Recidivism

A little bit more on Tuesday's Polly Toynbee column.  As ever, actually taking the time to go back and read the source material has cast her gift for inaccurate précis into sharp relief.  Take this quote from her column:

Crime Concern delivers many of those designed to draw in the children regarded as at highest risk of offending. On a Rochdale estate, it achieves a 70% fall in calls to the police complaining about young people. It cost £350,000 - but researchers estimate it saves £665,000.

The £350,000 vs. £665,000 comparison comes from an evaluation of the Community Merit Awards (CMA) programme, commissioned by the Youth Justice Board (pdf link here).  The figures do not refer to Rochdale, but to all CMA programmes (see p.13 of the report referenced above), and the efforts in Rochdale include a lot more than just the CMA, as this quote from page 12 of the report illustrates:

The Langley YIP is engaged in a number of other schemes, so it is difficult to give all the credit to the CMA

In other words, the "It" in "It cost £350,000 - but researchers estimate it saves £665,000" does not refer to anything mentioned previously in the column.

As for Polly's "[o]n a Rochdale estate, it achieves a 70% fall in calls to the police complaining about young people", the report is a little more nuanced:

For example, the positive view of the young people provided by wardens in Langley, Rochdale was backed up by the estate’s housing manager. He stated that, at a recent borough-wide meeting to discuss vandalism and anti-social behaviour, his was the only estate where these issues were not seen as a problem. He also estimated that calls regarding young people being a nuisance had fallen by 70% during the CMA scheme.

In other words, an estimate of calls (not necessarily to the police) by the estate's housing manager.  For an actual figure, I commend to you the Middleton Guardian's story "‘It must be YIP’ as youth crime halves", which says:

Between April and June youth nuisance calls have fallen by 41 per cent, with criminal damage down by 60 per cent. Areas not covered by the project have not reflected the reduction in trouble.
[emphasis added]

In other words, an excellent result, and one to be praised, but not the one Polly claims.

-o0o-

She then goes on to say:

The youth inclusion programme identifies the 50 children most likely to become offenders locally, achieving a 65% reduction in arrest rates

The 65% reduction in arrest rates among participants needs to be compared with a 44% reduction in the control group of "top 50" children most likely to become offenders who weren't included in the program.  On page 100 of Evaluation of the Youth Inclusion Programme (pdf link here), for example, comes this quote:

There was also a decrease in the levels of offending for the top 50 who were not engaged by the projects. However, the change was substantially lower than for those engaged, at 44% compared to 65% for those that were engaged

Again, a 21% reduction is a laudable result, but not the one claimed.

-o0o-

Polly also writes:

Home Office research shows that every 15% increase in incarceration only prevents 1% of crime.

Of course it doesn't.  Take, for example, page 130 of the 2000 Halliday Report - 'Making Punishments Work: A Review of the Sentencing Framework for England & Wales' (pdfs downloadble from here) which says:

Home Office modelling suggests that the prison population needs to increase by around 15% to result in a short-term reduction of crime of just 1%, assuming that the extra prisoners would have committed 13 recorded offences per year, if at liberty.

Now, these figures only try to calculate the number of crimes that would have been committed by the new prisoners, or to quote Halliday:

These figures represent the avoidance of crimes, arising from just imprisoning a person. They do no [sic] estimate the effect on the propensity to commit crime after their period of imprisonment or the deterrent effect on others.
[emphasis added]

And how about the assumption that "extra prisoners would have committed 13 recorded offences per year, if at liberty"?  Well, Halliday gives this as a basis:

A survey of self reported offending among males received into prison under sentence in early 2000, suggests that they commit offences at around 140 per year in the period at liberty, before they were imprisoned.

[From the footnotes] It is estimated that the self reported offending figure of 140 is equivalent to about 13 recorded offences per year

[Back to the main text] There are substantial differences in the extent of drug related offending, ranging from 22 offences per person per year for those not taking any drugs to 257 for those who take drugs and admit to their drug taking being a problem.

Or, in other words, you can be more selective about whom you incarcerate.  To quote Halliday again:

A 1% reduction in recorded crime can be achieved by targeting particular groups, but with a smaller overall increase in the prison population. For example, by increasing by 16% the prison population of persons who admit to taking a drug and to their drug taking being a problem. This is equivalent to a 7% increase in the overall prison population.

Or a 15% increase could achieve about a 2% reduction.  So it is not true that "every 15% increase in incarceration only prevents 1% of crime." [emphasis added]

-o0o-

The poor "every 15% increase in incarceration only prevents 1% of crime" citation is used as evidence for this sentence:

Prison is swallowing up the cash that might stop crime - and it doesn't work.

Now, the prison population is about 76,266 (source here -- not particularly up-to-date, but good enough for a rough and ready calculation).  The research Halliday cited claimed that each prisoner would have committed or taken part in  about 140 offences a year if at liberty, thus preventing 10.6m offences from occurring.  The budget for the prison service is about £2.8bn (£2,591m of current and £209m of capital expenditures. Source: Home Office Departmental Report 2004–05, pdf link here, p.124-5), which means it costs about £264 per offence prevented.

Separately, the Home Office has estimated that the average cost of a crime to society is about £2,000 (on page 4 of Home Office Research Study 217: The economic and social costs of crime, pdf link here, they talk about "16 million crimes were estimated to be committed in this category each year, at a total cost of around £32 billion."  Note that this is for crimes against individuals and households only).  This estimate includes the costs of the Criminal Justice System, so we'll strip these out to avoid double-counting.  These amount to about 20% (see p.55 -- note this is for the entire CJS, not just the prison service, so we are being conservative in our estimates here), so say the average cost per crime is £1,600.

Now, this is a back of the envelope calculation, and certainly not rigorous to make any policy recommendations as a result (e.g. it is not clear that all 140 offences would not occur, as some of them were committed with other people), but it is not immediately apparent that "[p]rison is swallowing up the cash that might stop crime" -- it looks like it is spending cash that does stop crime -- £264 per crime prevented, which would have cost £1,600 had it occurred.  (Here is where I got the idea for the calculation).