The figure for construction output within the latest “shock” GDP preliminary estimate figures showed construction output for the UK down 3.3% in the final quarter compared with the previous quarter.
The release also estimated that the snow effect on construction alone caused a drop in quarterly GDP of 0.1%.
Well did construction output fall by 3.3%?
The first thing to say is that the statisticians don’t actually know that construction was impacted by the snow to the degree they estimate, nor can they be certain that construction in the final month of the year actually continued the downward trend that was evident from earlier monthly data.
These are just pretty smart people making estimates on pretty incomplete data from a pretty volatile industry during a period of great economic volatility and extraordinary weather.
What is more the statisticians are, in the case of the construction data, working on a data series that is very new and where there is no historic data to provide a ready basis for monthly seasonal adjustments.
Further the statisticians have to follow process. So even if they suspect the process may need tweaking, they can’t just do it at a whim.
It is not my intention to open an epistemological debate here on how we know what we know about construction, but understanding at least broadly the provenance of data and the processes used to make sense of that data are very advisable before looking to draw too many conclusions. (Something I have to keep telling myself)
The rough rule I apply when engaging with numbers is that all data are potentially crap, but it is far better to have potentially crap data than to have none, especially if you can get a feel for how crap the data might be.
But what does this all mean for where construction is heading?
Well, the final quarter figure used for construction’s contribution to the GDP estimate was built mainly from the October and November data we already have for Great Britain (the GDP figure includes Northern Ireland estimates as well).
What we could surmise from the November data was that the industry had started a downward slide. This trend will have been taken into account by the statisticians when estimating what happened in December, because there is insufficient data for that month to draw but the scantiest of conclusions about the actual trend.
Then what the folk in Newport at the Office for National Statistics will have had to do was make an estimate on the probable impact of the snow. They will have had to look sensibly at whatever data they had for December and seek to gain insight into how much faster the fall was than the expected trend might suggest. This would provide some indication of the possible snow effect.
They will also have looked at the possible impact of the grotty weather in January on the data to see if that could tell us anything useful and perhaps looked at other severe periods of weather.
On this basis they will have, in effect, made an educated guess. That is to say a better guess than you or I would probably make – although that doesn’t mean they would necessarily be more accurate than every wild guess out there.
OK so far.
Well this would give them a figure for December and hence for the quarter as a whole. And that came out as 3.3% down on the previous quarter (when NI was included into the figures), with the snow effect in construction in December estimated to have reduce GDP by 0.1% in the quarter.
In crude terms, given that construction overall contributed to a 0.2% fall in GDP growth, that means half of the collapse in construction growth was due to snow in December. So we might be looking at, say, up to about 6% reduction in workload in the month down to snow, given things like December being a shorter working month than October or November.
So the 3.3% is, by dint of the limited data available and the circumstances, a very rough estimate. Additions to the pool of data will almost inevitably show that this estimate is wrong. The figures could be worse or better. Who knows?
So there you have it. Well not quite.
There is a matter of revisions, which occur as more data pours into the Newport offices.
Since the production of the third quarter construction output figures the method for estimating for missing data has changed after the statisticians undertook a regular review of past performance. This led initially to a hefty downward revision to the output figures.
Subsequently though, the revisions have tended to be positive. That is to say that when new data comes in it has tended to lead to an increase to the previously published numbers.
The upshot of this is that, if this tendency continues, we should expect to see the October and November construction output figures revised upward relative to earlier figures. This would suggest the scale of the estimated 3.3% drop might be reduced. Possibly good news, possibly not.
Either way I wouldn’t get too hung up on the estimate within the GDP release of what might have happened to construction in the final quarter of 2010.
Personally I would be much more worried about the overall figure showing a fall of -0.5% in GDP, because construction’s future in 2011 rests largely on the buoyancy of the broader economic recovery.