Statistics are what ministers say they are

When does management information become official statistics? Francis Maude, Minister for the Cabinet Office, and Sir Michael Scholar, chair of the UK Statistics Authority, have failed to reconcile their differences over this.
The disagreement may seem minor, but its implications are not. They make it plain that, like Humpty Dumpty, a word means just what ministers say it means, neither more nor less. And behind that there is an almost infinite capacity for mischief.
The argument started when Mr Maude produced a head-count of the public sector payroll, despite the existence of validated statistics from the Office for National Statistics on the subject already being available. Straight Statistics’ coverage of that is here and here.  
The release of these figures was hedged about with numerous cautions and Mr Maude’s himself admitted that they were “a bit rough and ready”. It now appears, from a letter from Mr Maude to Sir Michael, that the National Statistician Jil Matheson had intervened and attempted to prevent publication of the figures until they had been quality-assured.
Mr Maude overruled her, on political grounds. To hold up publication “would have prevented us from delivering the shock treatment that ministers judge was needed” he says in a letter now available on the authority’s website. 
Mr Maude draws a clear distinction between “management information” such as headcount, and national statistics. He says he does not feel than any input from ONS “would add much value” and that the collection of this sort of operational data “should not in my view be treated as a statistical exercise”.
He goes on: “I think it would be completely wrong and damaging for central government organisations to feel that these data are being collected for statistical purposes. Ministers must be able to decide at short notice to publish such information, obviously surrounded by caveats to make clear that the information may be incomplete or inaccurate.”
He then adds that one reason for releasing the headcount data in June wad to make the point “rather starkly” that government before then simply didn’t have the data at all. In fact it largely did, though the headcount included some employees on contracts not included in the ONS figures that already existed.
Mr Maude’s letter effectively argues that ministers can decide what are statistics and what are not, and publish what it admits are inaccurate figures to make a political point. It is thus a direct challenge to the authority’s efforts to wrest the control of statistical information out of the hands of ministers, and restore public faith in official statistics.
Sir Michael’s reply is diplomatic. He wants to do nothing to prevent departments gathering better management information, and says that nothing in the Code of Practice for Official Statistics should stand in the way.
“But when data collected initially as management information are aggregated into statistics that are used in public in support of government policy, we may need, under the Statistics Act, to comment publicly if these statistics are not compiled and issued in accordance with the statutory Code of Practice” he goes on.
“In the case of the public sector employment statistics the Authority notes that a set of National Statistics, produced by the ONS, already exists. The new management information collected by the Cabinet Office goes wider but partly covers the same ground as the published statistics.”
What he wants, he says, is for the Cabinet Office to ensure that any future published aggregates from the management information are reconciled with the official statistics.
“This need not mean that the two sets should fully agree, only that any differences are clearly explained. This, too, is part of being transparent. In pressing Cabinet Office officials to quality-assure the data released a few weeks ago, the National Statistician was properly concerned about the need for such reconciliation.”
However, the authority has agreed to review the guidance issued by the National Statistician on the Use of Management or Administrative Information, to clarify where the differences lie between such information and statistics.
Good luck to them, as the line is blurred. Indeed, in future more and more use is likely to be made of administrative information as the basis for official statistics, to replace the Census and to back up weak statistical areas such as the migration figures.