimate and that this point dropping out of the data caused October's figures to be a good deal higher.

It's important to note is that because the unemployment figures are moving averages that there is a delay between something happening in the economy and it feeding through to the figures. If unemployment actually rose in November then we may not see it immediately since the months of September and October would prop up the figures. What we would see initially would be the fall in unemployment dropping relative to the previous months. The 37,000 figure, being a good deal smaller than the previous month could be a sign of this.

A final thing to note is that that these numbers come with a pretty big health warning. First up, the 95% confidence interval for the unemployment figures is +/-91,000 meaning that that they are 95% confident in the most recent 3 months the change in unemployment was somewhere between +54,000 and -128,000. Second it's worth pointing out that although the figures are announced monthly they are only quarterly figures, unemployment has only fallen by 185,000 in the past year and the 95% confidence intervals here at -67,000 and -303,000.

The point I want to make from all this is that the government should be wary about what it boasts. The was the figures are presented means that the news can still look good even as the things start to look bad. And that while unemployment looks like it's moving in the right direction it's a long way back to anything resembling full employment and there are still plenty of pitfalls.

* I should also add that the figures are adjusted to take into account seasonal variations.

No comments: