Thursday, May 24, 2012

Non-existent recidivism statistics wrongly reported

This article from the BBC News website caught my eye today, headlined "Reoffending rates reach record level".  It also says that 
Ministry of Justice officials say the figures show a "clear trend" of a rising re-offending rate.
The Independent makes some similar claims.  The phrase "re-offending rate" is simple enough: it's the proportion of people who, having been convicted of a crime, go on to commit another one within a year, or within five years, or whatever.  Right?  

Well not according to the BBC article, which states that "90% of those sentenced in England and Wales had offended before."  This is the proportion of people convicted who were repeat offenders, and is really quite a different sort of figure.  However throughout the article, it is referred to as the "reoffending rate".  The proportion of people convicted who are repeat offenders could increase for various reasons: here I list the 3 most obvious.
  1. The reoffending rate could actually have increased.
  2. The total number of previously repeat offenders could have increased (because more people are being released from prison than before, perhaps).
  3. The number of convicted first time offenders could have gone down.

If the reoffending rate has actually gone up, then that would certainly be a cause for concern; however explanation number 3 ought to give cause for optimism (at least if the decrease is caused by a drop in crime, rather than by worse policing).  The article helpfully links to these Ministry of Justice figures, so I did a little digging.  Sure enough, it contains no mention of reoffending rates whatsoever.

In addition, the 90% claim is wrong, even as a measure of the proportion of those convicted who are repeat offenders.  It is noticeable that this figure does not appear in the Independent's article.  The MoJ's Figure 1.7 (reproduced below) very clearly states that 75% of offences resulting in a conviction or caution (or the juvenile equivalents of a caution) were committed by a previous offender (208,000 out of 841,000 offences.)
  If we charitably interpret "those sentenced" as meaning offenders actually convicted in court, this increases to 85% (519,000 out of 609,000), although it is hardly surprising that those with previous convictions are less likely to escape with a caution.

So where does 90% come from?  Well, later on the BBC article reports that the criminal justice system dealt with 2 million offenders in 2011, and 208,000 were first time entrants.  The latter figure is taken from Figure 1.7, (it is in fact 208,000 offences, subtly different to offenders)  and should therefore be interpreted in relation to 841,000 total offences.  Why the disparity between 2 million and 841,000?  Well at the beginning of the section on Offending Histories, the MoJ document helpfully explains:

Information presented in this section differs from previous sections on convictions, cautions and sentencing, in that all information in this chapter is taken from the Police National Computer (PNC). The main difference is that the PNC does not include a range of less serious summary offences (such as TV licence evasion and a range of motoring offences) and so the figures reported are not comparable or match figures reported in the previous chapters. [My emphasis, their grammar].
A simple case, then of comparing apples with pears.

The MoJ document summary does at least show that this 25% figure represents a decrease on previous years:

24.8 per cent of offences in 2011 were committed by offenders with no previous
criminal offences, a decrease by almost 10 percentage points since 2006. ... This is driven by the decline since 2006 in the use of cautions in dealing with first time entrants.
This reduction in the proportion of "new entrants" to the criminal justice system is reported by the BBC as an increase in the proportion of previously convicted criminals.  Both interpretations are correct, but it's difficult to conclude anything from this alone, since it is the overall number of convictions that should interest us.  Figure 7.3 provides exactly this information over the last 10 years: the number of first offences convicted has falled since 2007, and is now just below the level of 2001; the number of further offences grew from 2001 to 2008, and has since fallen back very slightly.

Thus interpretation number 3 from the list above is the most accurate one.  Cause for some celebration, one would think.

A question of who is to blame for this misleading and (in the case of the '90%' figure) false, presentation of official statistics is not entirely clear.  The BBC article does quote "Ministry of Justice officials," which suggests that political types may be responsible for providing the incorrect interpretation to the journalists.  However, I couldn't find any official MoJ press release relating to this data, and the BBC must take at least some of the blame for not bothering to read the document which they so helpfully cite (though all researchers are guilty of this to some degree).

The Independent quotes the MoJ's Chief Statistician, Iain Bell, and an MoJ Spokesperson, and their statements are much more nuanced than the quotations given in the BBC article.  It should be emphasised that the official MoJ document is very thoroughly written, carefully interpreted, and avoids making any of the dubious comparisons found in the news article.  It is very sad that hard working official statisticians have to put up with this rubbish from their political bosses, and further compounded by poor quality journalism.


  1. can the 90% refer to table 7.5? This data shows that 10% of all offenders in 2011 receiving all disposals were first time offenders. Surely this means that 90% were repeat offenders?

    1. Hmm - it could do, I hadn't noticed this figure before. This table refers to indictable (i.e. more serious) offences, and this qualification isn't mentioned anywhere in the news coverage, as far as I can see. I also find it difficult to believe that person/people who wrote the article chose to report this rather obscure number.

      Well spotted, in any case!

  2. Thank you - It was these sorts of guestimates that the Governement is using to justify its reorganising of the probation service(s) constituent parts and plan to privatise the majority of what has been reorganised and are now struggling to function in the reorganised way - even before the sales.

    There are all sorts of issues - any statistical comment, would I am sure be eagerly read by readers of the On Probation Blog.