Measuring GDP: Revisions are fact of statistical life

Factory assembly line

As the ONS publishes its latest estimates of UK Gross Domestic Product, the National Statistician, Sir Ian Diamond, explains the background to the recently announced revisions of economic growth in the pandemic period and the steps we’ve taken to underpin confidence in these key statistics.  

According to our latest estimate, GDP fell 0.5% in July. The UK is one of only a few countries in the world to produce a monthly reading of economic growth; GDP is published typically only every three months. We also update our methods and sources annually, whereas many countries do this much less often, meaning we can sometimes see the effects of shifts in the way the economy operates sooner than in other countries’ data.

Even in normal times, measuring a national economy worth around £2.5 trillion is not a simple task, but the impact of the pandemic provided some unprecedented challenges which have resulted in larger revisions to GDP.

When measuring the economy, you essentially have to do it in three different ways simultaneously – by adding up the output of businesses, business and household income and spending on investment and public services – and then reconcile the inevitable differences.

Huge volumes of data are required to do this, collected from households, businesses, government and the rest of the public sector. Some of it can be collected quite quickly, but some only with a long time lag – for example information on personal incomes from self-assessment tax returns or how the costs facing businesses are changing.

But people cannot wait years for the first estimates of what has been going on, so all statistical agencies produce preliminary estimates in the full knowledge that they will need to revise them later as more information becomes available. GDP revisions do not correct ‘errors’; they are a feature of the system, not a bug.

However, when we look at the quarterly revisions made to GDP, our quarterly estimates have been subject to some of the smallest revisions in the developed world.

But the COVID pandemic was a once-in-a-lifetime challenge for compilers and consumers of the National Accounts, because the size and composition of the economy both changed so substantially and so quickly. So larger revisions were almost inevitable.

One particular challenge is measuring the output of key public services like health and education. Many statistical agencies assume for simplicity that outputs simply track inputs. But the ONS tackles this head on – a tough job at the best of times, but conceptually and practically much more so when the NHS has to pivot from ‘business-as-usual’ to tackling a pandemic. Or, in education, when teaching is suddenly delivered remotely rather than in schools.

Connoisseurs of the annual ‘Blue Book’ revisions to the National Accounts were probably less surprised by the unusual scale and scope of the latest revisions than those who engage with them less frequently. But we need to ensure the widest possible confidence in – and understanding of – these data, including the way they evolve and the uncertainties around them.

We announced last week that we have invited the Office for Statistics Regulation to work with us to review our analysis of these revisions and how they are communicated, and to suggest anything we might improve. We want to underpin faith in our work, and this review will help us do that.

Looking ahead, we will continue to improve the quality and timeliness of the National Accounts – incorporating more information more quickly and exploiting big digital data more fully. But revisions will remain a fact of statistical life and it is our responsibility to continue to provide the best possible picture of the economy at any given moment, given the evidence available.

Picture of Professor Sir Ian Diamond

Sir Ian Diamond, National Statistician