, 2012; online edn, Oxford Academic , 1 Jan. 2014 ), https://doi.org/10.1093/acprof:oso/9780199697243.003.0008, accessed 9 Sept. 2024.
Navbar Search Filter Mobile Enter search term Search Navbar Search Filter Enter search term SearchThe offender characteristics highlighted by the analysis and modelling of the previous chapters are reiterated and the areas where policy might influence crime identified and reviewed. One particular policy initiative which was influenced by this research, the Persistent and other Priority Offender (PPO) programme, is outlined. The theory proposed in this book is used to explain the results of a policy evaluation of that programme, including some results which the evaluators found perplexing. The policy implications of the theory are discussed and the conditions under which particular policies would be effective as a crime reduction strategy are identified. In particular of the 20% reduction in BCS crime between 1994 and 2000 only 1.5% is explained by increased use of incarceration and 12% by demographic changes. Some frequently asked questions concerning the theory are answered.
Collection: Oxford Scholarship OnlineIn this book, we have proposed and tested a simple theory of criminal careers that makes exact quantitative predictions, and we have applied it to several different cross-sectional and longitudinal databases. In Chapter 7, we showed that it could be used to predict the prison population and the number of offenders in the national DNA database. In this chapter, we outline some of the policy implications of the theory, and especially what the theory says about the effect of conviction and imprisonment on criminal careers.
The theory has a number of implications for criminal justice policy. Some of these have been touched on earlier but are reviewed systematically here. The main conclusion is that the criminal justice system (CJS) does control crime. The most important single factor leading to significant reductions in crime would be to increase the probability of capture and conviction given an offence. In agreement with this, Farrington and Jolliffe (2005) found that aggregate crime rates in four different countries were negatively correlated with the probability of conviction. However, having convicted an offender, prison should be used only for those offenders who society is prepared to incarcerate for most of their active lives. Treatment programmes for offenders, in the community or in prison, should be targeted at the ‘high-risk’ category of offenders where the modest individual effects of the programmes will be amplified by repetition.
We begin by reviewing the main features of our theory. We believe that the analysis presented above provides convincing evidence in support of the theory and, if accepted, leads to the following implications:
Most offences are committed by members of three clearly distinguishable offender categories, whose characteristics are relatively fixed in early childhood and remain so until they desist from crime.
On reaching the age of criminal responsibility, the proportion of the population in each of the offender categories in any given year has been essentially constant since at least 1963.
The main reason for offenders desisting from crime is their interaction with the criminal justice system. Offenders do not simply ‘grow out’ of crime due to either an internal maturation process or by general interaction with their environment. The most important event which causes desistance is conviction.
Apart from those prisoners kept in prison for large fractions of their active lives, there is no overall long-term effect on crime simply due to putting some offenders in prison, unless the prison population is allowed to increase indefinitely.
We cannot, of course, claim that our theory is a true or unique reflection of reality. The categories we define are after all just inferences from the statistical properties of our data. However, what we can say is that our theory is based on simple but plausible assumptions about the offender population and process of conviction and reconviction which replicate those statistical properties. Models based on the theory also predict and explain a wide range other criminal career features and observations.
For each gender there are three categories of standard list offenders and a further category of non-standard list trivial offenders. The ‘trivial’ category of offenders commits a large amount of rather trivial crime (including a large proportion of all motoring offences). They rarely get sentenced to custody but may be very prolific. There may also be substantial overlap between the trivial category and the high-risk/low-rate category of Chapters 2 and 3, as they occasionally commit less serious standard list offences. At the other extreme there is a high-risk/high-rate category of offenders who on average accumulate five or six convictions, for serious offences, and on average are convicted about once a year during their criminal careers. Some high-risk offenders will accrue large numbers of convictions over many years. Finally there is a low-risk/low-rate category of offenders who typically are convicted only once or twice but for relatively serious offences. For low-risk offenders with more than one conviction, these would be on average four to five years apart.
The high-risk/high-rate and low-risk/low-rate categories have an average criminal career length, from first to last convictions, of about five years whereas the high-risk/low-rate and trivial categories have an average career length of 10 to 11 years. For all categories the range of career lengths (as measured by convictions) is from 0, a single offence/conviction, to the whole active life span but with a rapidly diminishing likelihood of careers longer than the average (see the age–crime curves in Chapters 3 and 4). Most offenders will also have been actively committing crimes prior to their first conviction.
Although the actual numbers (parameters) describing the categories differ, the nature of their behaviour as seen in terms of convictions is qualitatively the same. They are convicted at a constant Poisson rate (committing offences at random) and a constant proportion stop offending after each conviction. Criminality (the proportion of the general population sustaining one or more standard list convictions in their lifetime) was substantially constant for all the birth cohorts, as was the proportions in each of the risk/rate categories. There is much greater uncertainty concerning the trivial offender category. Their offences are technically crimes (regulatory offences, minor public disorder offences, motoring offences, etc) but many in the general population would not consider them as criminals and we have not included them, apart from the subcategory who also commit standard list offences, in our main criminal categories.
The theory and models are ‘large scale’ in two main respects. The theory does not specifically consider the social or psychological causes of the offending behaviour, although in Chapter 6 we identified significant correlations between our risk categories and psychological assessment scores. Also our group structure is not necessarily complete, in that there are almost certainly special categories and subcategories of offenders. Examples of special categories include those whose offending is induced by mental illness or psychological abnormality (eg paranoid schizophrenics, paedophile sex offenders, serial rapists, and mass murderers) who are all very different from offenders in general but thankfully relatively rare. Another such group probably consists of around half of life-sentence prisoners who have very low probabilities of recidivism. However, in the main, offenders are versatile and commit a wide range of offences over a criminal career. Selecting subsets of offenders leads to some variation in the parameters (proportions in the categories, recidivism probabilities and rates of conviction) although this variation is generally much less than the differences in parameters between the main categories.
With the exception of the mentally ill, our theory assumes that for individuals crime is a lifestyle choice (see also Walters 2006; West and Farrington 1977) and at each offence the individual makes a decision. Without consequences such behaviour is unlikely to change. Being caught is an essential element in desistance, and the greater the certainty of capture the greater the deterrent effect. However, being caught is not, in itself, sufficient. For those destined to join our criminal categories, offenders need to be made aware of the extent to which their behaviour is unacceptable to society at large. In our view it is the criminal justice system process 1 that appears to have the most influence on the life-choice decision to no longer commit crime, or at least to modify behaviour to avoid serious breaches of the law. The penalties imposed on conviction may well serve several other penological objectives, but there is no substantive evidence from our analysis that the particular sentence influences the decision to desist from further offending.
Crime prevention is of course a major vehicle for the reduction of crime, but it is not one that our analysis was concerned with (see Farrington and Welsh 2007; Welsh and Farrington 2009, 2012). Although reducing opportunities for crime could well impact on the parameters of our model it is not clear what that impact might be. We therefore concentrate here on policies which relate to (potential/actual) offenders and the criminal justice system: police, courts and corrections. Our theory suggests that overall crime reduction objectives might be achieved by policies to:
reduce the likelihood of individuals entering the criminal categories, by means of early intervention programmes for children at risk;
improve the effectiveness of interventions in early criminal careers, for example by informal actions including the handling of antisocial behaviour by society at large as well as the police, and improvements in the effectiveness of the warnings, reprimands and cautions issued by the police;
increase the efficiency of the criminal justice system. Doubling the probability of conviction given a crime halves the total amount of crime committed by known offenders in the long term;
make those small reductions in recidivism, mainly in the high recidivism group of offenders, which might be expected from some kinds of offender treatment programmes such as Enhanced Thinking Skills. These will have a disproportionate impact in reducing crime.
In the Cambridge Study in Delinquent Development, Farrington (2007; see also Farrington, Coid, and West 2009) identified a number of factors from childhood which could be used to predict later offending and conviction. In a systematic review of evaluations of early intervention programmes, Farrington and Welsh (2007) concluded that many programmes were effective in reducing childhood antisocial behaviour and later delinquency. Beelmann and Raabe (2009), from a review of articles and a meta-analyses of childhood interventions, found that prevention measures addressing high-risk categories produced higher effect sizes than universal strategies.
Common sense, the evidence base, and our theory suggests, that preventing the onset of offending would be a very effective strategy for reducing crime and early intervention would seem to be the most likely option for long-term criminality reduction (see Farrington 2007; Farrington and Welsh 2007). These programmes are not necessarily focused on crime but rather on community cohesion, improving parenting skills, the provision of facilities for families and improving social responsibility and good citizenship. All of these interventions are likely to reduce criminal involvement and have been in the political rhetoric for many years. There have been programmes based on these ideas but, in England and Wales over the time period covered by the Offenders Index (OI) cohorts, not on a scale that would show up in our analysis.
More recently, such programmes have been rolled out on a much larger scale, and in particular Sure Start, a comprehensive community based programme of early intervention and family support. Sure Start was initially aimed at children under school age and their families, particularly in less advantaged areas, and it has shown some encouraging results (Melhuish et al 2008). By 2011 there were more than 3,600 Sure Start Children’s Centres in England. Although the effectiveness of the Sure Start local programmes and children’s centres is being assessed on many of the anticipated outcomes, the impact on crime and criminality will not become apparent until about 2020–2030 and even then it may be difficult to separate the effects of Sure Start from other interventions and policies.
The next opportunity in the life course would be to deal more effectively with early-onset offending. During the period of the OI cohorts there has been a progressive tendency to divert offenders away from the criminal justice system and a reluctance to prosecute and impose formal sanctions on even quite serious young offenders, particularly for pre- and early teenagers.
In the 1953 cohort some 237 offenders, approximately 2 per cent, were convicted at age 10, but this number reduced in successive birth cohorts to 196, 123, 71, and 39 in the 1958, 1963, 1968, and 1973 cohorts respectively. The pattern is somewhat different if we consider those convicted up to the mid-teens. By age 16 about 20 per cent of offenders in the 1953 cohort had one or more convictions. This proportion reduced only slightly in the 1958 and 1963 cohorts to about 19 per cent of the estimated lifetime offender cohort size. But the proportion of under-16s reduced significantly to 14 per cent for the 1968 cohort and only 7 per cent for the 1973 cohort. Our estimates of criminality (see Table 2.6) remain essentially constant for all the cohorts, with an average of 23 per cent. The 1973 cohort estimate at 20.4 per cent seems low, but because the observation period ended at age 20 the criminality estimate is subject to greater censoring. However, the criminality estimate from the 1997 sentencing sample is also about 20 per cent, and should not be so error prone, suggesting that the 1973 cohort estimate might not be too low.
If correct, this reduction in criminality might be seen as evidence of the success of the police cautioning policies. But if the reduction is due only to those offenders who would have desisted after conviction now desisting after caution, the success is limited to savings in court time and the non-criminalization of young offenders. To effect a reduction in crime would require recidivism after cautioning to be lower than after conviction. However, after controlling for prior differences between cautioned and convicted youth, Farrington and Bennett (1981) and Mott (1983) found little difference in reoffending between them. Criminality estimates in successive cohorts from 1953 to 1968 suggest that, despite the progressively increased use of cautions up to age 16, criminality remained comparable to, or above, the 1953 cohort level. This in turn suggests that convictions were merely delayed with a commensurate increase in pre-conviction crimes (see also Farrington and Maughan 1999). The 1973 cohort and the 1997 sentencing sample, however, do seem to indicate a reduction in criminality. The authors have insufficient information on the effectiveness of cautioning in more recent years to resolve this issue but this is clearly an important area for more rigorous evaluation.
We define the efficiency of conviction as the ratio of convictions to recorded offences. We assume that offenders, in particular our high-risk offenders, commit many more offences than they are convicted of. For example, Farrington et al (2006) in the Cambridge Study found that there were 39 self-reported crimes for every conviction occasion. Based on a comparison between victim survey and conviction data, Farrington and Jolliffe (2004) concluded that, in England and Wales in 1999, there were only seven convictions per 1,000 burglars, 17 convictions per 1,000 vehicle thieves, six convictions per 1,000 robbers, and 25 convictions per 1,000 assaulters. Our theory suggests that conviction is the trigger for desistance. Therefore if offenders are convicted for more of their offences (increasing λ for convictions but not for offences) and the reconviction probability remains the same, then crime will be reduced.
In the cohort data there is a tendency for higher recidivism probabilities after shorter inter-conviction times. However this does not necessarily imply a causal relationship. In the 1953 cohort, following inter-conviction times of less than six months, the recidivism probability was 0.86, marginally greater than the high-risk value of 0.84. But this apparently higher recidivism probability could well be caused by the pseudo-reconviction problem, in which the pseudo-reconviction is for an offence committed prior to the previous conviction. For inter-conviction times between six months and a year the reconviction probability was 0.84, the same as for the high-risk category. For inter-conviction times greater than a year the probability decreased steadily to about 0.30 at 10 years, reflecting the increasing proportion of low-risk/low-rate offenders among those with longer inter-conviction times.
There is also some evidence that high-rate offenders in particular, tend to slow down between the penultimate and last conviction. This slowing down is unlikely to be an ageing effect as was demonstrated in Chapter 5 (Figures 5.4 and 5.5), where the inter-conviction survival time curves were parallel and parameter values were substantially constant over increasing ranges of appearance numbers and hence increasing age. It would also be counter-intuitive to believe that being caught and convicted more quickly would encourage recidivism rather than desistance. Policies to improve the efficiency of conviction should therefore hasten the end of the criminal career and hence reduce crime.
Goldblatt and Lewis (1998) summarized substantial evidence that various kinds of treatment programmes can reduce recidivism. Such programmes were not widely used in the period from 1963 until 1997, during which time the bulk of data for the research leading to our theories and models was collected. If such programmes are effective in practice they will lead to decreases in the recidivism parameters and would have a significant long-term effect on crime.
Criminal Justice: The Way Ahead (Home Office 2001) outlined many policies aimed at reducing crime and improving the operation of the criminal justice system. Among these policies was a commitment to target persistent offenders. The Prolific and other Priority Offender (PPO) programme was implemented in England and Wales on 6 September 2004. In each area PPOs were identified using criteria reflecting local concerns coupled with the offender’s criminal history. PPOs were then prioritized by the police, prisons, probation, and other local agencies, with a view to reducing their offending. The three objectives of the programme were to ‘Prevent and Deter’, ‘Catch and Convict’, and ‘Rehabilitate and Resettle’. The ‘catch and convict’ element of the PPO programme was intended to improve the efficiency of conviction for these offenders and hence shorten the criminal careers of the PPOs. In the short term, PPOs were also given rehabilitation training and post sentence support to aid resettlement. An important element of the programme was close coordination between all the agencies involved.
Dawson and Cuppleditch (2007) reported on an evaluation of the Prolific and other Priority Offender programme. The evaluation comprised three components: a reconviction analysis, offender interviews, and staff interviews. Although the interview components provided interesting and informative anecdotal evidence of the impact of the programme, our particular interest here is the reconviction analysis and how the theories developed in this book clarify and help to explain the results.
Offenders identified as PPOs were tracked on a computerized system known as JTrack. This system enabled the researchers to extract conviction information on all PPOs identified in the first two months of the programme (the PPO cohort). A monthly count of cohort convictions was made for the three and a half years prior to and one year and nine months after September 2004 (see Figure 8.1). 2
Monthly conviction count for PPO cohort and control group, before and after PPO selection
Source: Dawson and Cuppleditch (2007). Note: The overlaid curves are predictions derived from our theory and models.
Figure 8.1 shows an increasing trend in convictions up to the start of the PPO selection period followed by a 30 per cent reduction in the conviction rate in the subsequent three months and a steady decline over the remaining part of the observation period. Dawson and Cuppleditch (2007 p 7) interpreted this as ‘a steady rise in their criminal behaviour until they commence the PPO programme at which point there is a sharp decrease followed by a period of steady decline’. The inference was that the PPO programme halted the worsening criminal behaviour, reducing it significantly and setting in train a steady continuing reduction in convictions.
Although not specifically stated in the evaluation report, we suspect that selection as a PPO was triggered by a conviction in, or very close to, the selection period. The PPO sample size was 7,400 and the reported monthly conviction rate, in the selection period, was in the region of 3,500. However, for a sample of 7,400 offenders convicted at time t = 0, using our rate model with λ = 1.05 court appearances per annum, we would expect 1,236 offenders to have sustained at least one principal conviction in the previous month. The figure of 3,500 convictions suggests that Dawson and Cuppleditch (2007) counted all convictions rather than just court appearances.
We would also have expected a significant peak in the conviction rate at t = 0, in the PPO sample case, spread over two months or so. The high conviction rate is consistent with the PPOs being members of our high-risk/high-rate offender group and we can calculate the distribution of previous convictions prior to 6 September 2004 (conditioned on conviction in the approximately two month selection period): the solid smooth line on the graph shows the expected distribution of conviction rates up to 6 September 2004. Because of the conditioning of the sample and stability in the number of active offenders over time we would expect the build up of monthly convictions to follow the mirror image of the residual career length profile. The solid line up to 6 September shows the expected profile assuming p = 0.84 and λ = 1.05 court appearances per annum. On conviction in the selection period we would expect 16 per cent of the PPOs to desist and, had there been no intervention, the remainder to continue offending and desisting in line with the residual career length profile 3 as shown by the continuation of the solid line. The apparent rise in criminal behaviour is an artefact of the conditioning 4 of the PPO cohort sample (conviction in the selection period), which is in fact comprised of high-risk/high-rate offenders who are convicted at a constant Poisson rate λ.
In the PPO data the conviction rate reduces more slowly than predicted, over the first three months from September to December, but then continues to fall rapidly for a further two months. We would suggest that the slower than predicted initial decline is the result of the catch and convict element in the PPO programme which was intended to speed up the reconviction process with the aim of reducing the active population more quickly. This appears to have happened as after four months the conviction rate (which is proportional to the active population) is well below the prediction, continuing the initial trend. From February/March 2005 through to May/June 2006, the slope of the steady decline is steeper than the model predicts suggesting that, following involvement in the PPO programme, the residual career length is shorter. This can be explained either as an increase in the conviction rate λ as a result of catch and convict or as a reduction in the recidivism probability due to rehabilitate and resettle or a combination of the two. In any event the PPO programme appears to have been successful and the resulting conviction rate profile has a rational explanation in terms of the theories and models described in this book.
As part of the evaluation, Dawson and Cuppleditch (2007) set up a counterfactual sample of offenders based on Propensity Score Matching (PSM) to act as a control group. Offenders were matched case by case with the PPO cohort sample, on several characteristics including gender, age, and detailed criminal history. Judging from their description, the matching procedures were rigorous. The monthly conviction rates of the PSM sample are plotted as the irregular line in Figure 8.1. However the evaluators were perplexed by the profile obtained. The PSM conviction rate profile was very similar to the PPO rates from three and a half years to one year before the start of the PPO programme at which point the conviction rate started to decline consistently through to the end of the observation period. These results were described as ‘unexpected’.
In our view the problem has arisen due to the conditioning of the counterfactual sample. There is no mention of date of conviction in the matching criteria. From the graph we suspect that for the PSM offenders their last convictions prior to the start of the PPO selection period occurred evenly over the previous 12 months and that their inclusion in the sample was not conditional on a conviction during the PPO selection period. If that was the case, our theory would predict that each month, from the start of the PSM selection period, some 16 per cent of those convicted (recidivism = 0.84) would not be convicted again, representing some 6 per cent of the active offenders in the PSM sample. The number of active offenders would therefore be reduced by 6 per cent each month over the period from September 2003 up to, judging from the graph, possibly January 2005. The solid line, overlaying the PSM conviction profile, in Figure 8.1 shows the expected conviction profile over that period based on the above. The profile beyond January 2005 runs approximately parallel to the PPO predicted profile suggesting conformity to the residual career length distribution with λ and recidivism probability as for our high-risk/high-rate group.
The rehabilitate and resettle component of the PPO programme contained some elements of offender treatment and our analysis above suggests that recidivism may have been reduced significantly. That programme contained a variety of interventions including drug treatment, close supervision, and assistance with resettlement and it is not clear which element or combination of elements were responsible for the observed results or indeed whether the reduction in recidivism would persist beyond the observation period. Notwithstanding these reservations the indications are encouraging and the reduction in the conviction rate and the apparent shortening of the residual career length suggest a permanent change. A long term, up to 10 years, follow up of the PPO cohort and a more detailed analysis of recidivism and inter-conviction times would be needed to verify the effectiveness of the PPO programme.
The PPO analysis above has provided a practical example of the way our theory can be used to explain observations which otherwise can be easily misinterpreted. In the PPO programme example the theory provided a plausible counterfactual and was able to explain why the PSM sample produced an unhelpful control, which in the event is entirely consistent if the conditioning of the data is taken into account. In any evaluation research it is very important to have a clear understanding of the processes involved and a well founded expectation against which to measure the outcome. In this section we will outline what we might expect from policy interventions aimed at influencing criminality, recidivism, rate of offending or rate of conviction.
Perhaps the most important implication of our theory is that the criminal justice system does control crime. The vast majority of offenders cease offending because of the activities of the CJS. This means that changes to the CJS may lead to reduced or increased levels of crime. However, perhaps the most effective way to reduce crime is to reduce criminality, ie to stop individuals becoming criminal in the first place. For each potential high-risk offender diverted completely from a life of crime by early intervention, an average of between 19 and 31 recorded crimes would be averted. For potential low-risk offenders an average of between 4 and 7 recorded crimes would be averted. These estimates were based on our theory and ‘Offences Brought To Justice’ statistics (Ministry of Justice 2010).
For high-risk offenders the above estimates are probably very conservative; if unrecorded and unreported crimes are included the averages could be very much greater. Overall approximately 5 per cent of the population fall into our high-risk categories and in Chapter 6 we showed that a significant proportion of the high-risk category could be identified from psychological characteristics. If early interventions were focused on these most vulnerable individuals a disproportionate 5 reduction in crime could be achieved. Low-risk potential offenders would be more difficult to target as they appear to be similar, psychologically, to non-offenders and the bulk of non-standard list trivial offenders. Reducing the size of an offender category would result in a pro rata reduction in the crime committed by offenders in that category.
Policies aimed at reducing recidivism can also have a significant and disproportionate impact on crime. Again this is particularly true for high-risk offenders. In the Appendix we show that overall, at any given time, about 55 per cent of crime is committed by offenders prior to their first conviction. Thus reducing recidivism only impacts on 45 per cent of future crime (see Table A.1). We also show that crime is inversely proportional to the probability of desistance: thus reducing the recidivism probability for the high-risk group by 10 per cent, from 0.84 to 0.76, would result in a 34 per cent reduction in their future offending and, on reaching the steady state, a 14 per cent reduction in overall crime. In making these estimates we are assuming that offender treatment programmes are given to all high-risk offenders and that a 10 per cent recidivism reduction can be maintained. In practice, the results might be more modest.
The results would certainly be modest for low-risk offenders both because only 33 per cent, at most, of their crime would be affected by reducing their recidivism and because programmes aimed at first offenders in this group would be given to individuals of whom 60 per cent had already decided not to offend again. A 10 per cent reduction in the recidivism probability of low-risk offenders, from 0.30 to 0.27, would result in only a 0.5 per cent reduction in overall crime.
Policies aimed at reducing opportunities for crime may well reduce crime overall but there is some possibility that crime may simply be displaced to areas where opportunity reduction measures are not implemented or to other types of crime. On the other hand there is also evidence that the benefits of crime reduction measures may be diffused to neighbouring areas (Painter and Farrington 1999). Making crime more difficult may also reduce the frequency of offending but is unlikely to cause desistance. Thus, although λ for offending might reduce as a result of opportunity reduction, λ for conviction and recidivism will almost certainly be unaffected. This is because, as we have seen, offenders are versatile and according to our theory they will continue to offend until convicted at which point a constant proportion will desist.
Also, for opportunity reduction policies to impact on overall crime, we would require reductions in the offending rate to have no effect on the rate of conviction. In particular if reducing the offending rate simply increases the inter-conviction time then criminal careers would also be lengthened and overall crime would remain the same. Conversely, over the observation period of the OI cohorts there was a consistent increase in recorded crime through to the early 1990s but both criminality and conviction rates remained substantially constant. This suggests that, within reason, the two rates are only loosely related, possibly because of the dilution of police detection efforts or increases in the numbers of crimes dealt with at individual court appearances. Opportunity reduction by target hardening and surveillance is often very effective at a local level but the impact on overall crime is difficult to assess.
Policies aimed at increasing the probability of conviction, ie increasing the probability that an offence will result in the offender being caught and formally dealt with, will tend to increase λ for convictions. In the Appendix we show that the average residual career length is inversely proportional to both λ for convictions and the probability of desistance. Thus increasing λ for convictions will have the effect of shortening the criminal career provided that the recidivism probability does not increase. A 10 per cent reduction in average inter-conviction time (increasing λ for convictions from 0.86 to 0.956 convictions per annum) for high-risk offenders would result in a 2.25 per cent reduction in overall crime.
Over the first decade of the twenty-first century a new measure of overall CJS performance was introduced 6 in which offences brought to justice (OBTJ) were counted and reported quarterly in Home Office/Ministry of Justice statistical bulletins. An offence is brought to justice if it results in a conviction, caution, fixed-penalty, or is taken into consideration in the determination of sentence. The effectiveness of increasing the OBTJ count, as a crime reducing policy, will depend on whether λ for convictions is also increased. If more offences are brought to justice as offences taken into consideration, overall crime would be unaffected. If additional offences result in a conviction, through improved policing and prosecution procedures, then overall crime is much more likely to be reduced.
The impact on overall crime of additional crimes brought to justice through police cautions will depend on the effectiveness of cautioning and any interventions which may accompany the caution (‘cautioning plus’). The data available to the authors did not include any information on the subsequent criminal histories of cautioned offenders. An analysis along the lines of that in Chapters 2 and 3 of data from the Police National Computer, which includes cautioning information, would provide a very useful insight into the effectiveness of cautioning and its ability to divert offenders away from the courts. Fixed penalties are not issued for the more serious (standard list) crimes analysed here and are therefore unlikely to impact on the overall level of these crimes.
During the 1990s the political rhetoric in the UK, especially of Home Secretary Michael Howard, assured us that ‘prison works’. Courts were encouraged to deal severely with offenders with the result that from about 1993 the prison population started to increase by about 8 per cent per year. The average prison population increased from 45,600 in 1993 to 66,300 in 2001 and over the same period according to the British Crime Survey crime reduced by around 20 per cent. Proponents of policies advocating the increased use of prison claim that the fall in crime was caused by the incapacitation of offenders and possibly the deterrent effect of the more severe punishment. In Chapter 5, in our discussion of various criminal career theories, we concluded that there was no support in the OI data for fixed career length or age-based desistance theories. Any long-term incapacitation effect relies on such theories being true. We return to this issue in greater detail in the Appendix where we show that the size of the active offender population, in the steady state, is independent of the size of the prison population. However, under the above conditions of year on year increases in the prison population we would predict a reduction in crime, but only of the order of 1.5 per cent.
Our theory also predicted a fall in crime of 12.3 per cent due to demographic changes over the six-year period up to the year 2000. We cannot explain all of the reduction in BCS crime but our theory does explain well over half of it. The observed reduction in crime also suggests that more severe punishment increases deterrence but, from our analysis, those released from prison do not appear to have been deterred any more than those sentenced to community penalties. It might of course be that general deterrence is increased but it would be very difficult to establish a causal link or even to quantify deterrence at all. There may also be ‘feedback’ or ‘fashion’ effects whereby decreased offending due to a reduced number of active offenders leads to a reduced propensity to offend in those still active. Such a ‘non-linear’ effect was considered by Marris and Volterra Consulting (2003).
We consider here some of the objections to the arguments we have presented so far:
Aren’t there large numbers of offenders who are never convicted?
There are, no doubt, some offenders who never get convicted: one-off offenders or those who are effectively deterred from further offending by informal action and offenders specializing in types of crime where there are very low reporting or detection levels. However, in the major part of the analysis presented here, we are concerned with non-trivial volume crime and serious offending. Even for very low probabilities of getting caught for an individual offence, the vast majority of offenders who contribute importantly to the total level of crime, because of their repetitive offending, will be convicted at some point and thus fall within the scope of this analysis. The formulae in the Appendix show that about half of crime is committed by those who have yet to be convicted (see also Farrington et al 2006).
There are fewer offenders aged 30 than aged 17. How can you say that people don’t grow out of crime?
Older offenders will, on average, have been convicted more often than younger ones. As, under our theory, conviction triggers desistance, offenders are more likely to have given up lives of crime by age 30 than by age 17. Therefore there are fewer 30-year-old active offenders than 17-year-old ones. From the cohort data we know that almost half of offenders sustain their first conviction after the age of 20 and half of all convictions are of offenders over the age of 25. High-risk/high-rate (the most prolific) offenders dominate among juveniles and young adults. But, because they are convicted more often, their numbers diminish relatively quickly and less than one in ten of those active at age 17 will still be active at age 34. Of the high-risk/low-rate offenders active at age 17, 10 per cent will still be active well after retirement age and our low-risk group of offenders account for most of the convictions at ages 25 and above. In our view, it seems perverse to believe that only the most prolific offenders grow out of crime whereas age has relatively little influence on less criminal offenders. The constant probability of recidivism after conviction, independent of age, has convinced us that age is not a causal factor in desistance.
When someone is in prison they cannot commit a crime in the community. This must reduce the total amount of crime, mustn’t it?
For time spent in prison to reduce overall crime the sentence must either reduce the probability of recidivism or shorten the residual criminal career. Direct evidence from the cohort analysis suggests that neither outcome is achieved. After controlling for criminal history, the probability of recidivism is not significantly different after custody compared to community sentences for similar offence seriousness. There is also no significant difference in average reconviction times for these recidivists, after either custody or community penalties, who simply re-join the active offender pool with their expected residual criminal career unchanged. Their crimes are postponed rather than averted. As soon as we release recidivists from prison they carry on committing their delayed crimes at the same rate as they would have done if given a non-custodial sentence. Unless we let the prison population increase indefinitely, for each active offender entering prison one is released and the overall level of crime stays the same. It is like putting a dam across a river. While the dam fills up there is no flow downstream but when the dam is full the flow downstream is exactly the same as it was before the dam was built. The effect on crime, of simply imprisoning offenders, is the same.
Conviction rates depend on the operation of the CJS. Aren’t your results really just telling us about the capacity of the CJS to process people?
It is true that the parameters describing the categories are generated by the interaction of the CJS with the offending behaviour of offenders in the categories. As the theory is based on convictions it could not be otherwise. However, it can be seen in two ways that the parameters must reflect the underlying criminal behaviour. In the first place each category has very distinct parameters. If we were just seeing the processing of the CJS, we would not expect a distinct category structure. Either all the categories should have similar parameters or there should be a continuous range. Secondly, if we were seeing artefacts of the CJS we would expect to see the number of convictions follow the capacity of the CJS to process cases. What we actually see is the annual number of convictions rising and falling in line with demographics, as predicted by the models based on our theory. The effects of CJS capacity limitations may, however, explain why the trend in the number of prosecutions instigated does not closely follow that of convictions.
The sequence of events: arrest, charge and court appearances resulting in conviction and sentencing.
The data points in Figure 8.1 were taken from the published graph, Figure C in Dawson and Cuppleditch (2007, p 7), and may therefore not be exact as values were estimated to the nearest 50.
See discussions on incapacitation in the Appendix.
Had the sample been a random selection of active high risk/high rate offenders the conviction rate distribution would have been uniform over the observation period, prior to 6 September 2004, at about 600 convictions per month.
A well-targeted intervention could be up to 25 times more effective at reducing crime than one randomly allocated in the population at large.
See Criminal Justice: The Way Ahead (Home Office 2001, p 21) and Narrowing the Justice Gap (Crown Prosecution Service 2002).