As the Protection of Freedoms Bill approaches its report stage, due on October 10, the Home Office has finally released three analyses that inform the bill’s proposals on the retention of DNA profiles.

It took a Freedom of Information request from Chris Hughes, who chairs the National DNA Database Ethics Group – first rejected, then in response to his request for an internal review of the decision, accepted – to get these reports into the public domain. Why the Home Office was so bashful is not clear, as the report of the internal review that overturned the initial rejection is not yet available. But Mr Hughes has won a significant victory.

Three documents are now available on the Home Office website. They are an analysis dated 15 July 2010, based on arrest-to-conviction data from the Police National Computer, a second analysis dated February 2011 which assesses the risk of subsequent conviction of people arrested but not charged, and finally a revised version of the February paper, dated September 2011, which additionally includes an analysis of those charged but against whom no further action is taken.

Robert Clifford of the Home Office’s Information Action Team explains in a letter to Mr Hughes that the last of these essentially supercedes the other two and is being provided to Parliament to assist its consideration of the Protection of Freedoms bill.  If one wanted to quibble, it might have been more use to MPs to have the earlier analyses when Mr Hughes requested them and not now, when the committee stage of the bill is complete. However, it’s out now.

To recap, the bill essentially seeks to bring the law in England & Wales into line with the 2008 ruling of the European Court of Human Rights in the case of S & Marper vs the United Kingdom, where the court ruled that the “blanket and indiscriminate” retention of DNA from unconvicted individuals violated Article 8 of the European Convention on Human Rights.

This blanket retention policy was introduced in the 2001 Criminal Justice and Police Act, and widened in the 2003 Criminal Justice Act. The 2003 act provided for the indefinite retention of a DNA sample and the profile derived from it from anybody arrested for a criminal offence and detained in a police station, whether or not they were subsequently charged. In response to the European Court ruling, the then Labour Government introduced the Crime and Security Act 2010, which provided for the retention for six years of fingerprints and DNA profiles from people arrested for but not subsequently convicted of an offence.  Repeated attempts to discover the evidence justifying the six-year retention period, not least by this website, were ignored, though this did not prevent ministers making repeated claims that destroying the profiles would results in rapists and murderers going free.

The bill now going through Parliament is much more liberal. Profiles taken from adults or juveniles (under 18s) arrested but not convicted of a minor crime cannot be retained, though they can be checked against crime-scene profiles to see if there is a match.

Those arrested for serious crimes but not convicted, be they adults or juveniles, will have their profiles retained from three years plus a possible two-year extension if approved by a court. All adults convicted of any crime will have their profiles retained indefinitely, a regime which will also apply to juveniles convicted of a serious crime. But juveniles convicted of a minor crime will have a retention period of five years plus the length of any custodial sentence, and for a second conviction, indefinite.

The key questions relate to the three-year retention periods for those arrested for serious crimes but not convicted, and the more liberal treatment of juveniles convicted of a minor crime. Do the analyses now released suggest that these are the right periods of time to balance the needs of the individual against the risk that destroying the profile will prevent the solving of future crimes?

The analyses tackle this question by drawing hazard curves that trace the annual risk of further offences in the years following the taking of the DNA sample for people charged with a serious crime but not found guilty of it. The point at which these curves intersect the line representing the risk of offences in the population as a whole should tell us when the risk of offending has fallen to the community average and DNA retention becomes hard to justify.

The figure below traces these curves, making some assumptions. The upper and lower bounds of the confidence interval are also plotted, with two separate curves that reflect assumptions about  cases still in the pipeline – one treating them as all ending in guilty verdicts, the other as not guilty.

As can be seen, the lower bound of the curve intersects the annual “sanction risk” line for the general population at three years – so the earliest the offending risk in this charged group could fall to that of the general population is just over three years after the initial charge. That would suggest the three-year retention period has some basis in evidence.

However, there are a lot of assumptions, not the least of which is the risk of offending in the general population. This is calculated as just under 4 per cent per annum – but the general population includes people with previous cautions and convictions. If these people are excluded, the risk of offending falls to around 2 per cent, and the curves do not intersect on the timescale available for analysis, possibly on no reasonable timescale. Similarly, if the upper bound of the curves is taken, offending risk in this group might never fall to the level of the general population, however it is defined.

The table below is interesting because it casts some light on the situation that existed before the S and Marper ruling. It compares the risks four years after the taking of the DNA sample for a range of scenarios and shows that for adults in all cases short of a proven offence, the risk of a conviction or other sanction has fallen to that of the general population (or less) four years later. This was the group, bear in mind, that stood to have their profiles stored indefinitely had S and Marper not intervened.  Under the 2010 bill, such people would have had their profiles stored for six years – unjustified on the basis of this evidence, since by four years their risk had fallen to the same level as the general population.

The table also shows that for juveniles, the four-year period does not lead to their risk falling to the same level as the baseline. In other words, arrest alone is more predictive of future behaviour for juveniles than for adults.

The second table compares risks for both groups following a proven offence. Again, the hazard ratios are higher for juveniles than for adults four years after the first offence, questioning whether the more lenient retention periods for juveniles in the current bill are justified. Juveniles who have two brushes with the criminal justice system resulting in a non-custodial sentence are more than three times as likely to offend again within four years as are those who have no proven offences.

Despite the limitations of the analysis, acknowledged by its authors, it does suggest that the retention periods allowed under the 2001 and 2003 acts were unduly long, as were those proposed in the 2010 bill. The present bill, which is broadly similar to the law in Scotland, gets the balance more nearly right. That makes it all the odder that the Home Office was so reluctant to let it see the light of day.

Perhaps it was anxious that, if you exclude previous offenders from the background population risk, then the hazard curves for all scenarios fail to intersect the background line. This implies that anybody having any contact with the criminal justice system will have a future risk of offending higher than the public at large, if that public is chosen so as to exclude past offenders.

Which is the best background line to select? That of all people, or all people excluding past offenders? This is a tricky question the Home Office analysts don’t attempt to answer.