Home Office mum over mad March at the CRB
A secret criminal record of errors
The work of the Criminal Records Bureau (CRB) is undoubtedly important and often contentious.
So when on Monday this week it emerged that the error rate at the CRB had multiplied by a factor of 2.3 from a total of 680 in 2007/8 to 1,570 in 2008/9, it warranted further investigation.
Those error totals include false positives (people blocked from positions who should not have been) and false negatives (people allowed to take up positions they should not have been).
The Home Office issued an interesting statement alongside the figures. It said: "For the first 11 months of 2008-09 the CRB's accuracy rate exceeded the previous years. However, a dip in the quality in the final month brought down the overall figure."
Apparently intended to cast the 2008/9 figures in a positive light, with some rough calculations the statement in fact suggested a worryingly sudden collapse in the CRB's performance.
In 2007/8 the CRB ran 3.4 million background checks. In 2008 it ran 3.9 million, an increase by a factor of about 1.15. The 2.3 times increase in the number errors therefore couldn't be a simple function of more checks passing through CRB systems.
According to the Home Office statement, the proportion of errors from April 2008 through February 2009 was actually better than the overall rate for the previous year, which was 0.02 per cent.
So, at the end of those 11 months, assuming a constant number* of checks per month, there were fewer than 715 errors.
But by the end of March there had been 1,570, so there were at least 855 errors in one month. That compares to an average of fewer than 65 per month for the preceding 11 months.
For monthly error count to explode by order of magnitude, surely something must have gone drastically wrong.
We asked the Home Office press office for details. Perhaps there had been a technical glitch, new software, new staff or processes? Whatever the cause of the CRB's disastrous month, surely it was in the public interest to disclose it, and what has been done to ensure this "dip" is not repeated?
After we sent him our sums, a spokesman said: "Not one single factor can be attributed to this dip in the quality in the final month.
"Disclosure quality and accuracy is heavily reliant on the information supplied by the individual and is subject to fluctuations."
The implicit suggestion of his second sentence - that in one month more than 10 times more CRB check applications than normal contained dodgy identity data - is simply not credible.
But he also said not one single factor caused the "dip". Fair enough. So what were the factors? The Home Office refused to provide this information, while admitting its answers were "not helpful" and pointing out that it is possible to appeal against a CRB decision, if a complaint is lodged within 21 days.
Our attempts to extract an answer as to what went wrong at the CRB in March over the course of several telephone conversations on Monday and Tuesday were consistently and aggressively refused.
As a final effort to uncover why the CRB got it so wrong all of a sudden, we'll try a Freedom of Information Act request. In the meantime and possibly forever - assuming the Home Office really doesn't want anyone to know - the lesson of this story appears to be: Don't apply for a job in March. ®
*A reasonable assumption for these purposes. The Home Office told us there were 370,000 checks in March 2008, only slightly above the monthly mean for the year of 325,000.
Sponsored: Fast data protection ROI?