Time to exercise caution? Project CARA and out-of-court responses to domestic abuse

Abstract

Domestic abuse (DA) is a high-volume area of police work and the utility of criminal justice intervention has been questioned, especially in relation to cases perceived as low risk. This paper initiates debate on conditional cautions being used by some police forces as an alternative to a criminal charge. We highlight limitations of existing evaluations of the Cautioning Adult Relationship Abuse (CARA) intervention and potential unintended consequences of wider roll-out that remain untested. We suggest continuing to grow the evidence-base for behaviour change while addressing issues around implementation, with a greater emphasis on controlling and coercive behaviour.

INTRODUCTION

Responding to domestic abuse (DA) is a significant challenge for the police service in England and Wales. In the year to March 2022, the police recorded 910,980 DA crimes—a 116% increase since the year ending March 2016 (ONS, 2022). Though this increase may in part be due to greater willingness of victims to report (ONS, 2022), it is likely driven to a greater extent by a landmark 2014 inspection report on crime data integrity (HMIC, 2014) and subsequent police efforts to improve recording practices. DA now represents 17% of all crime recorded by police in England and Wales, and over a third of crime classified as violence against the person (ONS, 2022). Yet not all calls to police relating to DA are recorded as crimes. In the same period to March 2022, the police also recorded 589,389 non-crime DA related incidents (ONS, 2022).

It is beyond dispute then that responding to DA accounts for a significant proportion of police time. Yet alongside the increase in demand has been a squeeze on resources. Between 2010 and 2019, the number of full-time officers in England and Wales decreased by more than 20,000 (Home Office, 2019). There was also a significant decrease in numbers of police staff. This context of stretched resources appears to be particularly acute in specialist units tasked with investigating DA crimes (Myhill et al., 2022). Although an ‘uplift’ programme has sought more recently to replace the 20,000 officers, the issue of lost experience and specialism will take longer to address, and there has been no corresponding uplift programme for police staff.

Pre-dating even this context of increased demand and shrinking resources was a debate concerning the utility of criminal justice intervention in cases of DA. From the 1990s and following pressure from women’s advocates and critical research highlighting the low priority afforded to cases of DA (see for example, Edwards, 1989; Stanko, 1985), the police were encouraged to take a more interventionist approach. A Home Office circular issued to forces in 2000 introduced a policy of preferred arrest, and College of Policing authorized professional practice states currently ‘[o]fficers must be able to justify the decision not to arrest where the grounds exist and it would be a necessary and proportionate response’.¹ Similarly, CPS guidance states that ‘[g]iven the seriousness of DA offending, a prosecution will normally be required when the evidential requirements … are met’² and that the prosecution strategy should from the outset consider the possibility of proceeding without the support of the victim, if an evidence-led prosecution would not place the victim at greater risk of harm.

The interventionist approach has, however, had limited success in providing justice for victims. While cases classified as more serious may be charged and enter a recognized intervention pathway via Multi-agency Risk Assessment Conferences, many DA crimes result in ‘no further action’ (NFA) due to evidential difficulties including prosecution not being supported by the victim (Brennan and Myhill, 2022). Indeed, research has long identified that some DA victims don’t desire a criminal justice sanction for the perpetrator and seek instead police assistance while managing the abuse within the relationship (see for example, Hoyle and Sanders, 2000). Though the conviction rate for cases that are charged is currently 76%,³ recent research suggests cases of ongoing abuse are still frequently recorded and charged as discrete physical assaults that do not represent the totality of the offending and may not result in a significant sanction for the offender (Myhill et al., 2022).

‘Positive action’ policies have, then, resulted in significant numbers of arrests for ‘minor’ physical assaults that do not, in isolation and historically, attract a criminal justice sanction commensurate with what is frequently repeated abuse. Against this backdrop of ineffective criminal justice response and lack of recognized interventions for cases not assessed as ‘high risk’, one English constabulary instigated an experiment with a conditional caution, an out-of-court disposal not previously permitted in cases of DA. Currently operating in eight English forces, there have been calls for ‘project CARA’ to be adopted more widely (see Smith, 2022). This paper seeks to initiate (further) debate around the use of out-of-court disposals for DA. We should emphasize from the outset that we are not against in principle the use of conditional cautions for DA. We do however seek to highlight several issues we see as worthy of debate, including: limitations of the extant evidence-base; issues around implementation, including whether the format of the CARA intervention is appropriate for some of the offenders referred to it; and issues associated with upscaling this intervention.

CONDITIONAL CAUTIONS FOR DA

In England and Wales, conditional cautions ‘offer a proportionate response to low level offending’ as well as an opportunity for ‘offenders to be diverted at an early opportunity into rehabilitative services thereby reducing the likelihood of re-offending’ (Ministry of Justice, 2013: 4, emphasis added). Where an offender is given a conditional caution, criminal proceedings are halted but may be commenced should the offender not comply with the conditions of the caution (Ministry of Justice, 2013). As well as the offence being ‘low level’, the code of practice for adult conditional cautions states ‘[a] conditional caution is unlikely to be appropriate where the offence forms part of a pattern of offending’ (Ministry of Justice, 2013: 7). Historically, conditional cautions have been considered appropriate for cases of DA only in exceptional circumstances (Crown Prosecution Service, 2019).

In 2012, Hampshire Constabulary were granted permission by the Director of Public Prosecutions (DPP) to trial conditional cautions for certain cases of DA. The Cautioning and Relationship Abuse (CARA) trial involved randomly allocating offenders that had admitted an offence to receive a conditional caution requiring them either to desist from offending and attend two workshops (treatment) or desist from offending for a period of 4 months (control). The CARA workshops were 5 hours in length and utilized motivational interviewing techniques to encourage offenders to reflect and develop a sense of self-awareness, recognize abusive behaviour, and understand and use behaviour management strategies such as ‘time out’. The workshops introduced offenders to the Duluth power and control wheel, and emphasis was placed on ‘awareness raising’ while signposting men to other services (Chilton, 2012: 106–108). Further eligibility restrictions included the case not having been assessed as ‘high risk’ and the offender having no previous DA convictions (Chilton, 2012: 105, emphasis added).

Despite being conceived prior to the introduction of legislation to criminalize controlling and coercive behaviour in 2015, the genesis of the CARA intervention was a nuanced analysis by Jarman (2011) who drew on feminist accounts to acknowledge that it may be the cumulative effect of violence and intimidation (as opposed to the severity of a specific attack) that prompts victims to seek agency intervention, and that ‘minor’ domestic assaults are not minor for the victim. The official CARA project background document also suggested that while some offenders may have been arrested for ‘situational’ violence (see Johnson, 2008), a ‘high proportion of offenders will … be demonstrating signs of exerting power and control over their partners’ (see Chilton, 2012: 106). Yet subsequent accounts described the target population as ‘minor DA offending’ (Chilton, 2012: 2), with the background document also describing the trial as informing ‘the future handling of low-level DA incidents’ (Chilton, 2012: 105). The evaluators of the original CARA trial (see below) described a ‘vast haystack of low-harm, no repeat domestic callouts’ (Strang et al., 2017: 162).

We suggest there is an inherent contradiction between this presentation of CARA as a response to ‘low-level incidents’ and the way the intervention was conceived initially and is operating currently. At the time of writing, eight forces have dispensation to offer CARA in cases that have been assessed as either ‘standard’ or ‘medium’ risk. Evaluation data (Karavias et al., 2023) suggests more than half of cases referred to CARA are assessed as medium risk, whereby it is acknowledged that there are ‘identifiable indicators of risk of serious harm’ (ACPO, 2009). It is important therefore to assess the extent to which empirical evaluations of CARA have accounted for this issue in their assessment of impact and any potential risks as well as benefits associated with this intervention.

THE EMPIRICAL EVIDENCE-BASE FOR CARA

An initial evaluation of CARA via a randomized controlled trial (RCT) suggested the programme had a positive impact on ‘repeat offending’ (Strang et al., 2017: 171). Both outcome measures used in the evaluation have limitations. Almost one in five CARA participants were classified as having reoffended, based on traditional crime counts. Yet the measure reported—a further charge for a DA related offence (Strang et al., 2017: 170)—is one of the most restrictive. Comparatively few DA cases result in a charge (Brennan and Myhill, 2022) and any further contact with police might be seen as a more inclusive indicator of continued abuse. There is also evidence to suggest first response officers exercise discretion in relation both to recording crimes (Myhill and Johnson, 2016) and arresting suspects (Myhill, 2019) such that police data in the form of crimes, arrests and charges are limited and not entirely reliable measures of reoffending. And of course, there is no guarantee that victims will report further abuse to the police, so any outcome measure presented as ‘reoffending’ based on police data alone represents a huge assumption that no further contact equates to a desistance from abuse.

Strang et al.’s (2017: 169) preferred outcome measure was the Cambridge Crime Harm Index (CHI), and they reported ‘treatment group members were arrested for new DA crimes with 27% less CHI severity than the control group’. Yet while arrest is a broader measure than charge, applying the CHI to DA is problematic (for an in-depth critique, see Myhill and Kelly, 2023). Briefly, the CHI can lead to perverse outcomes in relation to DA due to the low sentence start-point for some crimes representative of violence against women. For example, many DA crimes, including a sizeable proportion of those referred to CARA, are recorded as assault occasioning actual bodily harm (ABH). An ABH assault scores ten points on the CHI. If a victim were to apply subsequently for a non-molestation order, a breach of that order by the offender would score five points on the CHI. Breach of a protective order in these circumstances would represent a 50% reduction in ‘harm’ according to the CHI, but very likely a continuation or escalation of abuse in the eyes of the victim. And the CHI is, of course, limited in the same way as simple crime counts in relation to the under- and mis-recording of crimes by the police.

Such concerns are less problematic if we are confident the RCT methodology eliminates bias that might compound these measurement issues for one group more than the other. The power of the RCT design rests with its internal validity. RCTs are sometimes regarded as the ‘gold standard’ for making causal claims, but these claims are weakened if the process of randomization is compromised. Of 1,469 cases deemed potentially eligible for CARA, only 293 were ultimately allocated randomly to the intervention or control conditions. Over 1,000 cases, then, were excluded from the trial and the evaluators acknowledged it was not clear to what extent this was a result of the application of pre-specified eligibility criteria, self-selection by victims or offenders, or first response officers failing to establish the eligibility of the case. Random allocation where the selection mechanism into the sample is unknown is not sufficient to eliminate confounding bias (Sampson, 2010) which is perhaps why Strang et al. (2017: 171) concluded their findings were sufficient to justify ‘further … trials’ as opposed to wider roll-out. There is also the question of external validity. Rarely in the social world are ‘randomized’ trials based on random probability or whole population samples (Sampson, 2010). In this instance, the initial sample would be influenced by officer discretion in relation to recording crimes (see Myhill and Johnson, 2016) and the highly variable nature of risk assessment (see Myhill et al., 2023). So even if the process of randomization was successful, some would argue that the non-random nature of the initial sample limits transferability of outcomes to other forces (whose crime recording and risk assessment processes might generate a non-equivalent sample).

A more recent evaluation by the University of Birmingham (Christie et al., 2022; Karavias et al., 2023) utilizing propensity score matching to create treatment and control conditions found reductions in both rearrests and ‘reoffending’ for offenders completing the CARA workshops, again based solely on police data. The evaluation covered two police forces areas. Similar proportions of CARA attendees were rearrested within 12 months in both areas (West Midlands 16%; Hampshire 14%). In Hampshire, the 12-month ‘reoffending’ rate for CARA attendees was 29%; in West Midlands it was 5%.⁴ Once again, any recontact with the police was not considered, nor other measures the evaluators themselves suggested as important outcomes, including the number of victims who report improved offender behaviour (Christie et al., 2022: 37–38). Notably, neither the Cambridge nor Birmingham evaluations investigated formally the ‘awareness raising’ element of CARA by assessing how many participants subsequently sought further (formal) help to change their behaviour.

This limited range of outcome measures, and especially the use of police data and absence of (qualitative) follow-up with victims, is problematic when considering the findings of an evaluation of full, Respect-accredited⁵ perpetrator programmes (project Mirabal). Though these programmes were longer in duration and served a somewhat different cohort of offenders,⁶ the methods employed mean there are relevant lessons from the evaluation. Kelly and Westmarland (2015), recognizing the difficulties associated with randomization and the near impossibility of generating a ‘clean’ comparison condition, chose instead a realist design. Longitudinal follow-up interviews with both victims and perpetrators were conducted across multiple sites to reveal a more complete picture of revictimization and illuminate the process of behaviour change. While Mirabal found significant reductions in physical violence, reductions in forms of abuse victims suggested were more salient to them—including respectful communication, ‘space for action’ and shared parenting—were less consistent. The evaluation identified a ‘continuum’ of change and suggested:

…any reports of absolutely no violence and abuse from the point men begin any perpetrator intervention would be highly suspicious given what over three decades of research on the patterns of violence and abuse has taught us (Kelly and Westmarland, 2015: 45–46).

They also concluded ‘[it] is the length and depth of [perpetrator programmes] which makes it possible to go beyond simple behaviour disruption’ (Kelly and Westmarland, 2015: 46). In other words, entrenched abusive behaviour requires a longer, more intensive intervention.

WHAT TYPE OF CASES ARE BEING REFERRED TO CARA?

Recall that the architects of CARA expected the intervention to serve both cases of ‘situational’ violence and cases where offenders are demonstrating ‘signs of’ controlling and coercive behaviour (Chilton, 2012: 106) but that the intervention was aimed primarily at ‘low level’ cases. The findings from the Mirabal evaluation would suggest it is situational violence and cases in which coercion and control are not well established which are most amenable to a shorter, less intensive intervention.

It has been suggested, however, that comparatively few cases of situational violence escalate to the point of requiring agency intervention (Johnson, 2008), due to the absence of an underlying pattern of abuse. Stark (2007: 235) suggests ‘partner assaults’ that come to the attention of the police are more likely to be repeated and act instrumentally to coerce and control. Indeed, qualitative research with survivors has shown victims do not typically disclose until they have attempted to manage the abuse (non-physical and/or physical) for a sometimes significant period, with the police frequently not the first agency to which victims make a disclosure. Kelly’s (1999) model of disclosure at a point of crisis is supported by more recent qualitative data (Evans and Feder, 2015; Puente-Martínez et al., 2023). In addition, Myhill (2015) found, using data from the Crime Survey for England and Wales, that victims were more likely to seek formal help if the abuse they experienced was ongoing and controlling and coercive in nature. Data from national population surveys in the USA (Cho et al., 2020), Canada (Akers and Kaukinen, 2009), and Spain (Domenech del Rio and Garcia del Valle, 2019) also associate reporting with more serious abuse, and reporting at a point of crisis has also been identified in samples of minority ethnic (Femi-Ajao et al., 2020) and male (Huntley et al., 2019) victims. It must therefore be open to question how much ‘minor’ or ‘low-level’ offending exists in the pool of (ex-) intimate partner DA cases reported to police and available to be considered for a conditional caution.

The University of Birmingham evaluation in fact gives an indication of the type of cases currently being referred to CARA.

…the Hampton Trust reported that whilst some of the male CARA Service users present with situational violence, the majority appear to be on a continuum more closely representing [coercive control] (Christie et al., 2022: 14).

Indeed, the list of offence types identified by the evaluation also suggests offenders displaying persistent abusive behaviour are being referred to the programme, as well as offence types which would appear to sit outside the normal criteria justifying a conditional caution. Across the two evaluation sites, the following offences have been referred to CARA: Malicious communications—conveying a threat (n = 66); breach of a non-molestation order (10); ‘revenge porn’ (9); grievous bodily harm (11); ill-treatment of a child (8); arson endangering life (3); harassment—putting in fear of violence (8); stalking (2); controlling and coercive behaviour (2); threats to kill (9) (Karavias et al., 2023: 9–10). Especially concerning is the acknowledgement that female ‘offenders’ referred to the programme are being identified by workshop facilitators as primary victims using resistive violence in response to ongoing abuse (Christie et al., 2022: 21).

THE (GREATER) SALIENCE OF COERCIVE CONTROL

Stark (2007) defines coercive control as a course of conduct involving multiple abusive behaviours, including physical assaults and sexual coercion. As abuse is experienced as continuous, harm is cumulative and not tied to a specific ‘incident’. When CARA was conceived, the concept of coercive control was not prominent in police policy and practice, despite having been central historically to (feminist) activist conceptions of DA. As stated previously, the initial trial of CARA pre-dated the introduction of the offence of Controlling or Coercive behaviour in an Intimate or Family Relationship, enacted in England and Wales in December 2015. Thus, there was always, in the absence of a course of conduct offence, potential for cases of entrenched coercive control, characterized by multiple non-physically abusive behaviours and/or non-injurious physical assaults, to be eligible for CARA.

Viewing DA through the lens of coercive control therefore problematizes the notion of minor or ‘low-level’ (incidents of) abuse. What appears ‘minor’ when shoehorned into administrative legal codes may not be reflective of a victim’s lived experience. It is therefore important that the gatekeepers for interventions such as CARA go beyond an overly simplistic framework of crime codes and ‘repeat’ incidents, especially in cases where there has been no previous contact with the police. As stated above, however, there is evidence from evaluations of CARA that processes for identifying suitable cases may not always pay due regard to ongoing coercion and control.

Cambridgeshire Constabulary has developed a two-stage tool for determining suitability for an out-of-court disposal which seeks to address the aforementioned issues around coercive control. The tool comes in two parts: the first is a coercive control screener which is completed with the victim; the second involves a detailed assessment of the offender by a specially trained team who examine all previous investigations, regardless of outcome, to identify patterns and characteristics which repeatedly appear in cases of domestic homicide. The coercive control screener goes beyond questions on the standard risk assessment to assess four domains: intimidation and threats, emotional abuse, economic abuse, and restricting freedom.

Cambridgeshire Constabulary also utilize a conditional caution alongside CARA that is aimed at offenders with more entrenched abusive behaviours. ‘Stepwise Relationships’ is a Ministry of Justice accredited programme delivered over 10 primarily group-based sessions by facilitators employed by the Probation Service. The programme employs cognitive behavioural techniques to help participants to identify how their learning history, thoughts and feelings can affect their behaviour. Individual sessions include what makes a healthy relationship, rules and expectations, communication, conflict management, and parenting/step-parenting. Contact with victims is maintained weekly by the force Out-of-Court Disposal Team for the duration of the programme, and all victims are referred to an Independent Domestic Violence Advisor (with approximately 40% engaging with the service). At the time of writing, the programme has been offered as a conditional caution for 6 months and there have been 41 referrals, with 36 deemed suitable for the longer intervention.

Thus, with Cambridgeshire’s tiered approach, the suitability and appropriateness of the intervention is based on in-depth analysis of the case, with offenders being referred to the most suitable intervention. Offenders displaying little or no evidence of coercive or controlling behaviour are placed on CARA, and those with low to moderate levels of coercive or controlling behaviour are sent for a suitability assessment to probation services for placement on the Stepwise Relationships programme. Cases in which offenders display high levels of coercion and control are deemed more suitable for an evidence-led prosecution.

An (anonymised) case example may be illustrative at this point, in part to highlight the type of case officers may deem appropriate for CARA, based solely on the crime recorded and the primary (first responder) risk assessment. A victim reported her ex-partner had assaulted her and damaged her mobile phone. The victim’s first account was captured on bodyworn video, but she indicated she would not support a prosecution. During interview, the man admitted destroying the phone but denied the assault. The officer in the case approached the out-of-court diversion team and requested a suitability assessment for CARA as the risk assessment had been graded ‘medium’. The coercive control screener suggested very high levels of coercion in all four domains; the second part of the test revealed that the suspect demonstrated behaviours which appeared on the domestic homicide timeline (Monkton Smith, 2020). The suspect was named in previous cases spanning multiple victims but had no previous convictions. He had utilized violence, strangulation and weapons and demonstrated controlling and coercive and stalking behaviours throughout these relationships. The case was declined and sent for an evidence-led prosecution with a specific request for a Stalking Protection Order and reclassification to ‘high’ risk.

It is important to emphasize that evaluation of the approach taken in Cambridgeshire is ongoing. We seek only to highlight the possibility and potential of a more nuanced approach.

CAN WE BE SURE CONDITIONAL CAUTIONS FOR DA ARE SAFE?

As discussed above, quantitative evaluations reliant on victims reporting further abuse to the police are limited in terms of what can be said about victim safety. There has been no longitudinal qualitative follow-up of victims whose abusers attended CARA workshops, save for a small study undertaken for a master thesis. Cornelius (2013) reported findings from telephone interviews with victims; 33 of the 216 participants had been subject to CARA. The structured interview schedule (Cornelius, 2013: 88) was phrased in terms of ‘incidents’ and ‘previous incidents’, and it is not clear how much interviewers probed around non-physical coercion and control. In addition, interviews took place around 4 months after the CARA workshops, as opposed to periodically over a longer follow-up period.

This gap in the evidence-base is significant, set against learning from Mirabal:

Short untested programmes run a number of risks, not least that they play into the instrumental orientation that many men have at the outset, and so are unlikely to address the deeper issues which matter to women in terms of their and children’s safety and the restoration of their voice and space for action (Kelly and Westmarland, 2015: 46).

While Cornelius (2013) provided evidence suggesting a short intervention may be appropriate for some offenders, it was apparent that some workshop participants perceived more entrenched behaviours in other attendees.

He is more open about himself and now talks about how he feels. The course scared him a little because he saw men that were really violent and he didn’t want to become like them (Cornelius, 2013: 69, emphasis added).

The victim follow-up showed—for all offenders, including those receiving non-conditional cautions—that, where there was a longer abuse history, victims were much less likely to report improved behaviour from their abuser and more likely to report deteriorating behaviour (Cornelius, 2013: 57). What extant evaluations have failed to explore is any potential (medium or long-term) backfire effect of a short intervention; specifically, are abusers whose behaviour is less amenable to change able to use completion of a short perpetrator intervention as a means to ‘gaslight’ victims or present themselves in a better light when engaging in ‘legal systems abuse’ (Douglas, 2018)?

DISCUSSION

There have been calls to dramatically increase the use of conditional cautions for DA by rolling out project CARA nationally (see Smith, 2022). Such calls have cited a ‘growing evidence-base’ which reflects satisfaction among victims, sustained behaviour change for offenders and supports wider adoption of the intervention. Yet we have highlighted that the current, published evidence-base consists of two studies that, while ostensibly strong designs, are limited by implementation issues, a reliance on police data and restrictive outcome measures. There is, to date, no published study demonstrating robust, longitudinal follow-up with victims and so the medium to long-term rehabilitative effect of CARA remains unclear. Indeed, the original CARA evaluation stopped short of recommending national roll-out. While Karavias et al. (2023: 11) suggest there is a ‘compelling case for CARA to be adopted nationally’, their evaluation highlighted issues around implementation, including potentially inappropriate referrals and failure to identify primary victims. That there may be problems with the implementation of conditional cautions for DA should not perhaps be surprising when considering findings reported by Westmarland et al. (2018) relating to police implementation of restorative justice in cases of DA.

Yet without an option for a conditional caution or alternative disposal the reality is that a greater number of cases will result in NFA, or a criminal charge not supported by the victim. Forces currently operating CARA suggest the former is more likely, due in part to resourcing pressures both in policing and in the CPS and courts. But is offering a potentially unsuitable intervention on a ‘probably better than nothing’ basis justifiable based in part on wider system failure?

A pragmatic way of navigating a route forward would seem to be continuing to grow the evidence-base for out-of-court disposals while addressing some of the implementation issues highlighted in the University of Birmingham evaluation. First, longitudinal, qualitative follow-up with victims (and ideally offenders) is crucial to improve understanding of behaviour change and any potential backfire effects of a short programme. This type of follow-up is resource intensive and logistically challenging. It is not impossible, however. One option is utilizing service providers and wider support services, as with the evaluation of the DRIVE programme (Hester et al., 2019). This approach is dependent though on meaningful engagement with and support for victims and the University of Birmingham evaluation highlighted differences between pilot sites in relation to victim engagement (see Christie et al., 2022). In Hampshire, where CARA originated, victim contact was undertaken by the service provider. In the other force, victim contact was the responsibility of neighbourhood policing teams, with workshop facilitators acknowledging there was ‘very little wraparound support for the victims’ (Christie et al., 2022: 32). Using police officers for victim contact was described as ‘challenging’, especially if this task was one among many for officers who had no specialist knowledge of DA or personal connection to the case (Christie et al., 2022: 32).

As a wider point, implementation issues become more salient when ‘scaling up’ an intervention. The evaluators of Mirabal emphasized the importance of skilled facilitators, a finding supported by Rowlinson (2015) based on observations of the delivery of CARA workshops. There is a danger that much broader roll-out of CARA results in unsuitable or less experienced providers chasing lucrative commissions, with an inevitable dilution of quality and programme effectiveness.⁷ The issues of victim-survivor support and multi-agency working are especially important. Integrated support for victims and a commitment to working with partners is a part of the Respect Standard for accredited programmes.⁸ Without these elements being mandated, there is a danger they are diluted, as the Birmingham evaluation showed. Support sector practitioners would suggest a co-ordinated community response is most effective in holding perpetrators to account, whatever the length of intervention.

In relation to the issue of inappropriate referrals, it ought to be possible to strengthen screening processes to take greater account of controlling and coercive behaviour and the degree to which the offender’s abusive behaviour is already entrenched. Without robust screening processes, operationalized by people with specialist knowledge of DA and coercive control, we believe there is a danger that a conditional caution becomes the easy or ‘go to’ disposal, in direct contradiction of the DPP’s assertion that the aim should not be to reduce the number of prosecutions (see Jarman, 2011: 10). While there will be cases where a criminal charge does not benefit victim or offender, recent research (Myhill et al., 2022) has suggested forces are underrecording and underutilizing the offence of controlling and coercive behaviour. It is possible that victims may be more supportive of prosecution if their abuser is arrested and charged with an offence they feel captures the entirety of the offending behaviour and attracts a meaningful sanction. Yet in the context of stretched resources (for both the police and the CPS), it is not unreasonable to suggest conditional cautions may be favoured over a resource intensive investigation in some cases where the nature of the offending behaviour merits (evidence-led) criminal prosecution and a more intensive perpetrator intervention.

CONCLUDING REMARKS

There is a valid conversation to be had around criminal justice intervention in cases of DA. Conducting a criminal investigation to the point of charge is resource intensive, and not always the desired outcome for victims. Conditional cautions may be beneficial in cases where offending behaviour in not entrenched, and the ‘awareness raising’ CARA workshops may prompt some offenders to ‘self-refer’ to full perpetrator programmes (Cornelius, 2013: 69). Yet there is also evidence that offending behaviour is more likely than not to be entrenched by the time it comes to the attention of the police which begs the question is it appropriate to knowingly refer such cases to what is acknowledged by the architects of CARA as ‘perpetrator programme lite’ (see Jarman, 2011: 56–7)? CARA appears to have become synonymous with conditional caution in discourse, yet it has proven possible in one force to undertake a thorough assessment of needs and to provide an alternative, more intensive, condition. There are precedents for DA interventions being introduced without robust evaluation and being found subsequently to suffer from issues around implementation, an example being the ‘DASH’ risk model (see Myhill et al., 2023). Given the potential for CARA to fill a gap in provision, it would be a shame if a rush to proclaim success resulted in overreach, unintended consequences, and a subsequent questioning of its suitability, even in cases where it may be effective.

Footnotes

Acknowledgements

We are grateful to Sara Kirkpatrick (Welsh Women’s Aid) and an anonymous peer reviewer for valuable comments that improved drafts of this paper.

REFERENCES