You are here: Home Publications Publications archive (pre-2011) 1997 Sentencing policy and guidance 3. The Rationales and Goals of Sentencing (1/2)

Archived publications

You are currently in the publications archive, and may not be accessing the most up to date information. The more recent publications can be found in the publications A - Z.


3. The Rationales and Goals of Sentencing (1/2)

There are some often cited rationales and goals for sentencing, namely: just deserts which in this chapter will cover both retribution and denunciation; deterrence; incapacitation; rehabilitation; and restitution.

Just deserts stands out from the others because it is a rationale for the imposition of sentences and the amount of sentence according to the offence itself and nothing else (particularly not future offending). Where retribution differs from denunciation is that the former is totally offender orientated and non-utilitarian and does not seek a wider social goal through sentencing, while denunciation acts principally for the benefit of the general population by publicly reinforcing the boundaries between acceptable and unacceptable behaviour. Denunciation is often linked to retribution because effective denunciation needs to be able to reflect different degrees of offence seriousness (see Young 1982). Deterrence, incapacitation, and rehabilitation on the other hand pursue crime reduction goals through sentencing. Restitution is primarily aimed at making redress to victims of crime. These systems justify sentencing on the basis of the end goals it is supposed to achieve, and thus they are often labelled as utilitarian. Just deserts simply provides a justification for sentencing on the premise that wrongdoing deserves punishment in proportion to the wrongdoing.

These rationales and goals are discussed in this chapter. The discussion focuses on the insight that theories of sentencing can provide on how society can best be served by its sentencing system, and what it can realistically expect that system to achieve. It also raises issues as to the extent that sentencing decisions should be guided by one or other, or some combination of, these theories. 

3.1 Just Deserts (Retribution and Denunciation) and Proportionality

Under a desert-based model the sentence is a punishment for wrongdoing. As such the focus is on the individual offender. Punishment is for the guilty only, and only for the offence(s) for which guilt is established. Just deserts sentencing does not attempt to predict or prevent future offending by the individual or other members of society.

The harm done by the offence is replied to by the state's imposition of some form of pain and deprivation on the offender. In New Zealand this involves some form of restriction on liberty (with imprisonment as the most extreme form) or a financial penalty, rather than the infliction of physical pain (e.g. corporal punishment).

The 'just deserts' approach can cover both a retributive element in that the sentence is deserved by virtue of the wrongdoing, and a denunciatory one whereby sentence is deserved because the wrongdoing must be publicly condemned. Proportionality is observed in both cases through the requirement that the sentence be just.

3.1.1 Why punish?

Just deserts is founded on the belief in punishment, so the first question to be addressed is how this punishment philosophy can be justified. There are two main justifications: punishment as a means of restoring the balance or moral equilibrium disturbed by the offence (retribution); and punishment as public censure (denunciation).

Punishment may be seen in terms of a symbolic restoration of the balance which was upset by the offence. This is sometimes explained in terms of a forfeiture of rights (Lacey 1988, p22). Through the act of committing an offence an offender violates his or her obligation to the state and its citizens to obey society's norms and the state is justified in depriving the offender of civil rights so that they do not obtain an unfair advantage over those who are law-abiding or cause the latter to be disadvantaged. A second explanation is one in terms of benefits and burdens. It is argued that everyone receives benefit from a law-abiding society, so long as they accept the burden of self restraint required to remain within the law. When someone breaks the law, they impose on others and get an unfair advantage; an extra benefit which is not available to law-abiding citizens. Punishment is designed to offset the unjustly obtained benefit and restore that balance or moral equilibrium. For this reason, the amount of punishment should be proportionate to the benefit unlawfully gained. Both these versions of desert theory exist within the social contract tradition in political philosophy, involving reciprocal political obligations of the state and its citizens.

Neither of the above justifications give a clear indication of the actual punishment appropriate in particular cases. What set of rights should be forfeited for violating certain norms through assault or burglary, for example? What punishment will remove an unfair advantage gained through an offence? A problem inherent in the latter question is that it is not always clear that an offender has actually gained an advantage. For property crimes the material benefit will be fairly obvious. This is not so clearly the case for many violent crimes and other offences such as vandalism and destruction. Any benefit to the offender is likely to be momentary: an opportunity to vent anger or frustration or exert power. Nevertheless, in a pattern of offending, such as domestic abuse, the offender can be seen to gain longer-term illicit benefits in terms of power and authority over others. It may, however, be straining credibility to argue that those who do not commit crimes really perceive this as a significant burden.

The other main theory sees punishment as censure, denunciation, or an expression of society's disapproval of the particular conduct associated with the offence. Denunciation's primary target is not the offender but the general population. Punishment is seen as positively reinforcing people's inhibitions against offending by encouraging respect for, and compliance with, the law and reaffirming the seriousness of society's rules. Through the sentence the offender will be made aware of society's attitude to his or her conduct. This is the view favoured by a number of Scandinavian theorists, and is similar to Hall's description of denunciation as emphasising:

that the criminal law is an educative and a cohesive force in the community, and, through the public nature of punishment of crime, an important symbolic statement as to the extent of society's indignation and condemnation of certain conduct is thereby expressed.

(Hall 1993-97, B/64)

The English Court of Criminal Appeal has stated that the sentencing function of the courts is the means by which society "shows[s] its abhorrence of particular types of crime" (Court of Criminal Appeal, England, (R v Sargeant (1974) 60 Cr App R 74, 77 per Lawton CJ).

American academic Andrew von Hirsch argues that censure is additionally justified as general crime prevention. The expressive function of sentencing reminds people of the negative consequences of offending, thereby "keeping predatory behaviour within tolerable limits" (von Hirsch 1992(a), p74). The New Zealand courts have advocated proportionate sentencing as part of general crime prevention. The Court of Appeal has stated that:

If a Court is weakly merciful, and does not impose a sentence commensurate with the seriousness of the crime it fails in its duty to see that the sentences are such as to operate as a powerful factor to prevent the commission of such offences.

(R v Radich [1954] NZLR 86,87 (CA))

These views of general crime prevention through punishment reinforcing society's norms are to be distinguished from views that would make deterrence (people fearing the consequences of offending) the primary justification for the imposition and amount of sentence, as discussed below in section 3.2.

Von Hirsch sees censure and disapproval as inherent in the imposition of sentence, and maintains that censure also provides a useful means of measuring the appropriate amount of punishment. He argues that:

If punishment is seen as an expression of blame for reprehensible conduct then the quantum of punishment should depend on how reprehensible the conduct is.

(Von Hirsch 1992(a), p68)

In New Zealand, the courts have emphasised the notion of sentencing as censure, particularly in cases of sexual offending, and in cases of violence against children. In one rape case, (R v Puru [1984] 1 NZLR 254) the Court of Appeal stated (per Woodhouse P) that "Society is entitled to exact a severe penalty from the offender so as to mark its condemnation of his conduct." In an incest case, Grieg J referred to "the community's abhorrence of incest", and stated that the offender should be imprisoned "for a period which marks the community's view" (Y v Police (High Court Timaru AP62/91, 9 Oct 1991) 14 TCL 42/9).

3.1.2 How much punishment?

If punishment is to restore the balance of benefits and burdens, just deserts dictates that it should be proportionate to the disruption caused by the offence. If it is an expression of blame for reprehensible conduct it should be proportionate to the reprehensiveness of the conduct. The guiding principle that the severity of the punishment should be in proportion to the gravity of the crime implies a hierarchy of offences and sentences, and that like offences should attract like sentences.

The proportionality principle has wide currency in the community. It appeals to common-sense notions of equity which are also evident with regard to sanctions which exist outside the criminal justice system. Few would feel comfortable with the employee who is ten minutes late being fired, while another who consistently abuses sick leave provisions is simply given an oral warning.

Support for proportionality was evident in a 1994 public opinion survey commissioned by the Department of Justice. Participants expressed concern at perceived inconsistency in sentencing for similar offences, and that sentences awarded did not seem to match the public's view of the hierarchy of offence seriousness. This was particularly expressed with respect to the relative sentencing of burglary and white collar crime, where a common perception was that the sentences for the former were greater than for the latter, despite white collar crime involving far greater monetary amounts.

Advocates of just deserts stress its value as a constraint on punishment, because it allows punishment only of the guilty and only in proportion to the seriousness of the offence(s) committed. This notion of a rational, justice-based restraint on punitive impulses has been stressed by the New Zealand Court of Appeal. In R v Puru [1984] 1 NZLR 248, 250, Woodhouse P stated that:

[The] judicial obligation is to ensure that the punishment [the courts] impose in the name of the community is itself a civilised reaction, determined not on impulse or emotion but in terms of justice and deliberations.

Similarly in R v Hiha 1987 2 NZLR 119,12, the Court reduced a sentence for rape and sodomy with the observation that it "is incumbent on the Courts to resist the temptation of being led by indignation to inflate punishments unnecessarily beyond an already condign and severe level."

The rights-respecting focus of just deserts means that the offender is treated as an autonomous individual. The offender's culpability, and harm done or risked, are the determining factors in the sentencing decision. This contrasts with the utilitarian goals of sentencing, under which the sentence is determined by what is necessary to achieve a greater social good. This lessens the focus on the offender's rights and, in the case of general deterrence at least, the offender becomes part of the means to a greater social good.

3.1.3 How is the sentencing scale set?

The real implications of a desert philosophy will of course depend on how the penalty structure is set. This is a subjective process, and one society's view of desert may vary considerably from another's, as will one person's from another's, when it comes to establishing the scale. So how does a society decide on the level of sentence proportionate to the offence?

Von Hirsch (1992 (a)) attempts to answer this question by distinguishing between ordinal proportionality and cardinal proportionality. Ordinal proportionality concerns the way offences are ranked in relation to each other, moving from the least serious to the most serious. Parity requires that offences of equal seriousness are ranked equally. The seriousness of offences is determined by the harm done or risked, and the culpability of the offender: factors of intent, motive, and circumstance which determine the extent to which the offender should be held accountable for the act. (This is presumably for a typical instance of the offence, which points to a problem with this approach, namely that it is difficult to know what a 'typical instance' is, particularly if the culpability of the offender is to be taken into account.)

Von Hirsch concedes that such offence ranking exercises are time-consuming, but the experience of sentencing commissions which have established detailed offence and penalty rankings shows that it is not impossible. This does not acknowledge that sentencing commissions have made little or no attempt to take into account typical levels of culpability when ranking offences. Culpability is rather a separate issue for the courts to take into account in deciding whether or not to depart from the guidelines.

In New Zealand a general ranking of penalties and offences already exists in the form of the maximum penalty structure of the Crimes Act, although this may be out of kilter in some respects because of the ad hoc way in which it has been changed over time. It is also a ranking in respect of the worst possible instance of each offence and many offences are very broadly defined.

Cardinal proportionality concerns the upper and lower anchoring points for each rank of the offence seriousness scale. These set the limits on the severity or leniency of the sanctions appropriate to each offence seriousness rank. Von Hirsch concedes that in this aspect, desert can only act as a limiting principle on the upper and lower limits of the sentencing scale which will already be fairly well established by political considerations and conventions.

In contrast to von Hirsch, Norval Morris argues that just deserts is a limiting principle for sentencing rather than a defining principle. Morris argues this on the basis that there is no unique answer to the deserved amount of punishment, but there is a strong intuitive understanding of what amount of punishment is undeserved. Just deserts enables determination of the appropriate maximum and minimum levels of sentence for an offence, but within that range there is room for variation to achieve other goals. The 'fine tuning' within the range should, in Morris's view, be done on the basis of utilitarian goals. For instance, deterrence could be pursued by imposing a sentence close to the maximum. Morris acknowledges that this does not equate with the principle of equality of sentence for like offences which desert theory stresses, but argues that equality can be no more than a guiding principle in a system which operates with so much inequality elsewhere, for instance in the prosecution decision (see von Hirsch and Ashworth 1992, pp201-6).

3.1.4 Just deserts and sentence severity

Criticism of just deserts is often based on its apparent association with escalation of sentencing levels. The penalty scale and its upper and lower limits will of course determine the severity or otherwise of a desert-based system. This is largely a matter of politics and conventions. In Finland the desert approach led to a scaling down of sentences, and there was a significant reduction in the prison population in the next decade and-a-half (von Hirsch 1990, p409). In Sweden the code was designed to be neutral in its impact on imprisonment levels, and in Minnesota, Washington, and Oregon the subsequent increases in imprisonment levels were lower than the increases in other states. In California, on the other hand, sentence levels were set by the legislature in a highly politicised manner, with a resultant increase in severity and imprisonment levels.

Proponents of just deserts argue that it actually provides a counter to pressures to increase sentences by its insistence on proportionality. Ad hoc changes to sentencing levels would disrupt proportionality, so any change necessitates a change in all sentencing levels. Unlike deterrence or incapacitation, just deserts sentencing is not designed to reduce future crime. This means it cannot be seen as failing and requiring a ratchetting up of sentencing levels if crime does not reduce.

3.1.5 Just deserts and social inequalities

Just deserts is one means of pursuing fairness and equality before the law by requiring like offences to be treated in like manner. A key objection to this is that it ignores differences between offenders which are often related to social inequalities.

Von Hirsch responds to this argument by querying whether social disadvantage can be held to diminish harmfulness and blameworthiness. Blameworthiness still attaches to acts against those who are not responsible for the offender's social disadvantage, and much crime is committed against victims who share the offenders' social disadvantage. Most crucially, social disadvantage does not reduce the harm done by the offence (von Hirsch 1990, p408).

Just deserts often draws negative reactions for its punitive approach and seemingly uncaring attitude to the causes of crime. This raises the question of how well suited sentencing is to address wider social issues and crime levels as a whole. Any major role in this respect seems ambitious in view of the fact that the sentencing system deals with a relatively small proportion of offenders (see Chapter One). In addition, social inequality afflicts many who do not come in contact with the sentencing system, so it is difficult to foresee how sentencing can address this issue at the general level. This is the province of wider social policy measures, not of a system set up to respond to crime following prosecution.

In striving for fairness in sentencing, just deserts can be seen as ensuring that socially disadvantaged offenders are not further disadvantaged in sentencing. This is particularly due to the exclusion in pure desert-based sentencing of offender status factors such as race and employment, on the basis that they are not material to culpability.

The traditional way in which the criminal justice system has addressed social disadvantage at the level of individual offenders is through rehabilitative programmes, and these are not incompatible with just deserts. The length and type of sentence can be set on a desert basis and within that sentence it is possible, and indeed desirable, to provide programmes designed to address the offending-related needs of the individual in an effort to minimise further offending. Similarly, with respect to community-based sentences of equal desert value, a choice could be made as to which particular one would be best suited to the offender's rehabilitative needs (von Hirsch and Ashworth 1992, p185). (See 3.4 for further discussion of rehabilitation and 3.6 for discussion of possible combinations of sentencing rationales.)

The just deserts focus on equal sentencing is said to ignore the fact that the same amount and type of sanction will not necessarily have the same impact on different offenders. It would, however, be an immense task to adapt sentencing to each person's individual sensibilities, and a concept of the general person is usually adopted when setting sentencing levels. (See 4.4 for a discussion of the principle of equality of impact.)

3.1.6 The use of just deserts in overseas jurisdictions

The just deserts philosophy of sentencing has gained increasing currency in recent decades, especially with the loss of faith in utilitarian goals of sentencing through the apparent failure of policies based on these goals, rehabilitation in particular. Pure desert theory requires offence seriousness and offender culpability to be the only influences on sentencing in every case, but it is not often applied in this pure form. Desert is frequently the main guiding principle in setting sentences, but it may be over-ridden in particular instances by other concerns such as incapacitation, and certain mitigating and aggravating features may also operate to increase or decrease the sentence.

Just deserts is the basis of the sentencing codes of Finland and Sweden. Article 6 of the 1976 Finnish Penal Code states that "punishment shall be measured so that it is in just proportion to the damage and the danger caused by the offence and to the guilt of the offender manifested in the offence". The 1989 Swedish Criminal Code, Chapter 29, has a similar statement of desert as the primary determinant of sentence. The key factors are "the harm, offence or risk which the conduct involved, what the accused realised or should have realised about it, and the intentions and motives of the accused" (Ashworth 1992, p83). A key exception is the provision for deterrent sentencing of serious drink drivers.

The German criminal law has an overriding desert philosophy. The amount of punishment must be proportionate to the offender's culpability and the seriousness of the offence. Rehabilitation and deterrence may be pursued within the narrow sentence ranges determined by proportionality (Albrecht 1995, pp6-7). The desert rationale has also been used in the production of highly structured sentencing guidelines. The Minnesota Sentencing Commission chose a modified desert rationale for their sentencing guidelines, with primary weight given to the seriousness of the current offence, and variations allowed for offenders with lengthy criminal histories. Just deserts was also the basis for guideline formulation in Washington and Oregon.

The England and Wales Criminal Justice Act 1991 made desert the primary rationale, but included special provisions for the courts to pursue incapacitation in the sentencing of serious violent and sexual offenders. Section 3(3) restricts the courts in determining the seriousness of the offence, to "the circumstances of the [current] offence", and directs them not to consider "information about the offender". The desert rationale was considerably weakened by amendments to the Act in 1993. Although s3(3) is unaltered, the courts' discretion has been widened as to what can be considered in regard to "the seriousness of the offence". Factors which are more related to the past conduct of the offender than the offence(s) for which sentence is being passed can now be considered. Thus, "seriousness of the offence" may include taking account of the offenders' previous convictions, and "any failure of [the offender] to respond to previous sentences." In addition, the fact that an offence was committed while on bail for another charge is to be considered an aggravating factor.

3.1.7 Summary

In a just deserts system the punishment emphasis is balanced by a consideration of rights. The objective is that the sentence is both just and deserved by being linked to the seriousness of the offence. The harm done or risked, and the offender's culpability are the key determinants of offence seriousness.

Just deserts does not pursue wider social goals through sentencing, although some theorists argue that public censure or denunciation, and the deprivation of money, time, or liberty which expresses this censure have a general crime prevention role. This is achieved through a positive reinforcement of the norms and values which society wishes to uphold. The extent of such effects is of course hard to measure.

One issue which the just deserts model has difficulty in addressing is that of restitution to victims and the community. The focus of just deserts is very much on the state responding to the offence on behalf of, and for the benefit of, the community. There is potential for conflict here between the censure deemed to be appropriate to the offence and the compensation or other actions required to meet victims' needs. For instance, desert may indicate a prison sentence as appropriate which may mean it is very difficult, if not impossible, for the offender to make any compensation to the victim. Just deserts and restitution are more compatible if a benefits and burdens approach is taken towards just deserts. Potential conflict still remains, however, in the emphasis of just deserts on fairness to the offender through like sentences for like offences, as opposed to restitution's requirements for fairness to the victim. This issue is further discussed in the section on restitution (3.5).

3.2 Deterrence

Individual and general deterrence are frequently cited as aims of, and rationales for, sentencing. Deterrence-based sentencing aims to influence future levels of offending through instilling fear of future consequences in either the individual offender, society at large, or both. It is based on the notion that the threat of the sentence is a major constraint on offending.

The justification for deterrent sentencing is encapsulated in the words of Jeremy Bentham, the nineteenth century 'father' of deterrence. Bentham argued that deterrence elevates a punishment from mere vengeance against the offender by transforming it into an "indispensable sacrifice to the common safety" (quoted in von Hirsch and Ashworth 1992, pp62-3).

Deterrence became prominent in the late 1970s after the loss of faith in rehabilitation. This was particularly reflected in the USA in the introduction of a range of mandatory minimum penalties to ensure that imprisonment was imposed. The pursuit of deterrence has been one of the factors in the continued escalation of sentence levels in response to rising crime rates.

If individual deterrence is the goal, then penalties are escalated once a person starts reoffending. The sentencing judge could, for example, choose to make an example of a persistent burglar by imprisoning him or her for the maximum term, even if the current offence is relatively minor. If general deterrence is the goal, any common offence could be met with very severe penalties no matter how serious the actual offence is. The legislature could provide a mandatory minimum penalty of imprisonment for any driver caught exceeding the speed limit in an attempt to discourage such behaviour.

Deterrence has widespread support as a goal of the sentencing system, and continued high rates of crime are frequently cited as evidence that current penalties are not harsh enough. This was a commonly expressed view among participants in a 1994 survey commissioned by the Department of Justice. In particular there was a view that prison is now too soft to be a deterrent. Individual deterrence was seen as especially appropriate for young and first time offenders, especially by older respondents. There was a minority view that corporal and capital punishment would be effective deterrents. The survey also revealed that shaming is seen as a necessary component to (individual) deterrence.

Few would dispute that the existence of a sentencing system and the fact that those who are convicted of crimes generally receive a sentence has some deterrent value. The question to address in discussion of whether deterrence should be a specific aim of sentencing and a guiding rationale for sentencing decisions is more concerned with the marginal deterrence to be achieved by increasing the sentence beyond what would normally be imposed, in order to send a message to the individual involved, and/or to the rest of the community. An alternative question is how much crime rates would be affected by a de-escalation of sentences.

The following discussion of deterrence relates to deterrence achieved through sentencing. It is important to bear in mind, however, that fear of the potential sentence is only an issue if the potential offender believes there is a good chance of being caught and convicted. The deterrent potential of changes in detection, investigation, and prosecution is a separate issue from the pursuit of deterrence through sentencing.

3.2.1 Deterrence as a rationale in current sentencing practice

There are particular instances in which the New Zealand courts have seen fit to impose exemplary punishments. This has occurred mainly for serious offences, where the offence is committed in a particularly grave form, and where it is prevalent in the community and is causing public alarm.

The courts have shown greatest readiness to impose deterrent sentences for pre-planned offences. In considering appeals related to the manufacturing of a class B drug, the Court of Appeal (per Cooke J) acknowledged that the extent of deterrence through severe punishment is contentious and varies between cases, and that the prospective sentence may not have a major influence on those already addicted to drugs. Notwithstanding this, the Court saw a major role for use of imprisonment as a general deterrent, stating that "there can be no time when it is more necessary for the courts to use their sentencing powers firmly in the hope of deterrence than at the early stage of the growth of a new social evil". In recognition of the planning and preparation involved in the manufacture of homebake, the Court found it reasonable to assume that "prospective penalties will from time to time deter at least some from a disastrous and insidious course" (R v Latta [1985] 2 NZLR 504 (CA)).

Deterrence has been seen as appropriate in fraud cases involving substantial amounts and abuse of positions of trust. In R v Raymer (CA 446/90 and 12/91, 23 July 1991) [1991] BCL 1703, the Court of Appeal (per Casey J) stated that "sentences must not only punish the offender and act as a deterrent to him; they must also serve the wider purpose of protecting society by deterring others of similar propensity ...." Deterrence has been cited in cases of commercial blackmail, where people have taken the law into their own hands, for crimes committed on contract, drink-driving offences, and aggravated robbery.

In the following sections the pursuit of deterrence is examined on ethical and effectiveness grounds.

3.2.2 Ethical issues

Deterrence may conflict with the sense that the criminal justice system should be humane and just in its operations. General deterrence does not respect the rights and autonomy of the individual, but merely treats him or her as a means to achieve the greater social goal of deterring others from like conduct. The sentence imposed has little to do with the offender and his or her culpability, and more to do with crime levels and concern about crime, which will vary over time and between localities.

Questions of fairness are raised if sentences for like offences differ markedly as the result of these external factors. Consider the case of white collar fraud. Offences of this type have caused substantial concern in recent years, and it is the type of planned crime which is often thought susceptible to deterrence. Should current offenders be given exemplary sentences to deter others? If rates of offending subsequently decline, should future offenders receive a sentence which is more commensurate with the seriousness of the offence and hence of a lower level?

Norval Morris argues on the other hand that general deterrence accords with the principle of using the least intrusive sentence possible, because it makes it unnecessary to sentence all the guilty equally. In fact, some can even be diverted from prosecution, if a few are selected for exemplary punishments to serve as a signal to the rest of the community. The additional deprivation suffered by those individuals selected for exemplary sentences is justified by the deterrent effect, and because the total deprivation is smaller than it would be under just deserts sentencing (because not all the guilty are sentenced). This cost/benefit analysis depends on the costs attached to the sacrifice of equality before the law (see von Hirsch and Ashworth 1992, pp202ff).

Individual deterrence, which is most often applied to repeat offenders, has a greater connection to the actual offender in that it relates to his or her past conduct. The risk here is that individual deterrence may amount to additional punishment for past offences. (This is further discussed in Chapter Six on the sentencing of persistent offenders.) Such sentencing is likely to impact strongly on the socially disadvantaged, as offenders from these groups are likely to have a higher probability of reconviction. To escalate their sentences for deterrent purposes would risk scapegoating this already vulnerable group (Ashworth 1992, p172).

In theory a deterrent rationale (both general and individual) provides no limits on sentence as long as it will deter. Since the aim is to create fear of penalties, there is a tendency for deterrence to inspire calls for harsh prison conditions, lengthy prison sentences, and capital punishment (for general deterrence only in this case), as these are the options which represent the extremes of fear-inducing sanctions in the modern mind. This tendency is exacerbated when past or current levels of sentence apparently fail to deter.

3.2.3 The effectiveness of deterrent sentencing

Many studies of deterrence are concerned with the deterrent effect of changes in levels of enforcement and prosecution rather than changes in sentencing. This section concentrates on the available evidence on the marginal deterrent effect of increases in sentencing levels.

Evaluation of deterrence often compares the crime rates of jurisdictions which differ in the probability or severity of sentencing, or they compare crime rates before and after a change in probability or severity of sentence in a given jurisdiction. If a negative relationship is found between crime rates and sentence probability or severity (i.e. if crime rates go down when probability or severity of sentencing is increased) this is seen as indicative of deterrence.

Deryck Beyleveld's review of such research concluded that:

Almost without exception, regardless of variations in design, correlation technique, control or modelling procedure, the expected relations have been found between probability of sanctions and offence rates, but have not been found between severity of sanctions and offence rates. In the latter case, the only consistent exception is found with homicide, where severity is measured by the length of imprisonment for homicide.

(see von Hirsch and Ashworth 1992, p79)

Beyleveld and others involved in such analysis caution that a negative relationship between crime rates and the severity or probability of sanctions is by no means conclusive evidence of deterrence. The 1978 (United States) National Panel on Research on Deterrent and Incapacitative Effects reported that:

Any conclusion that these negative associations reflect a deterrent effect ... is limited principally by the inability to eliminate other factors that could account for the observed relationship even in the absence of a deterrent effect.

(Blumstein, Cohen, and Nagin 1978, p46)

The panel was particularly concerned that the available research had not been able to eliminate the probable effect of the simultaneous relationship between crime rates and sanction probability, especially imprisonment probability. This means that crime rates are likely to influence sentencing probability or severity, as well as, or instead of, sentencing probability and severity influencing crime rates as assumed under deterrence theory. The former may occur if high crime rates overburden the criminal justice system to the extent that arrest becomes less likely, charges are dropped or bargained down for administrative convenience, and prosecution cases are not able to be prepared so rigorously, resulting in a drop in sentence probability and severity. The panel and others who have assessed such studies also urged caution in view of the questionable reliability of the crime rate data (due to recording errors, the fact that so much crime is unreported, and that reporting rates may change over time, possibly even as a result of changes in sentence certainty or severity).

Another issue is the need to establish causality between the observed changes in the crime rate and the sentencing variables; eliminating the possibility that the changes in both are caused by a third factor. Even in studies which control for socio-economic and demographic factors it is still possible that crime effects are due to other 'unknowns'. Most aggregate crime rate/sentencing probability and severity studies also have difficulty in separating out deterrent effects from the incapacitation effects which may be achieved by more use of imprisonment. Another consideration is that the crime rate and sentencing severity are both in fact likely to be causally related to degrees of tolerance of crime. Societies with low crime rates are likely to be relatively homogenous ones, which will accordingly have a low tolerance of deviance. This may in turn produce relatively punitive sentencing regimes. Hence, low crime rates may be associated with severe sentencing patterns and vice versa. This suggests that cross-jurisdictional comparisons of crime rates and sentencing severity are likely to be inadequate as proof of the efficacy or otherwise of general deterrence. These problems are not likely to be so acute in time series analyses within individual jurisdictions, since other relevant variables are more likely to be constant.

Despite these reservations and despite the panel's caution as to adducing evidence of deterrence, they concluded that there was more evidence for deterrence than against it, and that "criminal sanctions do influence at least some criminal behaviour by some individuals". The available research was deficient in that it was not able to shed light on "the magnitude of the effects of different sanctions on different crime types" (Blumstein, Cohen, and Nagin 1978, p47). It is of course just such information which is necessary for sentencing policies based on deterrence.

Interviews and questionnaires which focus on self-reported rates of offending and perceptions of probability and severity of penalty also suggest some deterrent effect for those with a perception of a high probability of apprehension, and among those who consider the perceived penalties to be particularly unpleasant. They do not indicate a greater deterrent effect according to the perceived length of sentence (Beyleveld in von Hirsch and Ashworth 1992, pp81-3).

Both the panel's work and that of Beyleveld were undertaken in the late 1970s. Studies since that time show some evidence of deterrence through sentencing with respect to some crimes. For instance, one study of the effects of mandatory sentencing for violent crime involving guns (homicide, assault, and robbery) in six United States cities suggested strong support for deterrence of homicides with a gun, while the results for gun assaults and robbery with guns showed little evidence of deterrent effect. One possible reason for this differential effect was that the higher volume and profile of media reporting of homicide cases compared with the other offences meant that sentence levels were much more effectively communicated for homicide. The researchers also proposed the alternative view that the assault and robbery results may have been masked by recording inconsistencies and errors in the crime rate data, which did not apply for homicides because these are generally recorded with greater accuracy and consistency. Another significant outcome was that the extent of the effects varied considerably from place to place, suggesting the need for further research to explain these differences. In addition, the follow-up periods were relatively short (twenty-four to thirty-six months), so the effects measured were short-term and it was not clear whether they would decline over time (McDowall, Loftin, Wiersema 1992, pp378-94).

Another review of studies of the deterrent effects of longer sentences concluded that:

The average estimates indicate that the deterrent effect is strongest for rape and assault, weakest for hijacking and fraud with robbery, burglary, auto-theft, larceny and murder in between. With the exception of murder, violent crimes tend to have larger (negative) elasticities than non-violent crime [meaning they are more deterrable by longer sentences].

(Lewis 1986, p49)

(It is of note that this result contradicts that of Beyleveld's review in respect of murder/homicide, see above). The author nevertheless acknowledges that the research under review may well be subject to the same problems as those identified by the National Panel on Research on Deterrent and Incapacitative Effects (Blumstein, Cohen, and Nagin 1978), so there is still reason for caution as to the findings. On the other hand, he makes the point that these difficulties may be exaggerated, or that there may also be difficulties which work the other way to mask the deterrent effect (Lewis 1986, p59).

The available research does not offer strong evidence of a deterrent effect from changes in sentence levels, but nor does it prove that deterrence does not occur. The research points instead to a number of questions which need to be addressed in the consideration of deterrence as a goal of sentencing. There is a need to ascertain the circumstances in which penalties will deter. There may be a sentence threshold beyond which no further deterrence will be achieved, especially as people tend to discount the future. Or it may be that deterrence, and hence crime reduction, is better pursued through increasing the certainty of apprehension and conviction, than through increases in the severity of penalties, and that sentence levels could even be reduced.

Perhaps the only thing which can really be concluded is that while some sentencing policies may deter in some circumstances, there will be many other instances in which they will not have a deterrent effect. The following section canvasses some evidence and theories as to the limitations of deterrent sentencing.

3.2.4 Factors which may limit deterrence through sentencing

Deterrence depends on potential offenders rationally weighing up the costs and benefits of crime and then determining whether it is worthwhile. This calculation relies on the potential offender having sufficient and accurate information as to the likelihood of being caught and convicted, and on the likely penalty.

Even with this information, the actual calculation of costs and benefits is likely to vary between individuals. Key factors in the calculation are the person's readiness to take risks, the strength of his or her preference for behaving in a law-abiding manner and, indeed, whether the person considers the contemplated behaviour to actually be a crime. Other crucial variables are the individual's assessment of the profit to be gained (particularly in the case of non-monetary profit) and the person's life circumstances, most particularly what they stand to lose from being caught, quite apart from the impact which any sentence imposed by the criminal justice system would have on them, i.e. how family, friends, or associates would react, and how it would affect relationships, employment prospects, and social standing.

Some indication of these individual variables is available from interviews and questionnaires which ask respondents to rank the factors which inhibit offending behaviour. Legal controls on behaviour appear to be generally less important than other controls such as moral inhibitions, although those with greater involvement in offending attach more importance to legal controls than those with lesser involvement. Respondents typically overestimate probabilities of arrest and have an inaccurate knowledge of penalties, although those with greater involvement in offending have lower estimates of arrest probability than the general population. The question arises whether this lower estimate makes them more willing to offend, or whether the lower estimate is a product of their offending experiences. Beyleveld concludes that even if these studies show some people are deterred by their beliefs about sentences, they do not provide any insight into the crucial question for sentencing policy; which is how do changes in actual sentence levels influence beliefs about sentences among those who are deterred by their beliefs (Beyleveld in von Hirsch and Ashworth 1992, pp81-3).

An insight into the assumptions of offenders comes from a 1980s study of English burglars. They did not think they would be caught; they were not worried about the consequences of being caught (either because the sentence was regarded as an 'occupational hazard' or because they did not think about the consequences at all); and the rewards of burglary were rarely known in advance (von Hirsch and Ashworth 1992, p58).

New Zealand researcher Dr Julie Leibrich's interviews with convicted offenders in New Zealand revealed that the actual sentence was only a factor insofar as it would cause them to lose something else they valued. Whether or not they continued offending was influenced very much by other life circumstances, and by whether the person actually believed that the contemplated behaviour was a crime. As Leibrich summarised it:

They were deterred from further crime by not wanting to put the things they most valued at risk - whether that was their family, their work or their self-respect. Sometimes the deterrence was simply their own sense of morality, sometimes it was an external threat of consequence.

(Leibrich 1993, p237)

Another theory as to why sanctions do not deter in certain situations is the defiance theory of Lawrence Sherman. According to Sherman's theory, defiance causes increases in the seriousness, prevalence, and incidence of offending as the result of proud shameless reactions to criminal sanctioning. Sanctions provoke future defiance of the law when offenders experience the sanctions as illegitimate (unfairly or disrespectfully imposed); when offenders have weak bonds to the sanctioning agent and the community represented by the agent; and when offenders deny their shame and become proud of their isolation from the sanctioning community. Under the opposite conditions sanctions will produce deterrence, or if the factors encouraging defiance and deterrence are evenly counterbalanced, the sanctions will become irrelevant to future offending. Specific defiance relates to the sanctioned offender's reaction; general defiance relates to the reaction of a group to the sanctioning of one or more of its members (Sherman 1993, pp448-9, 459).

On the other hand, deterrence may simply be limited by the impulsive nature of many offenders. This is the view expressed in the 1990 White Paper Crime, Justice and Protecting the Public issued by the British Government. Faith in deterrence was seen as over optimistic because:

Much crime is committed on impulse, given the opportunity presented by an open window or unlocked door, and it is committed by offenders who live from moment to moment .... It is unrealistic to construct sentencing arrangements on the assumption that most offenders will weigh up the possibilities in advance and base their conduct on rational calculation. Often they do not.

(quoted by Ashworth 1992, p60)

Other salient factors include the possibility that the deterrent effect of longer sentences may be lessened by the tendency of people to discount the future. This would be more likely for additions to already long sentences than for additions to shorter sentences.

It is possible that the deterrent effects of the system may change, but have little impact if they do not filter through to would-be offenders. Similarly, the system may operate at much the same level of deterrence, but changes in potential offenders' perceptions through their own or friends' experiences may increase the deterrent impact.

The degree to which changes in sentence severity will impact on offending will also vary according to different types of crime. Logic suggests that crimes which are a long time in the planning and are carefully thought out, e.g. fraud, may be more susceptible in this regard than crimes of passion or those committed under compulsion, such as stealing to support a drug habit. Leibrich's interviews suggest the limits of deterrence for compulsive offenders. Her interviews with persistent drink drivers showed that:

nothing they were aware of - not even prison - seemed an effective deterrent. Although three of them commented that jail was a strong possibility if they got caught, it was no deterrent. Partly this was because the addictive behaviour itself was so strong and also because people felt the risk of being caught was so small. Indeed their experience suggested they were right. Even when the judge had made jail seem a certainty next time, the certainty of punishment was irrelevant if the risk of getting caught was slight.

(Leibrich 1993, p97)

In another case the offender saw her drug dealing as such a financial necessity that jail was not a deterrent. She had even made arrangements for the care of her children if she did get put away (Leibrich 1993, p100).

Finally, policies aimed at deterrence may have effects other than the prevention of crime. Offenders may react to the greater risk by making greater efforts to avoid detection, or by escalating the degree of the offending so that the potential benefit continues to outweigh the risk. In the latter instance, the white collar criminal would actually embezzle more money in the face of a higher likely penalty. Alternatively, action taken against one type of offence may cause a substitution effect whereby potential offenders move to another type of crime. On the other hand, it is possible that deterrent action against one type of crime may also decrease complementary crimes, for instance increased enforcement against drug use may cause a decrease in dealing.

3.2.5 Summary

While it is reasonable to assume that the very existence of the criminal justice system has some deterrent value, there is little evidence to support the view that increasing the level of sentences will deter the individual offender or would-be offenders in general.

One of the difficulties with deterrent sentencing is that there are numerous potential influences on offending behaviour, and these will vary between individuals. Even if the potential for conviction and sentencing acts as a general deterrent, this does not necessarily mean that an increased level of sentence will be a greater deterrent. Another limitation is that potential offenders will not necessarily have full information about the likely level of penalty. The limitations of deterrent sentencing are perhaps best summed up by Leibrich's conclusion from her interviews with offenders that:

Penalties alone will not reduce offending. People who have nothing of value have nothing to lose. People who have no sense of belonging have no social status to risk.

(Leibrich 1993, p240)

Aside from these reservations as to how effective deterrent sentencing can be, there are ethical concerns with the premise of deterrent sentencing. Individual deterrence in effect punishes the offender again for previous convictions. General deterrence turns the offender into an example to others. Fairness between offenders of like culpability is clearly sacrificed if one or some are chosen as an example, or if some happen to be sentenced at a time when there is a particular concern about the level of offending. The question is whether, and to what degree, fairness should be sacrificed in return for an expectation of future crime prevention which has little empirical foundation.

3.3 Incapacitation

Incapacitative sentencing aims to protect the public from future offending by the particular offender being sentenced. The offender is not sentenced on the basis of the conviction offence, but rather according to predictions as to his or her likelihood of reoffending. Predictions of future offending are usually heavily influenced by the level and type of the individual's past convictions.

In this section, different types of incapacitative sentencing are outlined and issues addressed as to both the ethical implications and effectiveness of such strategies.

3.3.1 Types of incapacitation

Incapacitation is generally perceived to be achieved through imprisonment, as this takes the offender out of society and thus deprives him or her of the opportunity to commit offences (in society at least).

Collective incapacitation refers to a general lengthening of prison sentences and/or an extension of the use of imprisonment, often by picking out particular offence types for mandatory or mandatory minimum sentencing. This is why it is sometimes called charge-based incapacitation. Offenders are selected for longer sentences on the basis that they belong to a group (e.g. those that commit an offence classified as a serious violent one) believed to constitute a high risk of reoffending. This is based on the premise that locking up more known offenders for longer will help to prevent crime. Collective incapacitation is one of the reasons for the rapid growth in the United States prison population since the 1970s.

Selective incapacitation, as the name suggests, attempts to target prison resources towards offenders who are assessed as likely to commit further offences either at such a frequency, or of such seriousness (or both), as to cause considerable public nuisance or an actual threat to public safety. It gives courts the discretion to select a particularly long sentence for special high-risk cases on the basis of an individual assessment of the offender's likely future offending.

3.3.2 Incapacitation in New Zealand

In New Zealand, incapacitation is one of the goals traditionally pursued through the sentencing system. The Court of Appeal has affirmed that protection of the public is a legitimate factor to be taken into account in sentencing. The Court has, nevertheless, emphasised that any increases in sentence for incapacitative reasons should not take the sentence beyond what is proportionate to the gravity of the offence. On this basis, the Court has reduced a number of lengthy finite sentences imposed on persistent offenders in order to bring the sentences to levels proportionate to the offences. The leading case is R v Ward [1976] NZLR 588, where the Court stated that the "desirability of prevention must be balanced against [the] gravity" of the offence. The Court saw this limitation as necessary because incapacitative sentencing, based as it often is on assumptions drawn from past offending, may in fact amount to a repeat punishment for that earlier offending.

New Zealand law also includes the sentence of preventive detention which has a specifically incapacitative focus. This is an indefinite sentence, with a standard non-parole period of 10 years. Amendments to the Criminal Justice Act in 1993 now allow the sentencing judge to impose a non-parole period longer than ten years in "exceptional cases". There is no limit on the non-parole period which the judge may impose.

Preventive detention is an example of selective incapacitation in that it is focused on offenders perceived as particularly dangerous. It can only be imposed on offenders who are at least twenty-one years old, and who either have a history of violent or sexual offending since the age of seventeen or have been convicted of sexual violation. The sentencing judge may impose this sentence if he or she is "satisfied that it is expedient for the protection of the public that [the offender] ... should be detained in custody for a substantial period". In the case of an offender convicted for the first time of sexual violation who has no other previous conviction for a qualifying offence, the judge must first obtain a psychiatric report on the offender, and be satisfied that there is a "substantial risk" that he or she would commit one of the specified violent or sexual offences on release (Criminal Justice Act 1985, s75).

Whereas the incapacitative effect of a finite sentence only lasts for the length of the sentence, preventive detention has far greater potential by virtue of its indefinite nature. It is possible to hold an offender in prison for as long as he or she (always a male to date) is considered a danger to the public. Furthermore, after release the sentence continues to run, so that the preventive detainee may be recalled to prison if his behaviour gives cause for concern.

While preventive detention offers considerable potential for public protection, it is a very extreme example of the exercise of the state's coercive power. It is for this reason that the Court of Appeal has emphasised that this sentence should be considered very much as a sentence of last resort. The Court of Appeal has made it clear that the "controlling principle" of proportionality as expressed in Ward does not apply to offenders who qualify for preventive detention (serious and persistent violent and/or sexual offenders). Lengthy finite sentences may, therefore, be imposed in such cases for incapacitative reasons. In fact, the Court has made it clear that such sentences should be considered ahead of preventive detention, which should only be imposed if adequate public protection cannot be achieved through a lengthy finite sentence. This has not been changed by the 1985 Criminal Justice Act or the 1987 amendment to it (R v Bidwell (CA249/85, 20 December 1985) [1986] BCL 245; R v Glen-Campbell (CA330/87, 24 March 1988) [1988] BCL513; R v White [1988] 1 NZLR 264; R v Mataira (1992) 8 CRNZ 194). On this basis, the Court has on occasion replaced preventive detention with lengthy finite sentences. (See for instance R v Pratt (CA61/78, 8 August 1978) [1978] BCL, 792; and R v K (1990) 6 CRNZ 210, 212).

New Zealand has two other incapacitative provisions in sections 80 and 105 of the Criminal Justice Act. Section 80 (the result of a 1993 amendment to the Act) provides that the court may, when the circumstances of the offence are exceptional, impose a minimum period of imprisonment greater than ten years in the case of indeterminate sentences or in excess of two-thirds of the sentence up to a maximum of ten years in the case of serious violent offenders sentenced to more than two years imprisonment.

As with preventive detention and s80, s105 is aimed at serious violent offenders, but it is not a sentencing provision. Instead it allows for the Parole Board, upon application by the chief executive of the Department of Corrections (prior to 1 October 1995 the Secretary for Justice), to order the postponement of an inmate's release until three months before the end of the nominal sentence if it is satisfied that, if released before that date, the inmate would commit one of the specified serious violent or sexual offences before the date on which he or she would be released under the order. Since this provision came into force in 1987 there have been twelve applications for a s105 order and the Parole Board has made nine orders. (Statistics as at 1 October 1996 from the Department of Corrections)

3.3.3 Ethical questions

The major objection to incapacitative sentencing is the ethical one that it infringes a citizen's civil rights to be punished not for something they have done, but for what it is believed they might do in the future. Incapacitative sentencing means that offenders are sentenced more severely than justified by the seriousness of the offence. Under selective incapacitation, sentences for like offences may be grossly dissimilar (because of differences in the perceived risks associated with the offender). Unless selective incapacitation is applied to only the most serious offences, a further consequence is disruption of the penalty scale; i.e. a lesser offence may receive a higher penalty than an offence of greater seriousness.

The use of past offending in predictions of future criminality means offenders may receive additional punishment on account of past offences. This is the danger cited by the Court of Appeal in its emphasis that incapacitation must be tempered by proportionality, except in the case of the serious violent and sexual offenders who qualify for preventive detention. Another potential problem is that predictions that an individual will reoffend may have a labelling effect and thus be self-fulfilling.

The acceptability of these consequences will depend on the degree to which proportionality of sentence and offence is accepted as the norm, and on the acceptability of a trade-off between a certain amount of disparity and a certain amount of crime reduction.

Whether incapacitative sentencing achieves crime reduction, and if so, how much, depends on the accuracy of the predictions of future offending. Prediction accuracy is further discussed below, but the general point is that prediction techniques remain problematic. The National Academy of Sciences commented in their 1978 examination of incapacitation that:

Poor prediction not only undermines the utilitarian justification for selectively incapacitating some convicted offenders, but it also introduces concern for the injustice suffered by those who are imprisoned because their future crime propensity is erroneously predicted to be higher than it is.

(Blumstein, Cohen, and Nagin, 1978, p76)

The problem persists. More recently, von Hirsch has commented on the problems in curtailing a person's liberty "merely because people like him or her will offend again, ... [when] we cannot specify which of them will actually do so" (von Hirsch and Ashworth 1992, p104). In von Hirsch's view there are strong pressures to over-predict criminality, because preventive imprisonment conceals erroneous confinements but reveals erroneous releases. This leads to calls for more preventive imprisonment, and inevitably more false predictions of future offending (von Hirsch and Ashworth 1992, pp120-1).

An analogy is sometimes drawn between confinement on the basis of erroneous predictions and conviction of innocents. Conviction requires 'proof beyond reasonable doubt', but if people are detained on the basis of a 33 per cent to 50 per cent probability of future offending this would seem to fall far short of the normal standards of criminal proof (Tonry in von Hirsch and Ashworth 1992, p172). (Many predictions of serious violent offending are assessed as having a one in three accuracy rate, with 50 per cent as the best success rate, as is noted below at 3.3.4.) However, the problem is not simply that predictions are inherently uncertain, but that even if a prediction could be based on reliable grounds (in that, for example, the offender is in fact clearly disposed towards a particular sort of conduct) that still does not mean that he or she will act in accordance with it, because other intervening factors may change the situation. For example, at the extreme end of possibilities, a crippling accident or serious illness may prevent someone reoffending upon being released from imprisonment, but this does not mean that the original prediction was erroneous. Therefore, arguing against predictions on the grounds that they cannot meet the standards of proof required for convictions is perhaps not that useful.

Another argument concerns the dangerous non-offender. If the risk of future offending constitutes a reason for sentencing an offender beyond what is justified on normal tariff principles, then could it also be a reason for confining a potential offender who has yet to offend? One response is that everyone has a right to be presumed harmless until that right is forfeited by committing a serious crime, which indicates a capacity to harbour and carry out harmful intentions. The rights of potential victims then come to take precedence over the rights of the offender.

This balancing of rights in favour of potential victims was central to the conclusions of the 1981 British Floud Report on dangerousness. The authors recognised the "grave harm" of unnecessary confinement on incapacitative grounds, and concluded that this could only be justified if it relieves potential victims of the substantial risk of grave harm. On this basis, "protective sentencing" should be restricted to "cases where the predicted harm from the offender is quite severe" (on the basis of previous harm inflicted). Those who have committed very serious crimes may be exposed to the risk of unnecessary confinement, but this is preferable to society bearing the risk of victimisation by the offender (Floud and Young in von Hirsch and Ashworth 1992, pp104-5).

Ethical issues are also raised in respect of the factors used in predictions. Factors such as gender, race, ethnicity, and age may correlate with recidivism and violence probabilities, but since these factors are beyond the person's control, should they be used in predictions? A similar argument relates to the use of alcohol and drug addiction, employment, and educational status variables. The use of these status factors for predictions would concentrate the impact of incapacitative sentencing disproportionately on already disadvantaged (and highly imprisoned) socio-economic and ethnic groups.

The 1986 United States National Panel on Research on Criminal Careers (Blumstein, Cohen, Roth, and Visher) concluded that past criminal offending is the best single indicator of future offending. This means omission of the above status factors may not have a great impact on the accuracy of predictions. One remaining issue, however, is how past offending is measured. If arrests are included this may unduly inflate an individual's previous offending record, especially as those with an offending history may have a greater likelihood of arrest, if only because they are known to police. It could be argued that if past offending is to be used for predictions to determine sentence length and curtailment of liberty, then only proven or admitted offending should count (see Tonry in von Hirsch and Ashworth 1992, p175). In New Zealand the use of anything other than offences where convictions have been entered would be contrary to the New Zealand Bill of Rights Act 1990 (s26).

3.3.4 The accuracy of predictions

The incapacitative effect of imprisoning an individual will depend on more than just whether he or she would have committed another crime. It will also relate to the frequency with which he or she would have continued to commit offences, the seriousness of those offences, and how long he or she would have continued to behave in this way. Considerable attention has been devoted in recent years to developing predictive models. The literature to date suggests, however, that there are severe limits on the predictive capacity of these models.

In any predictive strategy two types of errors are possible: false negatives and false positives. False negatives predict some persons to fail (reoffend) who actually succeed, while false positives predict some persons to succeed who actually fail. Raising the trigger point to classify fewer offenders as high risk will decrease the risk of false negatives, but may also result in more false positives, and hence a loss of credibility in incapacitation as a crime control strategy. On the other hand, incapacitative strategies which have a high proportion of false negatives, and therefore unnecessary imprisonment, will have substantial human and monetary costs. The same is true of predictions which overestimate the years of criminal career left to an individual. Detention will not be incapacitative when it continues beyond the point at which the individual would have ceased offending.

Incapacitation effects are also influenced by the potential for replacement of imprisoned offenders. This is particularly likely in respect of organised group activity such as car theft or drug dealing.

Selective incapacitation is often advocated for dangerous offenders, but very serious violent crimes are a relatively rare event. This makes them particularly hard to predict, because there is not a large body of data from which to draw statistical correlations between the observed characteristics of known offenders and their subsequent criminal conduct. In addition, violence is difficult to predict because of its situational quality, and because a number of factors external to the offender may influence whether it is reported (see von Hirsch and Ashworth 1992, pp116, 118).

In its look at prediction studies in respect of serious offending (and generally serious violent or sexual offending) the Floud Report found none with better than a 50 per cent success rate, and many which resulted in two false positives for every one true positive. The authors concluded that statutory tests of dangerousness were not feasible for the restricted range of offences for which they proposed incapacitative sentencing because of the inadequacy of predictive judgements, and because the number of offenders would be relatively small and their characteristics extremely varied. (The offences for which they advocated protective sentencing were those involving death; serious bodily injury; serious sexual assault; severe or prolonged pain or mental stress; loss of, or damage to, property which causes severe personal hardship, damage to the environment which has a severely adverse effect on public health or safety; and serious damage to the security of the state). They concluded that the imposition of these protective sentences should rely on the exercise of a broad discretion (Floud and Young in von Hirsch and Ashworth 1992, p128). In 1992, Ashworth reported that there had been no significant improvements on the Floud Report's findings on predictions (1992, p65).

Studies show that actuarial (statistical) methods of prediction based on selected objective characteristics of the offender have had a higher success rate than clinical predictions based on a diagnostic approach to the individual characteristics of an offender. Actuarial prediction is gaining ground, particularly in the United States, as its techniques become more sophisticated. It is part of a conceptual shift from a subjective approach involving a diagnostic assessment of an individual's psychology for indications of dangerousness, to a more 'objective' one of matching individuals to high risk factors statistically linked to the highest probability of future violent offending. This application of 'rationality' rather than human discretion is considered by its advocates to be more exact, consistent and transparent, and therefore fairer. (It is also cheaper.)

One example of the use of clinical prediction in New Zealand is the requirement for a psychiatric report before preventive detention can be imposed on a person convicted of sexual violation for the first time (Criminal Justice Act 1985 s75). Advocacy of clinical methods may be due to a concern that the basis for the decision is not individualised enough in statistical analysis, and important case-specific variables may be ignored. There may also be uneasiness over including socially sensitive variables such as ethnicity as a basis for sentencing. On the latter point, the finding of the United States National Panel on Research on Criminal Careers that past criminal offending is the best single indicator of future offending is salient (see 3.3.3). This implies predictions can exclude status factors without a great impact on the accuracy.

An important observation of relevance here is that the literature on offence specialisation indicates that offenders typically engage in a range of illegal behaviours of varying seriousness and type so that those convicted of very serious crimes are more likely to be subsequently reconvicted of a different category of offence rather than one within the same category. This has been supported by data on the post-release history of New Zealand prisoners sentenced to imprisonment for serious violent offences (Brown and Cameron 1995; Brown 1996). It can be stated that most offenders commit a variety of offences in an unpredictable fashion, probably determined by the situation they are in and the opportunities presented to them, so reliance on an offender's most recent conviction as an indicator of the type of future crime he or she will most probably commit is likely to be inadequate. In the case of serious offending, this is more unlikely to prove a reliable predictor of subsequent conduct through the operation of the statistical principle known as regression to the mean. This states that extreme events (such as a very serious offence) are by definition infrequent and are likely to be followed by less extreme events which lie closer to the average. In other words most offenders' behaviour is likely to be relatively minor and serious offences will be infrequent and will be both preceded and followed by periods of less serious criminal activity (Brown and Cameron 1995, p422). This principle was backed up by the New Zealand study noted above.

3.3.5 Some prediction studies

The main British prediction study was by Brody and Tarling in 1981. Only nine of the forty-eight people predicted as dangerous committed dangerous offences within five years of release from prison, while an equal number (nine) of dangerous offences were committed by offenders not classified as dangerous (quoted in Ashworth 1992, p65).

The 1986 United States National Panel on Research on Criminal Careers (Blumstein, Cohen, Roth, and Visher) examined the estimated effects of various collective, individualised, and offence-based incapacitation strategies. With respect to collective incapacitation, the panel concluded (with reference to the United States) that to reduce crime by 10 per cent would require existing prison populations to more than double. The panel noted that even this crime reduction may be an overestimate, as it did not take account of the possibility that other offenders may take the place of those who are imprisoned. Incapacitation was seen to have more potential when selectively applied to smaller high-risk populations. For example, it was estimated that a 5 to 10 per cent reduction in robberies could be achieved with 10 to 20 per cent increase in the population of imprisoned robbers.

These estimates were made according to the expected frequency of offences per offender per year imprisoned. They did not take account of the fact that some of the imprisoned offenders may have ceased to commit crimes during the time period. Nor did they account for the effect that the fact of imprisonment may have on the offenders' future criminal careers. Imprisonment may lengthen the career, through the so-called 'school of crime' effect; or perhaps shorten the career through rehabilitative programmes undertaken while in prison. These exclusions, and the potential influence of factors from outside the criminal justice system, mean that the crime control benefits of incapacitation are often overestimated.

Overall, the Panel concluded that "while incarceration leads to some reduction in crime through incapacitation, none of the ... strategies examined provide dramatic reductions" (Blumstein, Cohen, Roth, and Visher 1986, pp142-3). Cohen suggests an alternative focus on criminal careers. Research by Blumstein and Cohen on robbery and burglary offenders in Washington DC which was based solely on past and present criminal records indicated that they have relatively short careers, but commit the offences at a relatively high frequency. This suggests short prison sentences for these offenders should prevent a lot of crime. Cohen also argues that because the sentences would be similar for those with similar past and present offences, problems of disparity would not arise (Cohen 1983).

An American study followed 6,310 offenders who reflected a random sample of men incarcerated in Californian prisons in 1962-63 through a follow-up period of 26 years. During that time, men from this cohort were arrested 30,464 times following their release; which included nearly 10,000 arrests for serious offences (homicide, rape, kidnapping, assault, car theft, burglary, thefts, robberies). These 30,464 arrests involved 4,897 of the 6,310 men. None of the prediction models (for purposes of selective incapacitation) that the study applied was able to distinguish with a significant degree of accuracy the recidivists from the non-recidivists or the type of offence recidivists were most likely to commit. All of the subsequent crimes could have been prevented by keeping all the offenders in prison (collective incapacitation on the basis of their 1962-63 sentence). This would have been at the cost of incapacitating 1,413 men in error (if the arrests made were legitimate and if those not arrested did not in fact offend). The above findings led the researchers to conclude that "the utility of incapacitation as a crime control strategy will be constrained by the limits imposed by predictive validity" and perhaps "by the nature of criminal careers" (Gottfredson and Gottfredson 1994, p468).

In New Zealand, senior psychologists of the Department of Justice and others have developed variables to summarise criminal conviction histories of offenders convicted in the years 1983 and 1988 (Bakker, Riley, Deely, et al 1995). The variables were used to develop predictive models on the likelihood of offenders being reconvicted for offences of specified seriousness, being sentenced to imprisonment, and being sentenced to imprisonment for a specified term. When tested against unseen data of subsequent post-sentence conviction or non-conviction a very close relationship was found between the probability of reconviction and the sentence outcome for a group of individuals, and the actual outcome in terms of the percentage of the group who were reconvicted and received imprisonment, including imprisonment for a specified term. For instance, when the model gave an individual a 0.85 probability of reconviction then close to 85 per cent of the people with that probability were reconvicted. The largest over- or under-prediction probabilities were at the top and bottom ends of the probability scale, where the numbers of offenders were smallest.

Although this model looks promising, it is yet to be validated in a genuinely predictive manner. In addition, it was developed for use in assessing priorities among sentenced offenders for access to treatment and programme resources (i.e. priority to be given to those assessed as in the higher risk groups). Although potentially very significant in terms of resource allocation for correctional programmes, such use of prediction models is of an entirely different order to its use in determining whether, and for how long, an individual should be deprived of his or her liberty. The model seeks to identify whether the individual belongs to a high- medium- or low-risk group, but does not identify the risk associated with the specific individual.

3.3.6 Summary

The experience of the United States in recent decades suggests that collective incapacitation policies are extremely costly in terms of prison overcrowding, resources, and wasted human lives, while crime rates remain high. Similarly, in New Zealand crime rates have generally risen along with prison populations.

Selective incapacitation offers the possibility of a better return than collective policies because it is targeted at high-risk or frequent offenders. The difficulty is that its effectiveness is completely dependent on the ability to predict future criminal careers. The high likelihood of inaccurate predictions and their human and monetary costs suggest that selective incapacitation should be pursued in only the most extreme cases of public danger. That is, it should be applied only to those offenders who pose a real danger to the public or to certain persons. This means the policy would be restricted to a relatively small number of offenders, and this of itself would be a means of minimising the potential harm of wrongful confinement. The dilemma here is that serious violent offences are among the hardest to predict. This issue is discussed further in the section on the sentencing of dangerous offenders (Chapter Seven).

Chapter 3 continued