J Exp Criminol (2008) 4:403–423 DOI 10.1007/s11292-008-9063-3
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending Kevin Marsh & Chris Fox
Published online: 28 October 2008 # Springer Science + Business Media B.V. 2008
Abstract With prisons in the UK reaching full capacity, and with similar trends in other European countries and the USA, there is much political debate about the efficacy of prison and community sentences. This paper aims to inform this debate by testing the hypothesis that prisons are an effective and efficient way of reducing re-offending. A rapid review of effectiveness studies was performed to determine the relative impact of prison and community sentences on re-offending. An economic analysis was undertaken to transform the estimates of effect into estimates of the economic efficiency of alternative sentencing options in the context of the UK. When compared with standard prison sentences, a number of community-based interventions and enhancements of standard prison sentences were found to save money, both for the public sector and for society more broadly. Diverting adult offenders from standard prison sentences to alternative interventions saves the UK public sector between £19,000 and £88,000 per offender. When victim costs are considered, diverting offenders from standard prison sentences saves UK society between £17,500 and £203,000 per offender. It was concluded that standard prison sentences are not an economically efficient means for reducing re-offending. Keywords Community sentences . Cost benefit analysis . Custody . Economic efficiency . Effectiveness . Prison
Introduction Prison populations are on the rise in a number of western countries. For example, in the Netherlands, the numbers of prisoners rose from 10,249 in 1995 to 20,075 in K. Marsh (*) The Matrix Knowledge Group, Epworth House, 25 City Road, London EC1Y 1AA, UK e-mail:
[email protected] C. Fox Manchester Metropolitan University, Manchester, UK
NO9063; No of Pages
404
K. Marsh, C. Fox
2004 and in Spain from 40,157 in 1995 to 59,224 in 2004 (International Centre for Prison Studies 2008). In the USA the prison population rose from 1,585,586 in 1995 (a rate of 600 per 100,000) to 2,135,335 in 2004 (a rate of 723 per 100,000). By 2007 this figure had risen to 2,319,258 (a rate of 750 per 100,000), and the Pew Center on the States (2008: 5) reported that “more than 1 in 100 adults in now locked up in America”. A similar trend is observed in the UK, where prison populations continue to rise. At the time of writing they had exceeded 80,000 (Prison Reform Trust 2007). UK prison rates are currently the highest in Western Europe at 148 per 100,000 (ibid.) Plans by the UK government to increase prison places by 9,500 over the next few years will give the UK an estimated imprisonment rate of 166 per 100,000, a rate higher than that currently found in a number of eastern European countries (Slovakia and Romania currently have rates of 155 per 100,000 and Hungary a rate of 156 per 100,000) (Prison Reform Trust 2007). It is likely that the rise in prison populations on both sides of the Atlantic is driven by similar factors. Most serious commentators agree that rises in the numbers of prisoners are not driven by increases in crime but by policy decisions. In the UK the rise is due to a number of factors, which include an increase in sentencing severity (Carter 2003), more people serving time under licence for longer, and the greater use of recall to custody for breach of licence conditions (Prison Reform Trust 2007). Similar factors are at play in the USA where 'three strikes' sentencing policies and breaches of probation and parole conditions are driving the increases (The Pew Center on the States 2008). In the UK the debate engendered by rising prison populations has been relatively high-profile and has involved a wide range of contributions from politicians, academics and social commentators. However, much of the debate, whether for or against the greater use of imprisonment, tends to focus on the moral, political and social arguments. Relatively little has been written about the economic arguments for and against the use of prison. This is despite the substantial costs incurred by the state in building and maintaining prisons. In the UK it is estimated that each new prison place costs £119,000 and that the annual average cost for each prisoner exceeds £40,000 (Prison Reform Trust 2007: 5). The overall cost of the criminal justice system has risen from 2% of gross domestic product (GDP) to 2.5% of GDP over the past 10 years (ibid.). In 2008, the USA spent $44 billion on corrections compared to $10.6 billion in 1987 (in 2007, California alone spent $8.8 billion dollars on corrections). Allowing for inflation, this was an increase of 127% (The Pew Center on the States 2008: 4). There are many reasons for the lack of an economic dimension to the debate on imprisonment. One is that the demand for criminal justice services is driven by the decisions that sentencers make, and, until recently, in the UK, they had neither considered the resource implications of different sentencing options nor the capacity of the criminal justice system to manage different sentences. However, this position is starting to change. Recently, a series of Court of Appeal rulings in the UK has made clear that judges can take account of resources in the prison system when deciding a sentence in cases where they have a choice of penalty, and custody is not needed for the protection of the public (The Times, 5th November 2007). Most recently, Sir Igor Judge called for judges to take into account prison conditions (specifically over-crowding) when deciding the punitive element of a sentence (The
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending
405
Times, 5th November 2007). However, these are controversial debates, carrying, as they do, possible implications for the independence of judicial decision-making. Consideration of the economic dimensions of the policy debate has been developed further in some other countries than it has been in the UK. The Pew Center on the States (2008: 3) finds that there are signs that policy makers in the USA are diversifying their states' sentencing options with the intention of diverting low-risk offenders into lower cost sanctions. An example of such thinking comes from the Washington State Legislature, which, faced with long-term forecasts indicating the need for new prisons, explicitly set out to identify evidence-based options that would reduce the future need for prison beds (Aos et al. 2001). However, there are signs in the UK of a greater awareness of economic arguments for and against prison, at a policy level. For instance, in 2003, Lord Carter's review of the prison system in the UK noted that: “Decisions on changing the capacity and composition of prisons need to be set against other interventions to reduce crime and improve public confidence. For example it may be more effective to increase the number of police officers rather than to extend the period of community supervision for offenders. These investment decisions need to be informed by evidence of what works, when and for whom” (Carter 2003, p. 14). A second reason for the lack of an economic dimension to the debate on imprisonment is the relative paucity of data upon which to base such arguments. Economic arguments for and against the use of prison are based partially on estimates of the effect of prison and its alternatives. However, the evidence on the effectiveness of sentencing options is limited (for recent reviews of the effectiveness literature see Smith et al. 2002 and Villettaz et al. 2006). A third reason is that there is even less evidence for the economic efficiency of criminal justice interventions than there is evidence for their effectiveness. Cohen (2000) points out that, despite the widespread use of economic techniques in other policy domains, the use of economic evaluation tools is relatively rare in criminal justice. A number of authors have identified a dearth of rigorous applications of economic analysis to interventions in criminal justice (Bushway and Reuter 2005; Cohen 2000; DiIulio 1996). McDougall et al. (2003) recently undertook a review of economic studies of prison. Reinforcing the point that there is a dearth of economic analysis of criminal justice interventions, they conclude their review as follows: “It is evident that consideration should be given to determining at which point imprisonment ceases (or begins) to be cost-beneficial and a non-custodial alternative may or may not be appropriate. To date there is no specific research guidance on this, nor evidence on the types of offenders for which a custodial sentence is or is not cost-beneficial. This is an appropriate question for further research” (McDougall et al. 2003, p. 125). Research objective The aim of this research was to test the hypothesis that prisons are an economically efficient means of reducing re-offending. Given the jurisdiction-specific nature of economic data, the research was undertaken for the UK. For the purposes of the research,
406
K. Marsh, C. Fox
'standard prison sentences' were defined as confinement to prison or jail. The alternatives examined included community-based interventions and prison-based interventions that add elements onto standard prison sentences. The research focused on the efficiency of sentencing options for adult offenders (defined as being 18 years or older). Economic efficiency was specifically defined as the incremental cost of prison when compared with its alternatives, where cost was defined as comprising the following three components: the cost of implementing the sentence, the cost of the crime committed during the sentence, and the cost of the crime after sentence.
Method As there are limited economic data available on sentencing options, the methodology comprised the following two parts. First a rapid evidence assessment of effectiveness studies was undertaken. This identified the change in re-offending resulting from moving an offender from a standard prison sentence to an alternative sentence. Second, an economic analysis was undertaken to transform the data on change in reoffending into an estimate of the economic efficiency of moving an offender from a standard prison sentence to an alternative sentence. Review of effectiveness studies The following databases were searched to identify studies of the effectiveness of alternative sentences on reducing re-offending: Applied Social Science Index and Abstracts, British Library Direct, Criminal Justice Abstracts, CSA Social Services Abstracts, CSA Sociological Abstracts, Social Policy and Practice, and Web of Knowledge1. A total of 986 studies were identified in the search. A number of selection criteria were applied to determine which studies would be included in the review: 1. The study had to include at least two distinct groups: a group receiving a standard prison sentence and a group receiving an alternative sentence. 2. The study had to include at least one outcome measure of recidivism: offending, arrest, conviction, or incarceration. 3. The study had to include data to enable the calculation of relative effect of the interventions (odds ratio) and the variance of this effect. 1
The following search terms were used to interrogate the databases: [(offender* or prisoner*) and (violence and [domestic or person*]) or “partner abuse” or sexual or burglary or robbery or theft or rape* or fraud or paedophilia* or forger* or drug* or motoring or driving] and [conviction or recidivism or arrest* or re-arrest or reconviction or offending] and [discharge* or fine* or compensation or (community and [order*” or service or supervision or sentence*]) or probation or custody or reparation or curfew* or “house arrest” or suspended or surveillance or deferred] or [electronic and (monitor* or tagging)] or [licence or parole or “control order*” or licence* or parole or resettlement or imprisonment or incarcerat* or intervention* or attitude* or thinking or cognitive or “anger management” or “aggression replacement” or “victim aware*” or counsel* or therapy or mentor* or “offender management” or “after care” or “correctional intervention*” or “offender prevention” or restitution or mediation] and effectiv* not [cctv or Africa* or Asia* or India* or “far east” or “south America” or mexico or “middle east” or mediterr*]
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending
407
4. The study had to have been published in 1996 or after. 5. The study had to be published in the English language. 6. The target population in the study had to be over the age of criminal responsibility in the UK (10 years old). 7. The study had to have been undertaken in a western context. That is, it had to have been undertaken in Europe, North America, Australia or New Zealand. 8. The study had to have employed an experimental research design that was scored as 3 or above on the Maryland Scale of Methodological Rigor (Sherman et al. 1997). The abstracts for the papers were read by two researchers and were excluded if they did not meet the criteria outlined above. Following the review of abstracts, 121 studies remained in the review. Those papers were collected and again read by two researchers and were excluded if they failed to meet the inclusion criteria. This search was supplemented by data from Campbell Collaboration reviews. Data were extracted directly from these reviews if they met the inclusion criteria. In many instances, Campbell Collaboration reviews did not report sufficient data on individual studies to allow their inclusion in the analysis, and the scope of the research meant that these studies could not be collected and reviewed separately. A total of 15 studies with data on adult interventions were identified. The studies were read to extract data for the analysis. Data were extracted on the intervention type, its components and intensity, the counterfactual, the location of the study, the characteristics of the offending population (including gender, age, and offending risk factors), study design, and the effect of the intervention. A total of 91 estimates of the effect on re-offending of prison for adults when compared against alternative sentences were extracted from the studies. The number of data were greater than the number of studies, as some of the studies were themselves reviews and some of the studies evaluated more than one treatment– counterfactual combination. As some of the studies were reviews, the studies included in this review and those reviews included in this study were crossreferenced to ensure that no data were duplicated. A number of rules were applied to select the effect data from the studies: 1. Where multiple crime outcomes were reported, the outcome 'closest' to the offence was extracted. That is, reported offending was preferred to arrest, which was preferred to convictions, which was preferred to incarceration. 2. Where outcomes were reported for different follow-up periods, outcomes for the longest follow-up period were extracted. 3. Estimates of effect adjusted for confounding variables were preferred. Effect sizes were calculated as odds ratios, using Comprehensive Meta-Analysis. In some instances data were available to calculate effect sizes using more than one formula. Where this was the case, a hierarchy of effect size measurements was applied. First, odds ratios were calculated, based on the proportions of participants offending after intervention and the sample sizes in the treatment and control groups. Second, odds ratios were calculated from estimates of Cohen's d and the standard error associated with Cohen's d. Finally, odds ratios were calculated, using estimates of mean offending after intervention, the standard deviation associated with mean offending, and the sample sizes in the treatment and control groups.
408
K. Marsh, C. Fox
A random effects meta-analysis was undertaken to estimate an overall effect size from the data collected for each combination of sentencing options. For identification of the effect estimate most appropriate for UK policy, where there was heterogeneity in the estimate of effect size for each treatment–counterfactual combination (defined as I2 >60) a sub-group analysis was run by location. If there were a statistically significant difference (at a 95% confidence level) in the effect size identified in different locations, the effect size for the 'worst' location was removed from the analysis, where 'worst' was defined according to the following hierarchy: UK-based studies were preferred to those not based in the UK, and studies not based in the UK or in the USA were preferred to USAbased studies. Once the data for the 'worst' location had been removed, the metaanalysis was repeated. This process was repeated until the heterogeneity in the underlying data had been removed, there was no statistically significant difference in the effect in different locations, or only UK-based studies remained in the analysis. A similar process was undertaken to ensure that the estimate of effect was based on the best methods available. Once the process to select the most appropriate location had been complete, if there was still statistical heterogeneity within the data (again defined as I2 >60), a sub-group analysis was run by research design. If there was a statistically significant difference (at a 95% confidence level) in the effect size between research designs, the effect for the 'worst' research design was removed from the analysis, where 'worst' was defined in accordance with the following hierarchy: randomised controlled trials and quasiexperimental designs that attempted to control for differences between the treatment and control groups were preferred over quasi-experimental designs where no efforts were made to control for differences between the treatment and control groups. Once the data for the 'worst' research design had been removed, the meta-analysis was repeated. This process was repeated until the heterogeneity in the underlying data had been removed, there was no statistically significant difference in the effect in different research designs, or only randomised controlled trial designs remained in the analysis. Once the above process had been complete, the final meta-analysis incorporated 41 estimates of the effect size of adult interventions Economic analysis An economic analysis was undertaken to transform the data on change in reoffending into an estimate on the economic efficiency of alternatives to standard prison sentences. All costs were calculated in 2007 prices. The economic analysis involved four components. First, estimation of the change in the cost of the crime after sentence if an offender is diverted from prison. Second, estimation of the change in the cost of crime during the sentence if an offender is diverted from prison. Third, estimation of the incremental cost of alternative interventions compared with prison. Finally, the combination of these three components to estimate the net benefit of an alternative intervention compared with prison. The remainder of the methods section summarises the analysis undertaken for these components.
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending
409
Change in offending after sentence The first component in the economic analysis was the estimation of the change in the cost of crime committed after the sentence when an offender is diverted from prison to an alternative sentence—the rehabilitation and/or specific deterrence effects. The effect data identified during the review measured the change in re-offending after sentence when an offender is diverted from prison to an alternative sentence. The objective of this element of the economic analysis was to translate this effect into a monetary value. The first step in the analysis was the estimatation of the number of offences likely to be committed in each year after release if an offender is given a prison sentence. The analysis was conducted for a male offender released from prison at the age of 25 years. Equation 1 shows the function used to calculate this baseline level of re-offending. NumberCrimesy ¼ ConvictChances * ConvictAve * OffConvict * CrimeRatioy
ð1Þ
where NumberCrimesy is the number of crimes committed in year y by an offender in prison for offence type s 2, ConvictChances is the chance that an offender in prison for offence type s is convicted of any offence in the year following release (source: Cuncliffe and Sheperd 2007), ConvictAve is the average number of convictions in the first year after release per adult male offender convicted [source: Research Development and Statistics Directorate of the National Offender Management Service (RDS NOMS) 2007], OffConvict is the number of offences committed per conviction3, and CrimeRatioy is the ratio of the number of crimes committed in year y (where y=1 corresponds to the first year of release when the offender is 25 years old) to the number of crimes committed at the age of 25 years (source: Farrington et al. 2006). The next step in the analysis was to estimate the average cost of a crime committed after release. Equation 2 summarises the function used to calculate the average value of a crime: ValueCrimes ¼
R X O X
OffenceDistos *CostCrimeor
ð2Þ
r¼1 o¼1
where ValueCrimes is the average value of a crime committed after release from prison by an offender sentenced for offence s, OffenceDistos is the chance that a crime committed by an offender released from prison for offence s will be a particular crime type o4 (source: RDS NOMS 2007), and CostCrimeor is the cost of resource type r associated with offence o (source: Dubourg et al. 2005). The analysis was run from two different perspectives. First, a public sector perspective was employed, where resource type r included criminal justice costs and National Health Service (NHS) 2
The following offence types were included in the analysis: violence, robbery, sexual offences, burglary, theft and handling, fraud and forgery, motoring offences, drug offences and other offences.
3
The number of offences committed per conviction is calculated as the product of the average number of convictions per recorded offence (source: RDS NOMS 2007) and the average number of offences per recorded crime estimated from data from the British Crime Survey (source: Nicholas et al. 2005).
4
The model specifies the following crime types: violence, robbery, sexual offences, burglary, theft and handling, fraud and forgery, motoring offences, drug offences and other offences. The sum of the probabilities that a crime is of a particular type is equal to 1.
410
K. Marsh, C. Fox
costs. Second, a societal perspective was employed, where resource type r included criminal justice costs, NHS costs, property stolen and not recovered, property damaged, lost output and the physical and psychological suffering of the victim. The total cost of crime committed after release from prison until the age of 50 years is thus given by the equation: LifetimeCosts ¼
25 X ValueCrimes * NumberCrimesy y¼1
ð1 þ DRÞy1
ð3Þ
where LifetimeCosts is the cost of crime after release from prison until the age of 50 years for offenders sentenced for offence s and released at the age of 25 years, ValueCrimes is the average value of a crime committed after release for offenders released after being sentenced for offence s (from Eq. 2), NumberCrimesy is the number of crimes committed in year y by offenders released from prison after being sentenced for offence s (from Eq. 3), and DR is the discount rate. In line with Green Book guidance (Her Majesty's Treasury 2003), a discount rate of 3.5% was employed in the analysis. Table 1 shows the output from Eq. 3. It demonstrates that the most costly offenders released from prison are those originally sentenced for robbery. Those offenders sentenced for robbery and released from prison when aged 25 years go on to commit approximately £941,000 worth of crimes by the time they are aged 50 years. As the baseline cost of re-offending is calculated by original offence type, and those offenders who are being diverted from prison to a particular alternative intervention might have committed a number of different offence types, Eq. 4 summarises the function used to calculate the baseline cost of re-offending for each alternative intervention identified in the review: LifetimeCosti ¼
S X
ChanceOffenceis * LifetimeCosts
ð4Þ
s¼1
where LifetimeCosti is the lifetime cost of crime committed after release from prison of those offenders who could be diverted to alternative intervention i, ChanceOffenceis is the proportion of offenders currently sentenced to intervention i who have been sentenced for committing offence s (source: RDS NOMS 2007)5, and LifetimeCosts is the lifetime cost of crime after release if the offender had been sentenced to prison for offence s (from Eq. 3). The change in the cost of crime if an offender is diverted from prison to an alternative sentence was calculated as follows: ChangeCostPosti ¼ LifetimeCosti * Effecti
ð5Þ
where ChangeCostPosti is the change in the costs of crime after sentence as a result of diverting an offender from prison to intervention i, LifetimeCosti is the cost of crime 5
It is assumed that the offence distribution of community service and community supervision is the same as that for those receiving a community order in the UK, that the distribution of offence types for residential drug treatment and surveillance with drug treatment is the same as that for those receiving a Drug Treatment and Testing Requirement (DTTR) in the UK, and that the distribution of offence types for any prison sentence and surveillance is the same as that for those receiving a custodial sentence in the UK.
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending
411
Table 1 Average discounted economic value of the crime after release from prison, by sentence offence (2007 prices) Sentence Offence (s)
Total Cost of Offending after Release (LifetimeCosts)
Robbery Burglary Violence Motoring Other Drug Theft and Handling Fraud and Forgery Sexual
£940,674 £369,771 £352,358 £342,183 £245,056 £237,357 £190,851 £173,043 £90,399
after release from prison if those offenders diverted to intervention i had, instead, been given a prison sentence (source: Eq. 4), and Effecti is the relative risk of re-offending with intervention i compared with prison (source: effectiveness review6). Change in offending during the sentence The second component of the economic analysis was the calculation of the cost of crime during the sentence associated with diverting an offender from prison to an alternative sentence—the incapacitation effect. For instance, if an offender is diverted from prison to a community sentence, it is possible that this will increase offending, as the offender remains part of the community and is able to commit crime. However, if the alternative sentence is another type of custodial sentence, then there will be no change in crime. Thus, this component of the analysis only applies to community-based alternatives to prison. It was assumed that number of crimes in the first year after release would have been committed for each year in prison. Equation 6 summarises the cost of crime that would have been committed had the offender not been sent to prison: SentenceCosti ¼
S X
ChanceOffenceis * ðNumberCrimes * ValueCrimes Þ * LengthCust
s¼1
ð6Þ where SentenceCosti is the cost of crime that would have been committed had the offender not been sent to prison for those offenders who could be diverted to intervention i, ChanceOffenceis is the proportion of offenders currently sentenced to intervention i who have been sentenced for committing offence s (source: RDS NOMS 2007)7, NumberCrimes is the number of crimes committed in the first year 6
The odds ratios estimated in the meta-analysis were converted to relative risks by the following equation: RR = OR/[(1-EC)+(OR.EC)], where RR is the relative risk, OR is the odds ratio, and EC is the baseline risk.
7
It is assumed that the offence distribution of community service and community supervision is the same as that for those receiving a community order in the UK, that the distribution of offence types for residential drug treatment and surveillance with drug treatment is the same as that for those receiving a DTTR in the UK, and that the distribution of offence types for any prison sentence and surveillance is the same as that for those receiving a custodial sentence in the UK.
412
K. Marsh, C. Fox
after release by offenders in prison for offence type s 8 (see Eq. 1), ValueCrimes is the average value of a crime for those sentenced for offence s (from Eq. 2), and LengthCust is the average length of a prison sentence (source: RDS NOMS 2007). It was assumed that the effect of the community-based alternatives to prison takes effect at the point of first contact with the offender. Thus, the effect on re-offending of community-based interventions identified in the review of the literature is assumed to commence at the beginning of the sentence. The cost of crime whilst on a community sentence is calculated as follows: ChangeCostDuringi ¼ SentenceCosti * ð1 Effecti Þ
ð7Þ
where ChangeCostDuringi is the change in the cost of crime during the sentence associated with diverting an offender from prison to intervention i, SentenceCosti is the cost of crime that would have been committed had the offender not been sent to prison for those offenders who could be diverted to intervention i, and Effecti is the relative risk of re-offending with intervention i compared to prison (source: effectiveness review). The incremental cost of alternative interventions The third component of the economic analysis is the calculation of the incremental economic cost of implementing alternatives to prison. A review of existing studies was undertaken to identify the economic cost of implementing prison and non-prison sentences. Three existing reviews of economics studies of criminal justice interventions were identified: Cartwright (2000); Welsh and Farrington (2000); and McDougall et al. (2003). The existing reviews were supplemented with a search of the following databases for the period 2000 to 2006: Applied Social Science Index and Abstracts (ASSIA); Criminal Justice Abstracts; National Criminal Justice Reference Service (NCJRS); Social Policy & Practice; International Bibliography of the Social Sciences; British Library Direct; Sociological Abstracts; PsycInfo; Social Services Abstracts; Social Care Online; Web of Knowledge; and Dissertation Abstracts9. The review identified 61 estimates of the unit cost of sentences in the UK. The sentence types for which unit cost data had been collected were then matched against those identified in the effect review. Where more than one unit cost estimate was identified, the average of these estimates was used in the analysis. Table 2 summarises the annual cost of sentencing options extracted from the literature. Equation 8 summarises how these data were employed to estimate the incremental cost of alternatives to prison: InterventionCosti ¼ ðCosti * Lengthi Þ ðCostCustody * LengthCusti Þ
ð8Þ
where InterventionCosti is the incremental cost of intervention i compared with standard prison sentence, Costi is the annual cost of intervention i (source: review of 8
The following offence types were included in the analysis: violence, robbery, sexual offences, burglary, theft and handling, fraud and forgery, motoring offences, drug offences and other offences.
9
The following search strategy was employed to search the databases: (crim* or offend* or reoffend* or recidivis*) and (econ* or cost*) and (benefit*) and (intervention* or outcome*).
£2,641
£2,641
£5,299 £5,947 £8,604 £31,124
Community service (adult)
Community supervision (adult)
Residential drug treatment (adult) Surveillance (adult) Surveillance with drug treatment Prison with behavioural intervention
£29,646
£28,690
Prison with sex offender treatment
Prison with drug treatment
Prison with educational intervention £27,109
£26,033
Standard prison sentence
Average cost of local (£23,801), open (£20,704), and remand (£26,249) prison for male offenders (Bowles Pradiptyo 2004; 2002 prices) Average of: (1) £2,761 cost of supervision order (Bowles and Pradiptyo 2004), 2002 prices), (2) £2,400 cost of community order (Hansard (2007; 2007 prices), and (3) £2,400 cost of community order/suspended sentence (Hansard 2007; 2007 prices) Average of: (1) £2,761 cost of supervision order (Bowles and Pradiptyo 2004; 2002 prices), (2) £2,400 cost of c ommunity order (Hansard 2007; 2007 prices), and (3) £2,400 cost of community order/suspended sentence (Hansard 2007; 2007 prices) Source: data provided by the NTA Cost of ISSP programme (6 months): £2,761 (Bowles and Pradiptyo 2004; 2002 prices). Sum of surveillance and community drug treatment (above) Sum of (1) average cost of local (£23,801), open (£20,704), and remand (£26,249) prison for male offenders (Bowles and Pradiptyo 2004; 2002 prices), and (2) average cost of Enhanced Thinking Skills Course (approximately 40 hours) in prison for high-risk offenders (£7,000) and general offenders (£2,000) (SEU 2002; 2002 prices) Cost of prison [average cost of local (£23,801), open (£20,704), and remand (£26,249) prison for male offenders (Bowles and Pradiptyo 2004;, 2002 prices) multiplied by the ratio of the cost educational training in prison in the USA ($46,859; source Aos et al. 2001) to the cost of standard prison in the USA ($45,000, source Aos et al. 2001)] Cost of prison [average cost of local (£23,801), open (£20,704), and remand (£26,249) prison for male offenders (Bowles and Pradiptyo 2004; 2002 prices) multiplied by the ratio of the cost of sex offender treatment in prison in the USA ($51,246; source Aos et al. 2001) to the cost of standard prison in the USA ($45,000; source Aos et al. 2001)] Sum of (1) cost of prison [average cost of local (£23,801), open (£20,704), and remand (£26,249) prison for male offenders (Bowles and Pradiptyo 2004; 2002 prices)] and (2) cost of drug treatment [average of (i) £2,770 cost of inpatient substance misuse treatment, and £2,841 methadone treatment (Godfrey et al. 2004; 2000 prices), and (ii) NTA unit costs of tier 3 and 4 treatment £2,920 (data provided by NTA, 2007 prices)]
£ (2007 Prices) Source
Intervention
Table 2 Summary of data on annual cost of sentencing options (NTA National Treatment Agency, ISSP Intensive Supervision and Surveillance Programme, SEU Social Exclusion Unit)
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending 413
414
K. Marsh, C. Fox
economic data), Lengthi is the average length of intervention i (source: RDS NOMS 200710, CostCustody is the annual cost of standard prison (source: review of economic data), and LengthCusti is the average length of a standard prison sentence (source: RDS NOMS 2007). Net benefit of diverting offenders away from standard prison sentences The final step of the economic analysis was to combine the estimates of the incremental implementation cost and the change in the cost of crime both during and after the sentence to estimate the net economic benefit associated with diverting an offender from prison to an alternative intervention. Equation 9 summarises this calculation: NetBenefiti ¼ ChangeCostPosti ChangeCostDuringi InterventionCosti
ð9Þ
where NetBenefiti is the net economic benefit associated with diverting an offender from prison to intervention i, ChangeCostPosti is the change in the costs of crime after sentence as a result of diverting an offender from prison to intervention i (from Eq. 5), ChangeCostDuringi is the change in the cost of crime during the sentence associated with diverting an offender from prison to intervention i (from Eq. 7), and InterventionCosti is the incremental cost of intervention i compared with standard prison (from Eq. 8).
Findings Is prison effective at reducing re-offending? Table 3 summarises the results of the meta-analysis of the effectiveness of alternatives to standard prison for adult offenders. In each case the effect represents the impact on offending of the alternative to standard prison when compared with prison. The result of the meta-analysis for each treatment–counterfactual combination was graded for methodology and transferability to the UK policy context according to the following scales: 1. Methodology: + the meta-analysis included at least one effect size calculated from randomised controlled trials or quasi-experimental designs combined with statistical techniques for controlling for differences in the treatment and control groups − the meta-analysis included only quasi-experimental designs that made no attempt to control for differences between the treatment and control groups
10
It is assumed that the length of community service and community supervision is the same as that for a community order in the UK, that the length of residential drug treatment and surveillance with drug treatment is the same as that for a DTTR in the UK, and that the length of any prison sentence and surveillance is same as that for custodial sentence in the UK.
0.513 0.568
0.453 0.463
0.580 0.697
na na 0.1188 na 0.8716 na <0.0001
+ −
+ − − − − ++ −
+ +
+ − − + − − −
Q-Statistic Transferability Method (P Value)
99.50 <0.0001 94.83 <0.0001
0.00 0.00 58.89 0.00 0.00 0.00 99.94
4,711 4,556
1.924 1.821 0.581 0.741 0.822 0.938 0.802
14 13
0.325 0.368 0.331 0.408 0.731 0.889 0.707
125 286 786 6,841 161 164 7,623
1 1 2 1 2 1 6
Community vs prison Community service Community supervision Residential drug treatment Surveillance Surveillance with drug treatment Prison with added Prison with behavioural intervention elements vs prison Prison with educational/vocational Intervention Prison with sex offender treatment Prison with drug treatment
0.791 0.818 0.439 0.550 0.775 0.913 0.753
Studies Number Effect 95% Confidence 95% Confidence I2 (OR) Interval (Low) Interval (High
Intervention
Table 3 Meta-analysis of alternatives to standard prison sentences for adult offenders (na not applicable)
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending 415
416
K. Marsh, C. Fox
2. Transferability to the UK ++ the meta-analysis included UK-based studies + the meta-analysis included no UK-based studies but at least one study that was not UK based or US based (studies from Australia, Canada, Europe or New Zealand) − the meta analysis included only US-based studies or studies for which the location was not known
Table 3 demonstrates that the following interventions reduced re-offending when compared with standard prison sentences: & &
& & & &
&
Residential drug treatment, specifically the Drug Treatment Alternative to Prison Programme (DTAP), which diverts non-violent drug-addicted offenders into community-based facilities. Surveillance in the community, specifically the Intensive Supervision Programme (ISP), which is a widely used programme with components that commonly include reduced caseloads, surveillance, drug testing, treatment, and employment. ISP differs from standard probation as it involves more intensive monitoring of, and interaction with, the offender. Surveillance in the community with drug treatment, specifically ISP and house arrest with electronic monitoring. Offender behaviour interventions undertaken in prison, specifically a cognitive behavioural programme designed to change thinking patterns and attitudes associated with offending. Educational/vocational interventions undertaken in prison, specifically basic education, prison-based vocational training, support in finding employment after release, and prison industry employment and/or training. Sex offender treatment undertaken in prison, specifically cognitive behavioural and psychotherapeutic approaches, including both individual and group treatments. Relapse-prevention work was explicitly stated as a feature of some treatments. Drug treatment undertaken in prison, specifically therapeutic community programmes in prison, case management approaches and programmes preparing for parole.
Table 3 also identifies two intervention types for which the evidence that met our criteria showed no statistically significant difference in offending when compared with standard prison: community service and community-based supervision while on probation. Variability in effect on re-offending The mean change in re-offending from diverting offenders from standard prison to an alternative sentence tells an incomplete story. Despite attempts to group the analysis by intervention type and to focus on the 'best' available methods and 'most appropriate' study locations, there is still substantial variability in effect sizes across
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending
417
studies, as indicated by the I2 statistics and the P values on the Q statistics in Table 2. This is particularly the case for educational intervention in prison, sex offender interventions in prison, and drug treatment in prison. Is prison efficient at reducing re-offending? Figure 1 shows the net benefit to the public sector of adult sentencing options compared with standard prison. It demonstrates that the following interventions all save public sector costs when compared with investment in standard prison sentences11: & & & & & &
Residential drug treatment [net benefit ca. £88,500 per offender, 95% confidence interval (CI) ca. £61,000–ca. £106,000]. Surveillance (net benefit ca. £57,500 per offender, 95% CI ca. £39,500– ca. £71,000). Surveillance with drug treatment (net benefit ca. £41,500 per offender, 95% CI ca. £36,000–ca. £46,500). Educational and vocational interventions in prison (net benefit ca. £19,000 per offender, 95% CI ca. £15,000–ca. £23,000). Sex offender treatment in prison (net benefit ca. £35,000 per offender, 95% CI ca. £30,000–ca. £40,000). Drug treatment in prison (net benefit ca. £30,000 per offender, 95% CI ca. £21,000–ca. £41,000).
Three interventions in Fig. 1 have a mean net benefit per offender to the public sector when compared with standard prison, but the difference in economic cost between these interventions and standard prison is not statistically significant: community service, community supervision and general offender behavioural interventions in prison. Figure 2 shows the net benefit to society of adult sentencing options compared with standard prison. It demonstrates that the following interventions all save social costs when compared with investment in standard prison sentences: & & & &
11
Residential drug treatment (net benefit ca. £203,000 per offender, 95% CI ca. £110,000–ca. £262,500). Surveillance (net benefit ca. £130,500 per offender, 95% CI ca. £70,000– ca. £175,500). Surveillance with drug treatment (net benefit ca. £61,500 per offender, 95% CI ca. £43,000–ca. £78,500). General offender behaviour interventions in prison (net benefit ca. £17,500 per offender, 95% CI ca. £10,500–ca. £24,000).
The confidence intervals around the estimates of net benefit are derived by applying the economic model to the confidence intervals on the estimates of effect derived from the meta-analysis. No other estimates of uncertainty are incorporated into the confidence intervals. Any uncertainty in the other parameters of the economics model are considered in the sensitivity analysis.
418
K. Marsh, C. Fox
£300,000
£200,000
Net benefit
£100,000
£0
-£100,000
-£200,000
Prison with drug treatment
Prison with sex offender treatment
Prison with edu./voc.
Prison with behavioural intervention
Surveillance with drug treatment
Surveillance
Residential drug treatment
Community supervision
Community service
-£300,000
Fig. 1 Net benefit per offender to the public sector of adult sentencing options compared with standard prison (mean and 95% confidence intervals)
& & &
Educational and vocational interventions in prison (net benefit ca. £67,000 per offender, 95% CI ca. £53,500–ca. £80,000) Sex offender treatment in prison (net benefit ca. £130,500 per offender, 95% CI ca. £112,000–ca. £147,500) Drug treatment in prison (net benefit ca. £116,500 per offender, 95% CI ca. £80,500–ca. £145,500) £300,000
£200,000
Net benefit
£100,000
£0
-£100,000
-£200,000
Prison with drug treatment
Prison with sex offender treatment
Prison with edu./voc.
Prison with behavioural intervention
Surveillance with drug treatment
Surveillance
Residential drug treatment
Community supervision
Community service
-£300,000
Fig. 2 Net benefit per offender to society of adult sentencing options compared with standard prison (mean and 95% confidence intervals)
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending
419
Compared with the analysis undertaken from a public sector perspective, two differences are noticed from a societal perspective: 1. The mean net benefit per offender from a societal perspective is greater than that from a public sector perspective, ranging from ca. £60,000 to ca. £120,000. 2. Prison with general offender behavioural interventions now saves costs when compared with standard prison. As with the public sector perspective, the difference in economic cost between standard prison and community service or community supervision is still not statistically significant when a societal perspective is adopted. Sensitivity analysis The economic analysis is subject to a number of caveats. First, the economic model uses the post-release reconviction rates for average offenders to construct the baseline offending rate—the offending rate associated with standard prison. However, it is likely that the decision to employ sentencing options other than prison will apply to offenders with a lower than average chance of re-offending. Second, the economic model assumes a zero decay rate in the effect of the sentencing options. That is, it is assumed that any reduction in offending with, for instance, drug treatment compared with standard prison identified in the short term by effectiveness studies will be maintained for the lifetime of the offender. This assumption will obviously overestimate the effect of the alternatives to standard prison. Sensitivity analysis was run to test the impact of these caveats, as well as any possible overestimation of effect resulting from publication bias, on the results of the economic analysis. The following changes in the parameters of the economic analysis would be required before the direction of the analysis results would be changed: 1. Reconviction rate: the 2-year post-release reconviction rate would have to be at least 90% lower than the average rate used in the model before the conclusions of the model are changed. 2. Decay rate: the conclusion that surveillance with drug treatment is of net benefit compared with standard prison is changed if the decay rate is ca. 20% or higher. However, the following alternatives to prison are still of net benefit when the decay rate is 100%: drug treatment and residential drug treatment. 3. Effect size: the effect that alternative sentences would have compared with standard prison would have to be exaggerated by at least 70% before the conclusion that the sentences are of net benefit compared with standard prison is reversed.
Discussion and conclusions The hypothesis that prison is an economically efficient means of reducing reoffending in the UK is generally rejected. The analysis suggests that a number of sentencing options are more effective and more efficient than standard prison sentences, including community-based sentences and enhancements to standard prison sentences.
420
K. Marsh, C. Fox
Community-based interventions The analysis suggests that a number of community-based interventions reduce reoffending when compared with standard prison sentences, including residential drug treatment; surveillance; and surveillance with drug treatment. As these interventions are also less costly to implement than prison interventions, it is no surprise that they produce cost savings when compared with standard prison interventions. If offenders are given community sentences rather than standard prison sentences, the cost saving to the public sector ranges from ca. £41,500 per offender for surveillance with drug treatment to ca. £88,500 per offender for residential drug treatment. That is, diverting drug-misusing offenders from standard prison interventions to residential drug treatment would save the public sector ca. £88,500 per offender in terms of avoided costs of crime, as well as avoided intervention costs. The costs saved by diverting offenders from standard prison interventions to community-based sentences increases when a societal perspective is adopted. That is, diverting offenders from prison to community-based interventions reduces reoffending. This reduction in offending means fewer victim costs associated with crimes. When these avoided victim costs are added to the public sector cost savings, the total saving associated with diverting offenders from standard prison to community-based interventions ranges from ca. £61,500 per offender for surveillance with drug treatment to ca. £203,000 per offender for residential drug treatment. Enhancements to prison sentences The analysis also suggests that there are a number of enhanced prison sentences that both reduce re-offending and save public sector and social costs when compared with standard prison sentences. These include: educational and vocational interventions, sex offender treatment, offender behaviour interventions and drug treatment. If offenders are given enhanced prison sentences instead of standard prison sentences, the cost saving to the public sector ranges from ca. £19,000 per offender for educational and vocational interventions to ca. £35,000 per offender for sex offender treatment. That is, diverting sex offenders from standard prison interventions to sex offender treatment in prison would save the public sector ca. £35,000 per offender in terms of avoided costs of crime, as well as avoided intervention costs. As enhanced prison sentences costs more than standard prison sentences alone, it is possible for these interventions not to be economically efficient when compared with prison, even if they cause crime to fall. That is, if the extra cost of the intervention outweighs the value of the crime avoided, it will still make economic sense to prefer standard prison, rather than investing in additional interventions within prison. This is the case for prison with behavioural intervention when assessed from a public sector perspective. Despite the fact that such interventions reduce crime when compared with standard prison sentences, the avoided public sector costs associated with this reduction in crime are not sufficient to offset the extra cost of the intervention. However, when a societal perspective is adopted, and the cost to victims of crime is included in the analysis, the benefits associated with general offending behaviour interventions do outweigh their costs.
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending
421
The costs saved by diverting offenders from standard prison interventions to community interventions increase when a societal perspective is adopted. From a societal perspective, the total saving associated with diverting offenders from standard prison sentences to enhanced prison sentences ranges from ca. £17,500 per offender, for offender behaviour interventions in prison, to ca. £130,500 per offender, for sex offender treatment in prison. Limitations and further research There are a number of important caveats to the conclusion that these interventions are economically efficient alternatives to prison. First, there are limitations within the review of effect data. For instance, while all the studies included in the analysis employed research designs that are a grade 3 or above on the Maryland Scale of Methodological Rigor (Sherman et al. 1997), there is the possibility that some of the effect sizes are subject to sample selection bias. Also, despite efforts to focus the analysis on UK-based studies, much of the data available were collected in the USA. The impact on re-offending of only one alternative to prison was based on data collected in the UK: prison with behavioural interventions. One other alternative to standard prison sentences (adult sex offender treatment in prison) employed data collected in a non-USA context. The remainder of the estimate of effect are based on data collected in the USA. There were also large amounts of variability in the effect sizes included in the meta-analysis for some of the alternatives to prison. Finally, the analysis was based on a relatively small number of studies. In a number of instances, especially the analysis of community-based alternatives to prison, the analysis was undertaken on only one or two studies, reducing the likelihood that the results of the studies could be inferred to the UK. Secondly, due to a limited number of data on reconviction rates and the rate of decay in intervention effects, it is necessary to make a number of simplifying assumptions when constructing the economic model. The model employs postrelease reconviction rates for average offenders to construct the baseline offending rate—the offending rate associated with standard prison sentences. However, it is likely that the decision to employ sentencing options other than prison will apply to offenders with a lower than average likelihood of re-offending. The model also assumes a zero decay rate in the effect of the sentencing options. That is, it is assumed that any reduction in offending with, for instance, drug treatment compared with standard prison identified in the short term by effectiveness studies will be maintained for the lifetime of the offender. This assumption will obviously overestimate the effect of the alternatives to standard prison sentences. The analysis suggests that the conclusions are not very sensitive to estimates of intervention effect, baseline re-offending rate and decay rate impact. This provides some reassurance regarding the validity of the conclusions, despite the above caveats. However, it is important to acknowledge these caveats, and any further work in this area should seek to address them. The limitations of the analysis also suggest a number of areas where further research would be useful. First, more research needs to be undertaken on the relative effectiveness—compared with that of standard prison sentences—of a number of interventions, especially community-based ones. Studies on community-based
422
K. Marsh, C. Fox
interventions were identified as part of the review. However, no statistically significant difference between offending rates of the offenders receiving these interventions and those receiving prison sentences were identified. Second, as the hypothesis was related to the impact on re-offending of different sentencing options, the analysis focused on crime reduction in the form of the incapacitation and rehabilitation effects of alternative sentences. In the study reported here there were a number of functions of sentences that were excluded from the analysis, including the economic value of retribution and the general deterrent effect of prison. Furthermore, there were a number of costs associated with prison that were not included in the analysis, such as the impact on offenders' families and children. Halliday (2001: 58) notes the importance of incorporating these values into an economics analysis of sentences: “In appraising the costs and benefits of sentencing, ideally, it would be helpful to distinguish the costs of punishment from the costs of crime reduction and reparation, so as to link costs with the outcomes of the three purposes of sentencing, and identify any separable “price of punishment”. Before we can conclude that prison is not an economically efficient means of reducing crime, rather than just reducing re-offending, further work is required on the relative value of punishment provided by alternative sentences and the relative deterrent effects of alternative sentences. Acknowledgements The study was funded by the Monument Trust, the Lankelly Foundation and the Bromley Trust. The authors would like to acknowledge the contribution of researchers at The Matrix Knowledge Group, especially Kerry McCarthy and Rashmi Sarmah. The authors would also like to thank Paul Dolan and Carol Hedderman for their comments and support throughput the project.
References Aos, S., Phipps, P., & Barnoski, R., Lieb, R. (2001). The comparative costs and benefits of programs to reduce crime, document number: 01-05-1201, Washington State Institute for Public Policy, http:// www.wa.gov/wsipp. Bowles, R., & Pradiptyo, R. (2004). Reducing burglary initiative: an analysis of costs, benefits and cost effectiveness. Home Office Online Report 43/04. Bushway, S., & Reuter, P. (2005). Collaborating with economists. The Criminologist, 30(1), 1–4. Carter, P. (2003). Managing offenders, reducing crime: a new approach. Home Office. Cartwright, W. S. (2000). Cost-benefit analysis of drug treatment services: review of the literature. The Journal of Mental Health Policy and Economics, 3. Cohen, M. A. (2000). Measuring the costs and benefits of crime and justice. In: Volume 4 Measurement and analysis of crime and justice (pp. 263-316). National Institute of Justice, NCJ 182411). Cuncliffe, J., & Shepherd, A. (2007). Re-offending of adults: results from the 2004 cohort. Home Office Statistical Bulletin. DiIulio, J. (1996). Help wanted: economists, crime, and public policy. Journal of Economic Perspectives, 10, 1–23. Dubourg, R., et al. (2005). The economic and social costs of crime against individuals and households 2003/04. Home Office Online Report 30/05. Farrington, D. P., Coid, J. W., Harnett, L. M., Jolliffe, D., Soteriou, N., Turner, R. E., et al. (2006). Criminal careers up to age 50 and life success up to age 48: new findings from the Cambridge Study in Delinquent Development, 2nd ed. Home Office Research Study 299.
The benefit and cost of prison in the UK. The results of a model of lifetime re-offending
423
Godfrey, G., Stewart, D., & Gossop, M. (2004). Economic analysis of costs and consequences of the treatment of drug misuse: 2-year outcome data from the National Treatment Outcome Research Study. Addiction, 99. Hansard (2007). Written answer column WA208. 21st March 2007. Halliday, J. (2001). Making punishments work: report of a review of the sentencing framework for England and Wales. Home Office. Her Majesty's Treasury (2003). The green book: appraisal and evaluation in central government. London: TSO. International Centre for Prison Studies (2008) World prison briefing. Retrieved April 17, 2008 from http:// www.kcl.ac.uk/depsta/law/research/icps/worldbrief/. McDougall, C., Cohen, M., Swaray, R., & Perry, A. (2003). The costs and benefits of sentencing: a systematic review. The Annals of the American Academy of Political and Social Science, 587, 160–177. Nicholas, S., et al. (2005). Crime in England and Wales, 2004/2005. Home Office Statistical Bulletin, 11/05. Prison Reform Trust .(2007). Bromley briefing: prison factfile, December 2007. Prison Reform Trust. Research Development and Statistics Directorate of the National Offender Management Service (RDS NOMS). (2007). Sentencing statistics 2005, England and Wales. Home Office Statistical Bulletin. Social Exclusion Unit. (2002). Reducing re-offending by ex-prisoners. Office for the Deputy Prime Minister. Sherman, L. W., Gottfredson, D., MacKenzie, D., Eck, J., Reuter, P., & Bushaway, S. (1997). Preventing crime: What works, what doesn’t, what’s promising. Office of Justice Programs Research Report: U.S. Department of Justice. Smith, P., Goggin, C., & Gendreau, P. (2002). The effects of prison sentences and intermediate sanctions on recidivism: General effects and individual differences 2002-01. Public Works and Government Services Canada. The Pew Center on the States. (2008). One in one hundred: Behind bars in America 2008. http://www. pewcenteronthestates.org/uploadedFiles/8015PCTS_Prison08_FINAL_2-1-1_FORWEB.pdf. The Times, 5th November. (2007). Retrieved 30 November, 2007 from http://www.timesonline.co.uk/tol/ news/politics/article2806831.ece. Villettaz, P., Killias, M., & Zoder, I. (2006). The effects of custodial vs. non-custodial sentences on reoffending. A systematic review of the state of knowledge. A report to the Campbell Collaboration Crime and Justice Group. Welsh, B. C., & Farrington, D. P. (2000). Correctional intervention programs and cost-benefit analysis. Criminal Justice and Behaviour, 27, 115–133.
Kevin Marsh Kevin Marsh is head of economics at The Matrix Knowledge Group (TMKG), London. He completed his Ph.D. in Economics at the University of Bath, UK, specialising in monetary techniques for valuing environmental resources. Following a year at the Social Disadvantage Research Centre, Oxford University, Kevin joined TMKG in 2003. His research interests include the economic evaluation of public policy, in particular criminal justice and public health interventions.
Chris Fox Chris Fox is a Principal Lecturer in Criminology at Manchester Metropolitan University, Manchester, UK. He specialises in evaluations of social policy, with a particular focus on criminal justice and crime reduction. He is Joint Editor of Safer Communities, a journal for crime reduction and community justice practitioners. He is a trustee of Community Service Volunteers (CSV), the UK's largest volunteering and training organisation.