"Corrosive of public trust in official statistics"
Within hours of this blog posting concerns about the Home Office's selective use of knife crime statistics, the government has been severely rebuked over the release by the official body which oversees such matters.
While he was at the Treasury, Gordon Brown's anxiety over lack of trust in government statistics led to the creation of the UK Statistics Authority - an independent watchdog to ensure that official figures aren't used for political spin.
Now that very body accuses Number 10 of publishing premature, irregular and selective statistics on knife crime in yesterday's press release issued through the Home Office.
The chair of the authority, Sir Michael Scholar, in a letter to the Permanent Secretary at Number Ten, paints a picture of the prime minister's officials and advisors deliberately breaching protocol. The release claimed that the number of teenagers admitted to hospital for stabbing injuries fell by 27% in areas covered by the government's Knives Action Programme.
But the statistics had not been checked and quality assured. Sir Michael goes on to reveal how, in his words, "the statisticians who produced them, together with the National Statistician, tried unsuccessfully to prevent their premature, irregular and selective release".
He describes what happened as "corrosive of public trust in official statistics and incompatible with the high standards which we are all seeking to establish".
The authourity (motto: Building Trust in Statistics) is charged with overseeing a new code of practice enshrined by act of parliament. The Home Office protocols under the code appear to have been breached on a number of counts.
Publication of statistics should be pre-announced; they should be orderly and planned to ensure no perception of interference and policy and operational staff need to be involved to ensure quality assurance.
It would appear that none of these conditions were met. Sir Michael's letter demands no repetition of this breach, but the damage may already be done - to the credibility of crime figures, to confidence in official statistics generally, and to trust in politicians.