³ÉÈË¿ìÊÖ

Officials at the continue to at the way in which Gordon Brown's closest advisors issued a "fact sheet" on knife crime through the ³ÉÈË¿ìÊÖ Office last month.

You may recall how the authority's chairman, Sir Michael Scholar, wrote a stinking letter to the permanent secretary at Number Ten in which he described the premature release of "unchecked statistics" as "corrosive of public trust" and "incompatible with the high standards which we are all seeking to establish".

Today, the UKSA's has broadened its criticism of the press notice.

Sir Michael was exercised by the way Downing Street ignored the personal pleas of the Chief Statistician to delay publication of hospital statistics on A&E admissions for stab wounds because they hadn't been properly checked.

But it is now clear that almost every bit of data in the release has got the statistics watchdog .

The Authority's job is to check that ministers abide by their statutory requirements under the . Basically, they must ensure that politicians don't play fast and loose with official numbers.

Their conclusion is that Number Ten and the ³ÉÈË¿ìÊÖ Office drove a coach and horses through the regulations.

Annex A of their release is entitled "List of practices from the Code with which the 11 December release of knife crime statistics appears to be inconsistent". There then follows a long and detailed list of how the press release broke the rules.

Annex B continues in the same vein: a devastating demolition job in forensic style. The authority's monitors accuse the government of releasing statistics that are "unclear", "selective or inappropriate" and lacking in context. Some of the conclusions drawn are "unsafe" and in one case the claims are "unsubstantiated".

The analysis chimes with my own impressions at the time. It is not that the figures themselves are necessarily wrong - rather that numbers were being mangled and manipulated to make the case that the government's knife crime campaign was having a rapid impact.

For example, the UKSA's report notes: "Youth violence is reported as being 30% lower in Halloween week than in the previous year. 'Halloween week' is not a recognised period for statistical comparisons. And no evidence is given about the reasons for this change - it could be because of the weather or other external factors."

The ³ÉÈË¿ìÊÖ Secretary Jacqui Smith for being "too quick off the mark with the publication of one number in relation to the progress that had been made with tackling knife crime".

However, the criticism is not that they were too enthusiastic in trumpeting success, but that they quite deliberately ignored the entreaties of statisticians in order to make a political point.

Conservative leader David Cameron has described the episode as "an appalling way to behave", but the Tories are not immune from using unsubstantiated numbers to make their case.

Just after Christmas, shadow ³ÉÈË¿ìÊÖ Office minister James Brokenshire claimed that under Labour, knife crime "has soared". His evidence was figures obtained under the Freedom of Information Act on crimes initially recorded by police as homicides caused by sharp instruments.

Four points:

• Homicides by sharp instruments include items such as glass bottles, so how can we be sure that we are talking about knives?
• The figure obtained from police forces amounted to 277 such homicides compared with 258 the previous year which, while tragic in itself, does not seem to amount to a new epidemic.
• The figures were not ready for publication because statisticians routinely wait until police and courts have confirmed that the incidents were indeed homicides. This process often sees a reduction in the headline figure.
• Overall, homicide has increased under Labour, as it did under the previous Conservative administration. However, there is some evidence that murder rates may have fallen from a peak in 2001 (unaffected by Shipman or the July 7th London bombings).

For an historical perspective, see my posts from last month: Map of the Week - Murder UK and The history of homicide.

• Bookmark with:
• |
• |
• |
• |
• - What's this?