The Office for National Statistics has today published a paper defending its early estimates of Gross Domestic Product against criticisms that the figures are not worth the paper they are written on. Every sensible analyst acknowledges there is a trade-off between timeliness of data and its accuracy, but this is not the ONS’s argument why its first estimate is good.
Rather, it has plumped for the argument that its data cannot hope to predict future methodological changes in the series. Its argument boils down to – you cannot expect apples to be able to predict pears.
That is fine on its own terms, but irrelevant to most people outside the ONS. What they want is the best contemporary indicator of today’s economic conditions as we will know them to be far into the future.
Obviously, we cannot know what this will be. But if other contemporary indicators have been better at predicting this number than the initial ONS estimate in the past, it is quite a hard sell to get people to put these indicators aside and say that in future the initial ONS estimate will be best. There is no evidence for anyone to believe this to be so.
In the terminology above, if what we care about is pears, and oranges have been a better predictor of pears than apples, people are going to worry about the apples indicator today when it conflicts with oranges.