What works in criminal justice? Time to find out

What’s a guilty plea worth? One third off the custodial tariff (the current arrangement) or a 50 per cent reduction in sentence for an early plea?

It looks as if the Justice Secretary’s bid to switch to the second option, thereby saving a projected £130 million a year, has been overturned by fears that the Government appears weak on high-tariff crimes such as rape.

Principal aims of custody include incapacitation from committing offences during incarceration, and rehabilitation to discourage re-offending upon release. Both have the potential to save costs from reduced offending. The problem is we don’t know the balance between the two in terms of incarcerated versus rehabilitative inhibition, but we do know the cost both of incarceration and of re-convictions.

 When in doubt, randomize! A pragmatic randomised control trial (RCT) on criminal justice costs (both financial and in terms of re-offending) could answer the question.

If only the Ministry of Justice could institute a national RCT such that every eligible offender in England who had not entered a guilty plea was randomized once only between "third-off for guilty plea" versus "50 per cent reduction in custodial tariff" on the first occasion he or she was charged with an offence liable to end in a custodial sentence. Thereafter, for all subsequent offences, the offender would continue to be managed by the criminal justice system (CJS) in the mode thus-assigned.

Such an RCT could answer 7 questions by comparing between randomized groups:

  1. time from randomization to guilty plea by eligible offenders (ie those who, at randomization date, had not admitted guilt for an offence that was likely to command a custodial sentence)
  2. cost of criminal proceedings from date of randomization to initial sentence date
  3. initial custodial sentence before assigned-reduction
  4. custodial tariff after assigned-reduction
  5. cumulative prison-days from randomization date, R to, say R+1095 days, long enough to allow time for first sentence to be completed and later offences to come to light
  6. number of convictions for offences committed from randomization date, R to R+1095 days; and brought to justice before R+1460 days
  7. number of convictions for serious or violent offences {pre-defined} committed from randomization date, R to R+1095 days; and brought to justice before R+1460 days.

Rather than dogmatically carry on with one policy (one third off) or dogmatically switch to another (50% reduction), randomization plays one off against the other - but in an unbiased manner - and objectively measures by how much one set-up is better than the other at reducing costs and reducing re-offending.

The current debate, if the status quo gets the upper hand, risks missing out on all of the MOJ's projected saving. Randomization, on the other hand, both HALVES the projected loss AND objectively measures how the two policies actually compare in the real world of English courts rather than in prior assumptions.

If Ken Clarke’s 50 per cent option were implemented, it could be undermined by judges’ effectively increasing their sentences in order to make up for the promised reductions. Only by randomizing can one learn objectively how much of a hand this sort of  judicial discretion actually plays in the justice meted out.

Most shorter custodial sentences are multiples of 30 days so that a one-third reduction on 90 days is easily calculated as 60 days; one third reduction on 180 days brings the sentence down to 120 days; one third reduction on 270 days is 180 days.

Fifty percent reduction would mean that the judges now had to worry about multiples of 5, or adopt a different set of conventional sentences to avoid them. For example, 50 per cent reduction on 90 days is 45 days; 50 per cent reduction on 180 days is 90 days; but 50 per cent reduction on 270 days is 135 days . . .

Why are custodial sentences mainly multiples of 30 days - other than being easier to work out than multiples of 28 days? With mainly multiples of 30 days, a one third reduction keeps sums simple . . . But there must surely have been a better reason for the current set-up than that . . . ?

Whatever the reason was, the evidence-base to support it has not been prominently cited in the current debate. The above proposed RCT would, however, allow data to dominate opinion and would provide insight into judicial discretion.