GDP estimates becoming more accurate?
A little more on the ONS's defence of its numbers against criticism by the likes of Goldman Sachs, whose chief economist, Jim O'Neill I recently interviewed for the Today programme and other BBC outlets. I wrote up the Goldman Sachs analysis of the figures on the day the first estimate came out ("First draft of UK economic history").
The key claim in the ONS article [64KB PDF] is that, far from getting less reliable, the first estimate of GDP has been getting more accurate over time, despite the fact that it covers only 40% of the economic activity included in the more mature, Blue Book estimate published a year later.
Specifically, the ONS finds that between the first estimate and the one from two years later, the average revision since the mid-90s has been a mere +0.05 percentage points. Although there is a bias in the revisions - the numbers tend to be revised up more often than down - but it is much smaller than suggested by some critics.
As the authors of the article admit, this greater accuracy may simply be down to the fact that the economy was more predictable in this period. We didn't have a recession, and output and inflation were, in the words of Mervyn King, unusually nice.
In the wake of the controversy over the preliminary numbers, some economists have taken the ONS's side - notably Danny Gabay, at Fathom Consulting, who recently published a detailed rebuttal to the Goldman Sachs analysis making some of the same points as in the ONS article [120KB PDF].
Gabay's basic point against the Goldman Sachs economists was that they were including too many data revisions that were due to changes in the way the ONS collated the data - like the introduction of "chain-linking" in 2003. When new statistical methods are introduced, the ONS will usually go back and amend past figures, to make the data set consistent over time - in some cases making quite significant changes as far back as 1961. But it is a little odd to then criticise, on the basis of these revisions, the first estimate in, say 1971, for failing to anticipate later techniques.
So, the ONS has come out fighting. And they may have the last laugh. But whatever the past record, there's no getting around the increasingly wide discrepancy between the surveys, and the employment data, on the one hand, and the official data on the other.
Three European economists published their own study of the numbers yesterday, including a monetary economist at the European Central Bank.
There's no reason to suspect they have an axe to grind with the ONS. But they think the third quarter numbers are pretty odd too, especially now that the Eurozone is thought to have grown by 0.4% in the same period.
They argue that "the lag with respect to the Eurozone would be very much at odds with historical experience." (See chart below.) Using their own model, based on both survey data and GDP, they estimate that the UK grew by 0.15% in the third quarter. Clearly, the debate is not over yet.
Quarterly GDP growth, EU vs. UK