Prison works: but how well?

In 1993, Conservative Home Secretary Michael Howard famously proclaimed “Prison works!” and built more prisons.
Today, the Justice Secretary in the Conservative-Lib Dem coalition, Kenneth Clarke, is due to unveil plans for the reform of sentencing that will involve imprisoning fewer people and putting far greater emphasis on rehabilitation. He intends to reward organisations that can show, not just that they have good intentions, but that they can actually improve rehabilitation rates as measured by a reduction in reconvictions.
Mr Clarke admitted on Monday on BBC’s Newsnight that he had “no idea” if his scheme will have the effects he seeks. How possible is it to make policy without evidence? And why do we still lack the evidence we need on the effectiveness of different sentencing policies?
In a two articles this week I intend to examine prisons’ past performance; at how “good intentions” are formally tested in jurisdictions other than criminal justice; and finally how Mr Clarke’s idea of “payment by results” is actually being tested in a pilot programme at Peterborough Prison. There are serious questions to be asked about this programme, and whether it is suitably designed and sufficiently robust to produce any worthwhile answers.
Prisons protect public safety and inhibit offending. There are few escapes, even when prisoners are being transported to and fro. For example, Scotland’s privatized prisoner-escort was responsible for only 67 escapes in nearly 900,000 escorts in the financial years 2005/06 to 2009/10 (and down from 22 per 100,000 in their start-up year). For each escapee, Reliance is fined around £25,000.
Prisons provide for the healthcare of inmates: universal offering of hepatitis B immunization; reduction in prison suicides by younger inmates; harm reduction; confidential testing for, and treatment of, HIV and hepatitis C; together with methadone substitution for heroin-abusers.
But despite these undoubted successes, problems remain. Some crimes are committed by prisoners during remand or under sentence – incidents of concerted disorder, the settling of drug scores sufficiently to frighten debtor-inmates into asking for protection, bullying or inciting others to crime. Even the most serious offence, murder, can occur.
The transition back to the community after release is problematic, especially for those with a history of heroin injection. The high risk of overdose death in the first and second fortnights after release is now internationally recognised. In Scotland, drug-deaths in the 4 weeks after release account for one in eight of all heroin-related deaths.
Yet, the biggest question for prisons, judges and criminologists is how well imprisoning criminals,  especially for short periods, actually works in reducing recidivism when compared to the cheaper alternative of community orders.
The obvious approach would be to carry out a randomised control trial, in which convicted criminals would be randomised to either imprisonment or to tagging, drug-rehabilitative, or other community sentences. Just as doctors invite eligible patients to agree to be randomized in formal experiments in which – in effect – the toss of a coin determines whether the patient receives an experimental or the conventional treatment, judges could do likewise.
In practice, doctors may register their patient’s eligibility and consent by telephoning the trial centre, which then informs the doctor of the patient’s random assignment. In the same way, judges – having written out their conventional sentence and experimental disposal for an eligible offender – could register the offender’s eligibility and consent in a telephone call to a criminal-justice-trial-centre, for it to inform the judge whether to pronounce in court the experimental or conventional sentence.
In health, not only do we subject new pharmaceuticals or screening policies to formal experimentation (using random allocation) as their basis for licensing or implementation but we also require cost-effectiveness to be demonstrated by NICE, the National Institute for health and Clinical Excellence.
By contrast, criminologists have typically:

  • lacked randomization as a basis for like-with-like comparison,
  • started the clock for time-at-risk-of-recidivism differently for those given custody (for whom start-time is ‘after release’) versus a community order (for which start-time is sentence-date!),
  • conducted studies that were under-powered to detect modest beneficial effect-sizes, or inadequate to elucidate harms (such as: drug-related deaths or serious further offences),
  • failed to establish offenders’ survival status, such as by record-linkage to the deaths register, so that success (rehabilitation) versus failure (suicide or overdose death) are indistinguishable,
  • too often relied on interviews to follow-up criminal justice clients – with response-rates well below 60 per cent, and
  • worked out the cost of disposals, such as drug rehabilitation orders, but lacked any decent estimate of their effectiveness.

In UK terms, Lawrence Sherman’s RCTs of restorative justice versus conventional sentencing were a breakthrough but – even so – they were run alongside some non-randomized trials that relied, in part, on matching offenders selected for restorative justice to controls by some measure of propensity-to-reoffend. How very neat for science, but not at all surprising, that randomization gave better-matched comparison groups than propensity did!
Professor Sherman, Wolfson Professor of Criminology at Cambridge, next plans to test his theory that criminal sanctions should be graduated by some measure of how likely an offender is to create future mayhem (as summarised through a crime harm index, CHI). He wants to randomize N,000 offenders between “prosecution as usual” and “manage by CHI-calibrated sanctions”.
The intention is that some offenders would be prosecuted according to their CHI (low, middle, or top tier), with sanctions defined by a set of rules. Follow-up for two to four years would be essential, with potential outcomes including future offences, deaths, hospitalisations, opiate substitution prescriptions, offender-employments,  and social and criminal justice costs all measured – and then compared with the outcomes from “prosecution as usual”.  Random selection would determine which offenders were chosen for CHI-linked sanctions and which for prosecution as usual.  
This is a plan based on well-marshalled data and a methodology proven in other fields. But is the Ministry of Justice’s payment-by-results plan on similarly firm foundations? Tomorrow’s article will examine that in some detail.