How do you solve a problem like the ONS construction statistics?

How do you solve a problem like the ONS construction statistics?

You don’t need a map, satnav or signposts to drive a car from one place you know to another you don’t. But it helps. A guide is handy. It supports better choices. It saves time.

So, too, can good industry statistics. You don’t need them. But a good set of numbers can help to scale your market and provide hints at where it’s heading. Even fairly ropey stats and indicators help.

This brings me to the latest ONS release of the construction output data. These are bedrock figures for scaling the size of the industry. No other short-term indicator seeks to value the actual amount of construction work done each month in Britain.

But there are problems that need fixing and I believe the industry and its representatives should rally to press for a lasting solution to a number of problems that beset the series, those that are known and those that are latent.

These data are hugely important. Sadly their importance is greatly undervalued. Furthermore they have received growing criticism in recent years as a result of major revisions and findings that were seen as at odds with other industry indicators. They have lost the status of national statistics.

For example, a blunt reading of the latest data would say that in April the ONS decided that on an annualised constant price basis the industry was £3.7 billion bigger than it was thought a month earlier. Blimey.

In some ways this actually relieved pressure on the ONS, because the result brought the data more in line with what the industry and analysts expected, whatever the true figure might be. But this is not the first big revision. The construction output series is seen as troubled.

Revisions to construction output that in one go shift annual GDP estimates will raise eyebrows and potentially annoyance, not least when the bragging rights among some economists seem to rest on the slightest tweak in the economic numbers.

Inevitably this all undermines trust and value in the eyes of users and potential users. Sure, no statistical series is perfect and some of the criticisms of construction output data have been petty. But the scale of revisions, including the most recent, have been large. So concern is justified.

But ill-informed criticism based simply on the facts above would be inappropriate, misguided and dangerous. Many of the revisions have resulted from improvements to what is still a young data series that is bedding in.

Wagging fingers of blame and simply casting ONS and its construction data team as the villains here would be unhelpful, with unintended consequences potentially leading to a worse long-term outcome.

The industry should instead press for investment in the data series – a short-term boost to fund a deep analysis of its frailties and find long-term solutions. This I suspect would payback handsomely long term in better data and a better understanding of how to measure construction within the UK.

The alternative could easily be problems emerging periodically with quick fixes applied. That would steadily devalue the data and drain confidence. This could well lead to a retreat in its use and a consequent retreat in the coverage ONS offers in response, leaving the industry to rely increasingly on the partial light cast by trade indicators, useful though they are. Meanwhile, forecasts would have unsteady foundations on which to build a realistic view of future prospects. Our understanding of the industry would be diluted.

It’s worth reflecting on the challenge for ONS. It is relatively easy to set up a trade indicator to provide a reasonable gauge of sentiment and movement in the level of activity in a selected part of construction. I could, so it must be easy.

I couldn’t measure the size of construction output on a monthly basis. It’s a whole different order of task. Construction is an extraordinarily muddled mix of businesses, business interactions, outputs, odd seasonal effects and timeframes. Add to this the potential confusion of how land plays in the mix. It’s highly volatile and the muddled mix is in constant flux, morphing in odd ways. What’s more, simply defining construction is a task in itself.

So hats off to anyone who can measure on a monthly basis exactly how big this industry is.

This was the task ONS took on a few years ago and subsequently produced a new output series starting in 2010, which links to the previous series.

When the ONS took on the role I suspect the complexity of the challenge was underestimated, as would be the case with most people unaware of the idiosyncrasies of the construction industry. It has hit problems on the way and has had to deal with them.

The current set of revisions, which I stress are interim, relate to the output price indices. These provide the deflators used to adjust for inflation so the volume of work over time can be gauged.

Briefly, these inflation rates naturally vary between the various sectors and vary regionally. The previous approach to this has become increasingly unreliable over time. So the ONS needed a new approach. (You should read the background to all this on the ONS website.)

It has put together a temporary fix. This is far from perfect, as ONS recognises.

Those who have read the detail may have spotted that they are based on inflation rates in labour, materials and plant. But what of margins?

The reality is that changes in margins within such a muddled industry as construction are hard to measure. But you might fairly assume they shrink (even go negative) during a recession and expand when things pick up. Yes, the construction industry really does pay its clients for the privilege of working for them during a recession.

So there’s a fair chance that margins are expanding. If they are expanding faster than the other elements (labour, materials, plant) then the deflator will understate inflation in the system. This will lead to an inflated figure for construction activity in constant prices.

So basically the temporary fix has flaws.

The problem is not made easier by the lack of clarity over how the value of the margin appears within the various payments made throughout the life of a long contract and more importantly how this is express within the forms contractors fill in and send to ONS.

Is the overall margin proportionately allocated for each month? If not then how? Also, how do the amounts submitted by contractors monthly account for large and uncertain final settlements left to the end of a project?

Even if there is a neat way to solve this problem (if it is a problem), or to accommodate it, there are a host of other known unknowns. And probably plenty of unknown unknowns.

Consider seasonal adjustment. We pretty much know it is unbalanced across the year, seemingly leading to overstatement in the summer adjusted figures and understatement in the winter. Adjust for this and the growth rate in the latest two quarters may well be higher. But what procedures can you use to improve the seasonal adjustment which are statistically robust and within the protocols that determine good practice?

Furthermore we really don’t know on a month-by-month basis what is actually measured. What I mean is that some firms put in figures for work done that is chargeable, others state the sum invoiced, others point to orders received and some put in data for turnover or income paid in the month. What’s more this leads to a lag between when the work was done and when the work was reported.

This blend of different things measured with different lags will vary month to month.

And, then, we have the problem of what is actually being measured and what comes into the definition of construction. Not all construction is recorded. Construction done by firms that are not construction firms, retailers, manufacturers, charities, non-construction engineering firms would not be measured.

So, in theory, if a non-construction organisation takes construction work in house it disappears from measured construction output, popping up elsewhere in the national accounts.

These are just a few of the major challenges that need to be tackled. They alone represent a major investment in time.

So I think the industry needs to decide how much it values the construction data.

Personally I suspect it needs the data more than it recognises. In that case the industry should refrain from pointing the finger of blame and instead lobby hard for a root-and-branch review of the construction output data. This would require a well-resourced deep examination from the perspective of statistical measurement.

Among its tasks would be to review such basic things as exactly what the construction industry is, how factor inputs vary, how it does business and how this is changing, what administrative and other data it collects and in what form, how efficiently and effectively suitable data might be collected, what factors influence these data and how. The list could go on.

The key would be to build knowledge that would lead to a greater understanding of the industry and a more robust way of measuring it. From this should flow a long-term solutions.

The alternative is carping and criticising. But to what end?

Comments are closed.