Reconceptualising the Custody Threshold in England and Wales

The custody threshold provision in England and Wales was intended to operate as a limit on the use of custodial sentences, preserving what is the system’s most severe sanction for the most serious offences. However over the past few decades it has become apparent that the custody threshold is failing. Academics have discussed the reasons for this failure, which has seen the prison population double in space of a quarter of a century. This piece explores the custody threshold in the context of the use of custody in other Western European jurisdictions. It examines the courts’ response to the provision and various judicial attempts to amplify Parliament’s language. The authors then consider the academic critiques of the custody threshold provision, analysing the extent to which said criticism can be seen as a solution to the problem, before offering a new critique of their own. Finally, in a move towards more a more principled approach to the custody threshold, the piece offers a solution which would, it is argued, make the provision more effective and more theoretically sound.

fessional and personal lives of prisoners (and their families) with little compensating benefit in terms of rehabilitation. Longer-term imprisonment impairs the life prospects of ex-prisoners for years after release. There are also considerations beyond the individual offender. Many western jurisdictions have for years struggled to constrain their prison populations. England and Wales has both a rising prison population 3 and a use of custody which is significantly higher than most other western European nations such as Germany. For all these reasons, there is a clear imperative to resort to custody only when absolutely necessary, when no lesser, non-custodial sanction would be appropriate. This position is captured by the principle of restraint or parsimony.
Regulating the use of imprisonment creates challenges for legislatures. One method of constraining the use of custody involves codifying a direction to courts to use custody only for certain offences or offenders. This approach often takes the form of a custody ''threshold'' provision, whereby an offender may not be imprisoned unless the offence has crossed some threshold of seriousness. For example, England and Wales placed a seriousness-based custody threshold on a statutory footing (in 1991), instructing courts not to pass a custodial sentence unless the court was of the opinion that the offending behaviour was so serious that neither a fine alone nor a community sentence could be justified for the offence. 4 Other jurisdictions have introduced similar provisions. 5 Legislative directions of this kind represent a common approach to ensuring that imprisonment is reserved for the most serious offences. Yet sentencing and prison statistics confirm that custody threshold provisions have proven ineffective in restricting prison to the most serious offences, with the result that custodial populations in many western nations that have introduced them remain stubbornly high. In England and Wales, the focus of this article, the prison population remains near the top of the EU prison statistics table: a recent report found that England and Wales reported an imprisonment rate of 148 per 100,000 -significantly higher than other western European countries such as Germany (76), France (100) and Italy (86). 6 This article explores the statutory provision in England and Wales which currently regulates the use of custody as a sanction. We consider how this key provision could more effectively ensure that only the most serious cases receive a term of imprisonment. Criticism of the current provision has led to calls to abandon the concept altogether. 7 In contrast, we advocate revision rather than repeal of the custody threshold provision. There are several reasons for focusing on the custodial threshold as a potential mechanism to restrict the use of custody as a sanction. First, most common law countries have adopted some form of custody threshold; there is therefore a significant international experience with this strategy for directing courts away from the imposition of custody which can inform reform in this jurisdiction (just as, in turn, our analysis of the deficiencies of the current provision in England and Wales carries lessons for other jurisdictions). Second, amending a single provision is more feasible than larger-scale (or longer-term) reforms such as adjusting guideline sentence recommendations across a wide range of offences. 8 Third, a provision of this kind carries an important message to courts and also to the community about the need for penal restraint and the role of imprisonment.
Part I describes the origins of the custodial threshold, and summarises academic criticism. The current statutory provision has been criticised for being too vague, and for permitting a very subjective determination of the kinds of cases for which custody is ''inevitable''; an unclear provision leaves different judges with the discretion to set the threshold wherever they believe appropriate. Part II explores ways in which this element of sentencing law may be improved. We Criminal Law Review 593. 8 In a jurisdiction like Minnesota, the use of custody can be regulated directly through the presumptive sentencing guidelines. These mandate a sentence or range of sentence for specific offences or categories of offender. The Minnesota Sentencing Commission can therefore regulate the use of custody or alternatives to custody by assigning offences to different locations within the two-dimensional sentencing grid. Minnesota courts are required by statute to follow the sentence recommendations.
propose a re-framing of the custodial threshold, one which restricts the influence of factors such as prior convictions or plea which are unrelated to the gravity of the offence. The justification for focusing on the relationship between custody and previous convictions is that a common cause of imprisonment for offenders convicted of less serious offences is the offender's criminal history. This is true in all Anglo-Saxon jurisdictions but particularly England and Wales and the United States. 9 Part III makes some concluding remarks. We suggest that our proposed reforms would not only make the custody threshold provision more effective at restraining the use of custody as a sanction, but also make the concept more theoretically sound.

I ORIGINS AND CRITIQUES OF THE CUSTODY THRESH-OLD
The ''custody threshold'' in England and Wales is located in s.152(2) of the Criminal Justice Act 2003. This provision mirrors those found in other common law jurisdictions 10 and has been the subject of much academic criticism. The principal critique of the custody threshold concerns the lack of clarity of the statutory language, and the way in which the test has been interpreted by the courts. It is asserted that the provision is vague and allows (or even encourages) a subjective interpretation; as a result, it is inconsistently applied. 11 Moreover, commentators have said that despite the ''tight'' wording of the provision, it has proved ineffective in restricting custody to the most serious cases. Section 152(2) entitled ''General restrictions on imposing discretionary custodial sentences'' states: 9 For example, in a recent analysis of sentencing statistics Roberts and Watson demonstrated that a high percentage of offenders convicted of a minor offence such as shoplifting were committed to custody as a result of their prior convictions. See J.V. Roberts and G. Watson, ''Reducing female admissions to custody: Exploring the options at sentencing'' (2015) 15 Criminology and Criminal Justice 1. 10 For example, s. 16 of the Sentencing Act 2002 in New Zealand and s. 718.2(e) of the Criminal Code of Canada. The German law (which is addressed in Harrendorf's article in this issue) is rather different, in that what English law calls ''community orders'' can only be imposed on offenders as part of shaping the terms of a suspended sentence (or as a condition of a later release on parole). 11 Padfield, fn. 7 above, at 610, observed that ''the height of the custody Ôthreshold' seems to vary between sentencers and over time''.
The court must not pass a custodial sentence unless it is of the opinion that the offence, or the combination of the offence and one or more offences associated with it, was so serious that neither a fine alone nor a community sentence can be justified for the offence.
This provision had an important predecessor in the Criminal Justice Act 1991 which was considered by academics to constitute ''a landmark of sentencing legislation''. 12 Section 1(2) of the 1991 Act stated: ''Subject to subsection (3) below, the court shall not pass a custodial sentence on the offender unless it is of the opinion (a) that the offence, or the combination of the offence and one or more offences associated with it, was so serious that only such a sentence can be justified for the offence; or (b) where the offence is a violent or sexual offence, that only such a sentence would be adequate to protect the public from serious harm from him.'' Section 1(2) of the 1991 Act led to the use of the term ''custody threshold''. 13 However, the concept of a ''line in the sand'' between avoidable and unavoidable custodial sentences existed in common law prior to the Act. 14 Section 1(2) formalised that concept by creating a general statutory restriction on the use of custody. It was underpinned by the 1990 Government White Paper, Crime, Justice and Protecting the Public, 15 which was clear in its intention that with the 1991 Act, the government would pursue a more systematic approach to sentencing. Wasik and von Hirsch, who saw previous sentencing legislation as ''a rag-bag of proposals which reflect[ed] short-term considerations with little inherent coherence'', noted with approval that the 1990 White Paper was ''substantially along the right lines in advocating a coherent set of guiding principles for sentencing''. 16 The ''thoughtful approach'' pursued by the government therefore stems from an emphasis upon principle and considered sentencing policy, as opposed to short termism and politics. Commenting specifically upon the proposed custody threshold provision in the White Paper, however, Wasik and von Hirsch remarked that ''something clearer is needed''. 17 The 1991 Act largely failed to deliver, however, and was thought by some to have been let down (at least in part) by its drafting. 18 Section 1(2) of the 1991 Act was interpreted in R. v Cox 19 by Lord Chief Justice Lord Taylor, who indicated that the phrase ''so serious that only such a sentence can be justified for the offence'' meant the kind of offence which when committed by a young person would make right-thinking members of the public, knowing all the facts, feel that justice had not been done by the passing of any sentence other than a custodial one. 20 Subsequent cases did not clarify this more satisfactorily, merely applying the test without consideration as to its clarity. 21 Ashworth queried whether such a sweeping test as that formulated by Lord Taylor CJ was in fact the best that we can do, questioning both the propriety of such an approach and the accuracy with which the views of ''right-thinking'' people could be determined. 22 He proposed a presumption that certain types of offences would not attract custodial sentences. Although such a measure would result in a reduction in the use of custody in practice, its application would need to confront the 17 Ibid. at 517. 18 One such example was said to be that: ''The provision on the length of prison sentences stated only that sentences should be Ôcommensurate with the seriousness of the offence' and this was not enough to prevent the then Lord Chief Justice from distorting the provision by construing it to mean Ôcommensurate with the punishment and deterrence which the seriousness of the offence requires'. principle of ordinal proportionality, which may present some practical and theoretical challenges to this approach. 23 In an article published in 1997, von Hirsch and Ashworth asserted that the ''right-thinking member of the public test'' was ''conceptually flawed and empirically unsupported'', 24 and suggested that the aim should be a progression towards ''firmer criteria in sentencing'': although the custody threshold was ultimately left to the discretion of the sentencer, they argued that the test should ''require a normative judgment rather than an exercise of discretion…on an unprincipled basis''. 25 Subsequently, in R. v Howells, then Lord Chief Justice Lord Bingham considered that: There is no bright line which separates offences which are so serious that only a custodial sentence can be justified from offences which are not so serious as to require the passing of a custodial sentence. But it cannot be said that the Ôrightthinking members of the public' test is very helpful since the sentencing court has no means of ascertaining the views of right-thinking members of the public and inevitably attributes to such right-thinking members its own views. 26 His Lordship concluded that ''It would be dangerous and wrong for this court to lay down prescriptive rules governing the exercise of that judg[e]ment…'' clearly placing significant value in the exercise of judicial discretion. 27 This subsequent development did little to clear the muddy waters; the court had seemingly endorsed Ashworth's criticism of Cox but merely stated that ''test'' to apply was the court's own subjective view of what justice required on the facts before it. The test therefore remained rather opaque and the custody threshold difficult to identify. 23 If a sentencing system were to adopt such a measure, it would likely have to accept a dilution of the principle of ordinal proportionality in order for it to be effective, as observing ordinal proportionality would require the range of offences to which the presumption applied to be so wide or so narrow so as to make the presumption unworkable in practice. While this is not a bar to its successful operation, it presents a challenge to justify such an approach on theoretical grounds. 24  Koffman criticised the Court of Appeal for not providing constructive guidance on the interpretation of s.1(2). 28 Whether it had been the intention of the 1991 Act for the Court of Appeal to provide such guidance, is unclear. However, as set out above, several successive Lord Chief Justices ruled on the interpretation of s.1(2). As such, Koffman's critique is more of an indictment of the lack of clarity in the drafting of s.1(2) as opposed to the court's ''failure to play its part''. Koffman considered that decisions of the court in upholding immediate custodial sentences in cases involving minor property offences rendered the threshold ''largely meaningless'', 29 and endorsed the von Hirsch-Ashworth view that the decision not to interpret the provision in a manner consistent with the purpose of the 1991 Act had set the custody threshold at a point lower rather than higher than before, this being directly contrary to the intentions set out in the White Paper. 30 The custody threshold provision has changed very little between its original enactment in 1991 and its current form. 31 The main substantive difference between the 1991 and the 2003 provisions is the apparent removal of the second alternative of the 1991 provision for crossing the custody threshold -that ''the offence is a violent or sexual offence [and] only such a sentence would be adequate to protect the public from serious harm from [the offender]'' -from the 2003 provision. In fact, the public protection exception was retained but the way it was to apply was modified. The 2003 Act created a number of sentences for public protection and the custody threshold provision was expressed not to apply to such cases. As regards the (2): ''Subject to subsection (3) below, the court shall not pass a custodial sentence on the offender unless it is of the opinion:(a) that the offence, or the combination of the offence and one or more offences associated with it, was so serious that only such a sentence can be justified for the offence; or(b) where the offence is a violent or sexual offence, that only such a sentence would be adequate to protect the public from serious harm from him.'' That provision was repealed in 2005 and re-enacted in Criminal Justice Act 2003 s.152(2): ''The court must not pass a custodial sentence unless it is of the opinion that the offence, or the combination of the offence and one or more offences associated with it, was so serious that neither a fine alone nor a community sentence can be justified for the offence.'' first -and now only -alternative based on offence seriousness, the changed formulation is intended to draw the line exactly where it was drawn before, but more clearly so. There has been very little discussion of section 152 since its enactment -perhaps as a result of Lord Bingham CJ's comments in Howells -and any discussion has unfortunately been little more than descriptive. 32 It is clear, then, that since Howells the Court of Appeal has been consistent in its view that the threshold test is something of an enigma: attempts to provide guidance as to the interpretation of earlier incarnations have been dismissed and no such attempt has been made to provide such guidance pertaining to the current provision. Parliament's intention as to the effect of s.152(2) is that espoused in the 1990 White Paper: more offenders should be dealt with in the community. Although the courts have been silent on the issue since Howells, when Parliament's intention is set against the statistics for the use of imprisonment as a disposal it is clear that the custodial threshold requires reform.

Reactions to the Perceived Failure of the Custody Threshold Provision to Restrict the Use of Custody
Among the more radical proposals was that advocated by Padfield who rejected the notion of a ''threshold'', suggesting that a ''custody zone'' may be preferable. 33 The argument was premised upon the false sense of security created by the term ''threshold'' which, she argued, suggested clarity where there was none. Whilst it was Parliament's intention that s.152(2) operates as a restriction on the use of custody, as with its predecessors, Padfield pointed out that there ''is no easily ascertainable custody Ôthreshold'''. One reason given by Padfield is that s.152(2) is flanked by provisions concerning an offender's prior record that allow for custody to be imposed for an offence that would otherwise not warrant it; and by provisions allowing a guilty plea to lead to a non-custodial sentence even where the offence would otherwise warrant its imposition. (Both previous convictions and plea are discussed at greater length in the next section, 1.2, below). Thus, where an offence is so serious that it crosses the custody threshold, the eventual sentence may properly come beneath it and conversely, an offence which appears to fall beneath the Padfield concluded that the absence of a bright line representing the custody ''threshold'' rendered it a ''simplistic concept which gets in the way''. 34 As sentencing is an evaluative task that involves more complex thought processes than simply weighing aggravating and mitigating factors so as to locate the offence above or below the ''threshold'', she argued it was ''time to bury the custody threshold''. She proposed a diagram which permitted the disposals to overlap. While Padfield's critique made a strong case for reconceptualising the ''threshold'' as a ''zone'', her proposal does not constitute a substantial improvement on the status quo. The ''custody zone'' may be a better label for what the current law creates than a ''custody threshold'', but the problem remains of how to locate the custody zone, and determine which cases fall beneath it.
Ashworth focused his criticism on a different issue: the apparent failure of the judiciary to draw the line based on offence seriousness (and the requirement of proportionate punishment) as such. Reviewing cases in the Court of Appeal (Criminal Division), Ashworth contended that the conclusion in those cases that immediate custody was inevitable appeared not to be a judgment of proportionality but rather to derive from the deterrent rationale or from an expressive or denunciatory rationale, treating imprisonment as the only unambiguous way of demonstrating society's rejection of such conduct. 35 Accordingly, he questioned whether or not there should be another form of punishment that achieved public censure without imposing imprisonment. Recognising that such an approach may be divisive, he suggested tightening the statutory wording and including a requirement that courts explain why an offender cannot be dealt with by way of a community penalty when imposing any sentence of under two years' custody. 36 This, again, makes a compelling case for reform, but it is difficult to see how the statutory wording could be improved upon. 34 Padfield, fn. 7 above, at 612. This is a view arguably shared by Lord Bingham CJ, see above. 35  Ashworth's suggestion that courts might be required to explain why a community sentence could not be imposed in cases in which a custodial sentence will be of two years or less sounds attractive in that it appears to encourage the court to turn its mind to the issue of the custody threshold. However, two objections might be raised. First, this is surely little more than the current requirement to give reasons for sentence 37 ; arguably, the duty to explain why an offence cannot be dealt with by way of a community sentence already falls within the current duty imposed upon the court. Second, even if the proposal is discretely different from the current duty to give reasons for the sentence imposed on an offender, this duty is in practice ineffective at compelling sentencing courts to modify their practice, as a failure to give adequate reasons does not invalidate a custodial sentence. 38 The Sentencing Council 39 addressed the problem in 2016 when it consulted on a draft guideline concerning the imposition of community and custodial sentences. 40 However, the approach taken by the draft guideline failed to assuage concerns about the custody threshold. A joint academic response to the consultation paper criticised the ''[broad reproduction of] the relevant statutory provisions'' noting that ''it [contained] no guidance on how to apply them''. 41 The response acknowledged that the Court of Appeal had declined to expand upon the concept but maintained that it was ''vital'' that the Council provide ''substantive guidance'', offering some alternative methodologies for providing guidance on what was accepted to be a difficult issue. These suggestions notwithstanding, the definitive guideline (issued in late 2016) simply restates the statutory test and asserts that there is no general definition of where the custody threshold lies. 42 The definitive guideline additionally repeats the statement from the 1990 White Paper that the intention of the provision is to reserve custody for the most serious cases, but it offers no further guidance on how to identify these cases.
In circumstances where the Court of Appeal (Criminal Division), the Sentencing Council and academics have been unable to improve on the wording of the custody threshold provision, the conclusion appears to be that -at least in its current form -the wording cannot be improved upon and the concept remains vague and problematic. The position has therefore not advanced since the early 1990s.

The Relationship between the Custody Threshold and Surrounding Provisions on Prior Convictions and on Pleas
We think that there is an important additional reason why the custody threshold provision has so far failed to achieve the objective of restricting the use of custody to cases where the seriousness of the offence necessitates a custodial sentence. Although the provision purports to focus upon offence seriousness, the test in fact expands to encompass case seriousness, opening the door to consideration of the offender's previous convictions and other matters unrelated to the offence of conviction such as bail status at the time of the offence. The provision (s.152) does not stand alone and a related provision engages the offender's prior convictions in the context of determining seriousness. Sentencers are given further guidance as to how to apply the test for imposing a custodial sentence in the form of other provisions in the 2003 Act. 43 Section 143, entitled ''determining the seriousness of the offence'', engages the offender's prior convictions, as follows: ''(1) In considering the seriousness of any offence, the court must consider the offender's culpability in committing the offence and any harm which the offence caused, was intended to cause or might foreseeably have caused. 42 Sentencing Council, Imposition of Community and Custodial Sentences: Definitive Guideline (London: Sentencing Council, 2016), available at https://www.senten cingcouncil.org.uk/wp-content/uploads/Definitive-Guideline-Imposition-of-CCS-fi nal-web.pdf. 43 Sections 145 and 146 concern aggravation of an offence by reference to race, religion, disability, sexual orientation or transgender identity, and mandate that such features are treated as an aggravating factor when a court is determining the seriousness of an offence. Section 145 excludes racial or religious aggravation for offences under Crime and Disorder Act 1998 ss. 29-32 where racial or religious aggravation is an ingredient of the offence.
(2) In considering the seriousness of an offence (Ôthe current offence') committed by an offender who has one or more previous convictions, the court must treat each previous conviction as an aggravating factor if (in the case of that conviction) the court considers that it can reasonably be so treated having regard, in particular, to (a) the nature of the offence to which the conviction relates and its relevance to the current offence, and (b) the time that has elapsed since the conviction.'' It can be seen that for the purposes of s.152, the ''seriousness of the offence'' is broadly defined; it includes not only the offence, assessed by the culpability and harm approach now familiar in the English and Welsh sentencing guidelines, but also factors beyond the offence, specifically limited to any relevant previous convictions, and whether or not the offence was committed on bail. It excludes other nonoffence factors such as personal mitigation, equity mitigation and any plea of guilty. Section 144 of the 2003 Act also requires a court to consider the stage at which a guilty plea was indicated, and in what circumstances it was entered, before deciding the nature of the sentence to impose. 44 The combination of s.152(2) and s.144 is contradictory: in determining whether or not a custodial sentence should be imposed, the court is to consider the seriousness of the offence (which does not include a guilty plea as a relevant consideration) yet a guilty plea must be considered prior to determining what kind of sentence to impose.
The upshot is a test which is theoretically inconsistent. By identifying offence seriousness as the guiding consideration in determining whether the custody threshold has been crossed, s.152 reflects a desert-based approach to sentencing. That being the case, an offender's prior convictions should play only a limited role in the determination of the severity of sentence. Although there are competing accounts of the role of previous convictions in desert-based sentencing, the consensual position among proponents of proportionality is that prior misconduct should not exercise a significant influence on sentence severity. The seriousness of the offence is unaffected by whether the offender has a short or long criminal his- 44 Criminal Justice Act 2003 s.144(1) states: ''In determining what sentence to pass on an offender who has pleaded guilty to an offence in proceedings before that or another court, a court must take into account:(a) the stage in the proceedings for the offence at which the offender indicated his intention to plead guilty, and(b) the circumstances in which this indication was given.'' A natural interpretation of the language therefore suggests that a guilty plea is relevant to the nature of the sentence to be imposed, not simply its length. tory. 45 Whether repeat offenders are more culpable is a disputed question; many retributivists deny any link between culpability and previous offending. 46 Even those who see some retributive significance in an offender's record regard it as a modest source of aggravation. 47 The methodology which emerges from the ensemble of statutory provisions and the case law is thus unnecessarily complex. First, s.152 requires a court to determine whether the custodial threshold has been passed. This determination is guided by the seriousness of the offence, a retributive consideration. However, in determining seriousness the court must move beyond the offence to examine the offender's record, but not factors amounting to personal mitigation which could be relevant to culpability. Section 143 states that a recent and relevant criminal history should increase the seriousness of the offence (unless it would be unreasonable to do so). The consequence is that there will be cases where the gravity of the offence alone is insufficient to move the case across the custodial threshold, but the offender nevertheless ends there by virtue of his prior convictions, a factor generally considered irrelevant to the retributive approach. If this happens, the offender now sits on the wrong side of the threshold, but may be brought back over to the non-custodial side by virtue of personal mitigation: factors such as good character, or primary caregiver which again, are unrelated to retributive sentencing. 48 This approach is reflected in the sentencing guidelines issued by the Sentencing Council. The concept of the custody threshold is incorporated into the various categories in the Council's offencebased guidelines. The assessment of seriousness (for both the custodial threshold and determining the length of the offence) is informed by the consideration of culpability and harm as the first step in the process. Whereas this demonstrates to sentencers the location of the custody threshold by reference to the offence categories, 49 in reality it merely transfers the exercise of discretion from identifying the custody threshold to determining which category to place the offence and thus the result remains the same. Accordingly, even with the advent of guidelines in England and Wales, the custody threshold remains a vague and poorly defined concept as the guidelines have not engaged with the threshold as a concept to any significant extent. This problem is compounded by the Council's decision not to attempt to define the custody threshold in the Imposition of Community and Custodial Sentences Definitive Guideline. 50 Taken together, these critiques help explain why the custody threshold has not restrained the use of custody as a sanction as was intended by Parliament; the case for reform remains, we suggest, compelling.
If the custody threshold is working well, the prison population should be composed of offenders convicted of serious crimes for which no community based order was appropriate. There is compelling evidence in England and Wales that this is not the case. This may be illustrated by reference to recent sentencing statistics. Theft accounts for a high volume of cases. Almost 100,000 offenders were sentenced for a theft offence: approximately 10% of all sentenced cases. In 2014, approximately one-third of offenders convicted of theft received a sentence of immediate custody or a suspended sentence order. 51 Since theft is (generally) a low seriousness crime, it is 49 For example, in the Council's Assault Offences Definitive Guideline (available at http://www.sentencingcouncil.org.uk/wp-content/uploads/Assault_definitive_guide line_-_Crown_Court.pdf), Category 2 specifies for offences causing ''actual bodily harm (ABH) a range of a low-level community order to 26 weeks' custody. There fore, with offences falling within Category 2 being those which typically involve greater harm and lesser culpability or lesser harm and higher culpability (but not both) sentencers are more easily able to ascertain the location of the custody threshold by a consideration of where a Category 2 offences falls within the category range.
50 Above fn 42. Interestingly, this is in contrast to the approach to the imposition of suspended sentence orders. The guideline provides general guidance as to the type of case which may be suitable for a suspended sentence order, see p. 8 of the guideline. 51 Before a court may impose a suspended sentence order (SSO) it must impose a term of imprisonment. An SSO is therefore a term of imprisonment, albeit one which is suspended. Statistics on the use of these sentences for theft can be found in Sentencing Council, Consultation on Draft Guideline for Imposition of Community & Custodial Sentences (London: Sentencing Council, 2015). plausible to conclude that many of these offenders were committed to custody as a result of their prior offending histories.
The sentencing pattern for theft suggests one way in which the custody threshold has misfired: although designed to ensure that crime seriousness determines whether the offender receives a term of custody rather than a community penalty, many offenders convicted of low seriousness crimes nevertheless are sentenced to prison. In our view, a custody threshold should exercise greater control over the decision to imprison. It should ensure that crime seriousness does indeed determine whether custody is imposed. In this way the underlying commitment to proportionate punishment is respected. Since serious offences represent a small proportion of the total caseload appearing for sentencing, an effective custody threshold would incarnate the spirit of restraint identified at the beginning of this article. An ancillary benefit of an effective custody threshold would be to exercise some restraint over admissions to custody, thus preventing the prisons from filling up with less serious cases.
In Part II of this article, we address how the custody provision in England and Wales could be made more effective, developing a proposal to reformulate the existing provision.

II REFORMULATING THE CUSTODY THRESHOLD IN ENGLAND AND WALES
2.1 Clarifying s.152 and its Relationship to s.143 If the sentencing regime reflects a proportionality model, the determination of whether the custodial threshold has been passed should focus on the offence and the offender's culpability for that offence. Any wider inquiry -into personal mitigation, the impact of the sentence on third parties and so on -would result in a loss of clarity in terms of penal censure. The sentence should reflect state censure for a specific act. This is not to deny a role for such other factors in ultimately determining whether the offender should be imprisoned. These (and related) circumstances should result in many cases in which the custodial threshold is passed yet where a non-custodial alternative is imposed. But incorporating considerations unrelated to the offence in the initial determination obscures the sentencing message of penal censure, which is for a specific proscribed act. Our reconceptualisation recognises that there are two principal ways in which the custodial threshold fails to restrict admissions to custody to the most serious cases. Some cases are committed to custody when the court has over-estimated the gravity of the offence, and imposed custody when a community order would have been sufficient. It is hard to determine how many admissions to custody are explained on this basis, or even how an independent authority might establish whether this was the case. After all, in a pluralistic society, some people will see significant harm in conduct others regard as relatively trivial infractions of the criminal law. One way of establishing the degree to which this occurs would involve an external audit of a sample of decisions to imprison. This might be conducted by an independent judicial authority and correspond to reviews of decisions to prosecute. 52 The second and more common cause of incarceration for conduct insufficiently serious to justify custody involves offenders convicted of minor crimes and who have lengthy records of recent, related prior convictions. At present, a significant (but unknown) number of offenders are committed to custody, notwithstanding the custodial threshold provision, by virtue of their criminal records. In such cases, a court may feel compelled to reflect the record by incarcerating the offender. This may occur through the application of section 143 which encourages such an approach, or because the court, having repeatedly imposed non-custodial sentences, sees no reasonable alternative except escalating the severity of the court's response by changing the nature of the sanction. 53 Our proposed reform would address the latter of these two sources of prison admission, by modifying the relationship between prior convictions and the custodial threshold.
Currently, and as noted, offenders may be propelled across the custodial threshold as a result of their prior convictions alone. This phenomenon is described as ''push-in'' in the US guidelines; it occurs when an offender convicted of an offence that would normally attract a non-custodial sentence is ''pushed-in'' to the custodial zone of the sentencing grid solely by virtue of his previous convictions. 54 We take 52 For further discussion of this proposal, see J.V. Roberts and L. Harris, ''Addressing the Problems of the Prison Estate: The Role of Sentencing Policy'', Prison Service Journal, in press. 53 On this latter point, a number of jurisdictions have codified directions to courts to encourage them to consider alternatives to custody, even when the imposition of community-based sentences in the past has failed to prevent re-offending. 54 For example, a first offender offender convicted of a seriousness level VII offence receives a stayed sentence. If he has 3 criminal history points the presumptive sentence becomes 54 months' imprisonment. For discussion, see R. the view that while an offender's past should be taken into account at sentencing, prior convictions alone should not make the difference between community and custody, except in exceptional circumstances. The aggravating effect of prior convictions should be limited to increasing the severity of the non-custodial sentence or increasing the length of the term of imprisonment but should not change the nature of the sanction.
Why should the power of previous convictions to aggravate sentence severity be normally restricted to increasing the quantum 55 of punishment within a particular sanction range? The reason returns us to the notion of ''just deserts''. If a mitigating or aggravating factor is of marginal (or worse, no) relevance to the offence, giving it undue weight undermines the concept of proportionality. In this sense, restricting the power of previous convictions is simply part of a broader requirement to preserve proportionality.
How then, should previous convictions affect sentence severity? Where the offence has crossed the custody threshold, the matter is easily dealt with by a modest increase in the duration of imprisonment. However, where the offence has not crossed the custody threshold, but is near to it -i.e., a high level community order -the solution is not so simple. There are multiple options, but our preferred method would involve an extension of existing community order requirements. By enabling the court to impose (a) a certain number of hours, or additional hours of unpaid work; or (b) an extended operational period of an electronically monitored curfew, to reflect the offender's previous convictions, the sentence for the offence and the increase for the prior record are separately marked, enhancing transparency and public understanding. Additionally, for an appellate court, the sentence is clearer and easier to understand. 56 Preventing a court from imprisoning an offender on the basis of his record alone represents as significant departure from current practice in all common law jurisdictions. In England and Wales the statutory provision regulating the consideration of previous convictions allows this basis for committal to custody. In addition, a sig-55 It is of course not just about sanction-change; a quasi-retributive factor should not reduce a sentence from 15 to five years for the same reason. Yet the shift from community to custody is so fundamental and is more radical than, say, the transformation from a fine to a community order. 56 In some sentencing appeals, the Court of Appeal has to make assumptions as to the sentencing judge's methodology and approach when they only have the final sentence imposed to inform their discussions. This is further complicated by reductions for totality and guilty pleas. nificant number of offenders in states such as Minnesota are sent to prison as a result of their criminal history score alone. It would be unrealistic to expect courts to abandon this practice. For this reason, we advocate an exceptional circumstance provision which would permit a court to imprison certain highly repetitive offenders. This might adopt wording which would explicitly proscribe a policy of incarcerating on record: An offender's prior convictions are normally relevant only to the quantum of punishment imposed, and may not alone justify committal to custody. Only when the offender has a very extensive record of prior convictions and the court has exhausted all non-custodial sentencing options may a term of custody be imposed to reflect this.
A provision of this kind would at least inhibit the use of imprisonment in cases where the offence(s) of conviction alone do not justify committal to custody. As with all sentences of imprisonment the court should give reasons, and impose the shortest period of imprisonment.

The Relationship Between a Guilty Plea and the Decision to Imprison
For similar reasons as those pertaining to previous convictions, we propose a more limited role for the consideration of a guilty plea. 57 We advocate the view that a guilty plea is, in the context of factors relevant to sentencing, sui generis; it is a transaction between the state and the defendant borne out of pragmatism, as opposed to the more traditional view that in pleading guilty a defendant demonstrates remorse for his or her actions and an insight into the effect of the offence on the victim(s). In pleading guilty, a defendant ''buys'' a shorter sentence in exchange for the financial and social benefits of not having a trial. This view is supported by the Sentencing Council; in their guidelines, remorse is listed as a mitigating factor at Step Two, distinct from the consideration of any guilty plea. Further, in its recent consultation paper on the use of custodial sentences, the Council offered the following rationale: 57 Plea is important not just because of the statutory direction for courts to take this factor into account but because, like previous convictions, it affects a very high percentage of cases appearing for sentencing. Approximately 90% of cases sentenced in the courts involve a guilty plea rather than a conviction following trial; see J.V. Roberts and B. Bradford, ÔSentence Reductions for a Guilty Plea: New Empirical Evidence from England and Wales' (2015) 12 Journal of Empirical Legal Studies 187.
The purpose of reducing the sentence for a guilty plea is to yield the benefits described above and the guilty plea should be considered by the court to be independent of the offender's personal mitigation. Thus factors such as admissions at interview, co-operation with the investigation and demonstrations of remorse should not be taken into account in determining the level of reduction. Rather, they should be considered separately and prior to any guilty plea reduction, as potential mitigating factors. 58 As noted above, giving a factor that is of no clear relevance to harm or culpability such prominence in the determination of sentence type would be to undermine proportionality. Where the offence is so serious that only a custodial sentence can be justified, there is no justifiable reason why a guilty plea -a factor divorced from the offence entirely -should play a prominent (even determinative) role in the decision as to the nature of sentence. However, that is not to underplay the importance of a guilty plea. Our view is merely that because it does not relate to the seriousness of the offence (judged by the harm caused by that offence or the offender's culpability for that offence) its role in determining sentence should be circumscribed. As such, we propose that once the custody threshold has been crossed, absent exceptional circumstances, a guilty plea cannot bring an offender back beneath the threshold. It would however, be possible for a guilty plea to result in a suspended sentence order in place of an immediate custodial term, as this is a form of custody. 59

Proposed Methodology
We propose a staged approach to determining whether to imprison the offender, one which maps easily on to the sentencing guidelines step-by-step methodology. 60 The methodology consists of four stages: the first two stages determine the nature of the disposal, principally custody or community, whereas the second two stages determine only the extent to which the severity of sentence is increased or decreased. The methodology is as follows. . 60 The Council's guidelines all follow the same approach, namely creating a series of steps for all courts to follow when sentencing an offender.
First, a court should consider all factors relevant to the seriousness of the offence and the offender's culpability. 61 If all sanctions other than custody would fail to adequately reflect the total harm and culpability, the custodial threshold has been passed. This calculation excludes consideration of the offender's prior record for the purposes of determining whether imprisonment is necessary, 62 for the reasons discussed earlier in this article. 63 Relevant prior convictions should not weigh against the offender whose disposal hangs in the balance, effectively consigning him into prison on the basis of his record. We would also exclude consideration of a plea of guilty from this calculation for the reasons discussed above.
At the second stage, in the absence of any relevant prior convictions the offender is entitled to first offender mitigation, and this should exercise a powerful influence over the decision to imprison. 64 This stage also includes an assessment of all other aggravating and mitigating factors concerned with the offence and the offender. These factors may also have an important influence upon the nature of the disposal.
Once a court has resolved whether the offender has ended up in the custodial or non-custodial zone, it should extend the term of custody or enhance the punitiveness of the non-custodial sanction to the degree that the offender has recent, relevant convictions. 65 In this way, prior convictions aggravate the sentence imposed 66 but would not alter the nature of the sanction, consistent with the approach we advocate here. A court should also consider plea at this point, working in mitigation. A guilty plea is a non-retributive factor, and should not make the difference between community and custody. A guilty plea should reduce the sentence length (or the duration and onerousness of a community order) according to the timing of the plea as recommended by the definitive sentencing guideline. Since a suspended sentence order is a form of custody, it seems appropriate that a guilty plea should have the potential to move an offender from a term of immediate custody to one of suspended imprisonment. Similarly, a significant criminal record could operate in the opposite direction, disentitling the offender to a suspended sentence order, where the court had already determined that the custodial threshold has been passed. But as noted, neither consideration should have the power to change the sentence from non-custodial to custodial, or vice versa.

III CONCLUSION
The custodial threshold provision has to date failed to achieve its stated objective of ensuring that only the most serious cases are committed to custody. Rather than abandon the concept of a custodial threshold we have suggested reformulating the provisions in a way that focuses more clearly on the seriousness of the current offence. By restricting the role of non-retributive factors such as prior convictions and focusing a court's attention more tightly on the seriousness of the offence, this formulation would have two principal effects. First, it would lower the volume of offenders committed to immediate custody by reducing the number of offenders who are ''pushed into prison'' by virtue of their prior convictions. Second, the cases committed to custody would conform more closely to the profile envisaged by a retributive sentencing philosophy.
Reducing the use of custody as a sanction will be neither easy nor expeditious. It will require some political leadership, or at least acceptance that the current levels of imprisonment are unacceptably high, as well as a range of specific strategies to reduce the number of offenders sent to prison in a principled way. A useful, and symbolically important first step entails amending the key provisions which attempt, albeit ineffectively, to regulate the use of custody.
The reform proposed here would enhance the proportionality of current sentencing decisions in England and Wales by restricting the role of factors unrelated to harm or culpability. At the same time it may well reduce the volume of admissions to custody by lowering the number of cases in which an offender is imprisoned principally to reflect his prior convictions.