Following the release of numbers suggesting construction jobs increased by 47,000 over 2008 I asked for clarification from the statisticians. To my mind and in the view of anyone in the industry I have spoken to these figures were wildly wrong.

I have now received my reply from the Office of National Statistics on its view of the workforce jobs figures for construction. Read it below.

Can't say I am happy with the response, as it doesn't really provide me with anything I didn't already know.

I guess, though, what I wanted was either a hands-up answer of "these figures are rubbish" or an explanation that removed my concerns that these figures are rubbish. Well as you can see I got neither.

I will say now that I have huge sympathy with the statisticians trying to measure construction jobs. It is far from easy and the parsimonious approach by successive governments towards the statistics service does not help.

This is somewhat ironic, laughable even, given this Government's constant references to evidence-based policy.

I now wonder whether there is any point in this data series as far as construction is concerned.

Worse, I fear that publishing these figures probably does more harm than good.

Based on this "evidence" the appropriate policy should reflect the "fact" that there is no problem with jobs in construction.

Now I was always taught to keep open the possibility that I may be wrong even if it seems extremely unlikely. Healthy advice, but to my mind and probably in the minds of about 2 million construction folk and their families there were many fewer jobs in construction at the end of last year than at the beginning of the year.

Surely this should be reflected in the figures unless there is mass collective delusion.

While I get the bit about sampling variability, this is now at least two surveys in a row which appear to fly in the face of the real world. This being so there is a significantly smaller chance of the mismatch being due to sampling variability and a vastly increased chance that there is a systemic problem.

What is rather bothering me is that the trade bodies and unions that are supposed to be representing the industry and its workforce seem little bothered by numbers that suggest construction is a safe haven for workers.

Statistics may not be sexy but they are vital if we are to get policy right.

Reply from ONS:

On 17th December 2008 the construction workforce jobs figures (WFJ) were published showing an increase of 31,000 over the quarter to September 2008.

The latest figures, covering the December period, are showing a fall of 2,000 on the quarter for the same series.

The increase in construction jobs to September 2008 was unexpected. There were no method changes. Generally, the employee jobs component of Workforce Jobs (WFJ) is sourced from employer surveys, but for construction the Labour Force Survey (LFS), a household survey, is currently used. For self-employed jobs the Labour Force Survey is used for all industries, including construction.

The latest Construction Output release published by ONS on 6 March shows construction output for Q4 2008 was 7 per cent lower than in Q3. While there has been a quarterly fall in construction of new housing, there has been a small increase in public sector non-housing with housing and repair maintenance staying broadly flat.

Although construction job figures may not completely reflect the output contraction, the LFS based redundancy figures in the sector have grown markedly with the numbers made redundant in the sector reaching 48,000 for Oct-Dec 2008. This contrast is explained somewhat by increases in self-employment jobs offsetting falls in employee jobs, as is re-employment rates being comparatively high in construction, with around 30 per cent of employees who were made redundant in the past 3 months in construction estimated to be back in employment within the sector.

With all survey data, there is sampling variability associated with the estimates, and the quarterly change in construction jobs in September of 31,000 is not a significant change. The sampling variability on the quarterly change is estimated to be +/- 56,000 (that is, if the survey was run a 100 times, it is estimated that in 95 cases the estimated quarterly change would fall in the range of -25,000 to +87,000).

These are currently the best ONS estimates available but with all our survey data we are continually looking to improve coverage and methodology in the future.