Attrition – Part 2

Commissioned by The Ministry of Women’s Affairs
Authors: Sue Triggs, Elaine Mossman, Jan Jordan and Venezia Kingi

Return to Responding to sexual violence research reports

Content

List of Boxes
List of Tables
List of Figures
Acknowledgements
Executive Summary
1    Introduction
2    Methodology
3    Profiles of victims, offenders and incidents

4    Attrition rates during the police investigation
4.1    Definitions
4.2    Overview of attrition rates during the investigation
4.3    Prosecution rate
4.4    Investigation outcome by offence
4.5    Victim characteristics by investigation outcome
4.6    Investigation outcome by police district
4.7    Suspect identified, but no charges laid
4.8    No suspect identified
4.9    ‘No offence’ and ‘false complaints’
4.10   Victim withdrawal
4.11   Attrition before recording
4.12   Summary of attrition during the police investigation

5    Attrition rates through the court process
5.1    Definitions
5.2    Overview of attrition rates during the court process
5.3    Conviction rates
5.4    Cases not convicted
5.5    Victim characteristics by prosecution outcome
5.6    Prosecution outcome by police district
5.7    Summary of attrition during the court process

6    Factors influencing outcomes
6.1    Overview
6.2    Approach
6.3    Factors tested
6.4    ‘False complaints’
6.5    No offence disclosed
6.6    Suspect identification rate
6.7    Suspect prosecution rate
6.8    Conviction rate
6.9    Summary of factors influencing attrition
6.10   Conclusion

Appendix A: Sexual offences included in this study

Appendix B: Literature reviewed

References

4.0 Attrition rates during the police investigation

 The focus of this chapter is the attrition rate as cases progress through the police investigation. Attrition rates through the court system, including conviction rates, are covered in chapter 5, while chapter 6 presents the results of a multivariate analysis of the factors influencing attrition throughout the process.
Chapter 4 begins with a note of definitions (section 4.1) and a summary of attrition rates at the police investigation stage for both cases (section 4.2) and offences (section 4.3). It is important that the summary results are not read in isolation.
In particular, there are several ways of calculating prosecution rates, a wide variety of case types within the broad outcome categories, and some cases that may be misclassified (sections 4.3, 4.4, and 4.7 to 4.10). The victim characteristics of each outcome category are examined in section 4.5, and attrition rates by police district are in section 4.6. The final section summarises key results.

As will be seen, most of the attrition within the criminal justice process occurred during the investigation phase – with charges laid in only three out of ten recorded cases. Where data were available, attrition points in this phase are examined in more detail, especially cases classed as ‘no offence’ and where the victim withdrew.

4.1    Definitions

The unit of measure is the case, which may be a single offence or a group of related offences (at least one of which was coded as sexual violation) against a victim documented in a single police file. For cases that relate to more than one offence, the outcome selected to represent the case is the one with the most serious outcome (i.e. the outcome that proceeds the furthest towards prosecution). Box 3 defines the outcome categories for the police investigation stage.

Box 3: Definitions: investigation outcome categories

 4.2    Overview of attrition rates during the investigation

Figure 6 summarises the stages of attrition during the investigation process for the 1,955 cases recorded as sexual violation in this study, based on the outcomes recorded in the police NIA database.

Figure 6: Attrition of cases through the police investigation stage

Notes: All percentages are of total recorded cases. Eight cases that had charges laid, but were subsequently found to be ‘false complaints’, were included in both the relevant categories in this flow diagram, but were treated as ‘false complaints’ for the rest of this chapter. The charges-laid group includes six cases where the offender was charged with a related offence and one case dealt with by Family Group Conference.

In summary, the overall case-based prosecution rate (the percentage of all cases in which charges were laid) was 31 percent. However, this overall rate is not the only way to estimate the prosecution rate, as highlighted in the following two sections. For example, almost half of all sexual violation offences resulted in a prosecution compared with less than a third of cases.

Thus, the majority of cases did not result in charges being laid against the alleged offender, as has been found in international research. For example, a meta-analysis of 75 studies found that an average of 35 percent of cases resulted in a prosecution (Daly and Bouhours, 2008). Prosecution rates may be even lower in Australia.
In New South Wales (Fitzgerald, 2006) and Victoria (OWP, 2006) charges were laid in 15 percent of cases. The latter study had a high rate of non-finalised cases (21 percent).

One of the main reasons for attrition studies is to examine what happens to the cases that do not result in a prosecution, because official statistics often code these cases into very broad categories such as ‘undetected’ or ‘not cleared’. As pointed out in HMIC and HMCPSI (2007: 10), ‘undetected’ is not a helpful classification because ‘the suspect was either known or identified following investigation in 80 per cent of cases classified as undetected’, a consequence of the low rate of rape by a stranger.

The findings of the current study in relation to cases that did not result in a prosecution were that:

• one in four cases did not result in charges being laid despite a suspect being identified (as discussed in section 4.7)
• one in nine cases did not proceed because no suspect was identified, with around two-thirds of these case files being closed (section 4.8)
• one in three cases was classified as ‘no offence’, including 8 percent documented as ‘false complaints’ (section 4.9).

Victim withdrawal was not included in the data set as a specific category, but further analysis suggested that in at least one in five cases the victim withdrew from the investigation process. These cases were mainly included within the ‘no offence’, ‘no suspect/file closed’ and ‘suspect/file closed’ categories, as discussed in section 4.10.

4.3    Prosecution rate

As shown in Figure 7, at least one offender was charged in 31 percent of all recorded cases.
The earlier New Zealand study (Stace, 1983) reported a prosecution rate of 33 percent for adult victims of rape or attempted rape, which was higher than the estimated rate of 26 percent for rape or attempted rape in the current study.

However, the current estimate of the overall case-based prosecution rate may be a slight underestimate, as some offenders may yet be apprehended and charged. In sections 4.7 and 4.8, it is estimated that as many as 124 case files that are open and active could result in charges being laid (although, after a lapse of six months or more, the actual impact is likely to be less). Excluding all these cases results in a case-based prosecution rate of 33 percent.

Official statistics in many countries (including New Zealand) exclude cases classified as ‘no crime’ or ‘no offence’. Therefore, prosecution rates calculated using official statistics indicate the likelihood of charges being laid in sexual violation cases that are identified by the police as valid offences. The prosecution rate in the current study was 46 percent, if all ‘no offence’ cases were excluded.

4.4    Investigation outcome by offence

As noted in section 4.3, 31 percent of cases resulted in at least one charge being laid (the case-based prosecution rate). Rape cases had a lower prosecution rate (26 percent) than cases involving other offences (43 percent to 35 percent) (as shown in the shaded column of Table 5). Rape cases also had higher rates of ‘false complaints’ and other ‘no offence outcomes’.

The prosecution rate for individual sexual violation offences was considerably higher than the case-based rate, with almost half of all offences resulting in charges being laid. Again, the prosecution rate was lower and the ‘false complaint’ and ‘no offence’ rates were higher for rape compared with rates for other offences.

Table 5: Investigation outcome by offence, for case-based and offence-based data

Notes: In the case-based analysis, the case is represented by the offence with the most serious outcome. In the offence-based analysis, all offences are represented, including multiple offences within a single case. USC = unlawful sexual connection. ‘Other’ offences include incest, inducing sexual connection and sexual exploitation of an impaired person.

The reason the offence-based prosecution rate (49 percent) was so much higher than the case-based rate (31 percent) was that cases involving multiple offences were much more likely to result in charges being laid for all offences in the case. Thus, for cases involving just one sexual violation offence, the prosecution rate was 20 percent. This increased to 71 percent for cases with two offences (which count as one case prosecuted but two offences prosecuted). The prosecution rate was 83 percent for cases involving three to five offences, and 98 percent for cases involving six or more offences.

Multivariate analysis (chapter 6) confirmed that the number and type of offences were among the most significant predictor variables for all types of outcome.

4.5    Victim characteristics by investigation outcome

Table 6 profiles the characteristics of victims associated with each investigation outcome category. Categories were grouped where sample sizes were small.

The relationship between the victim and offender showed substantial variation among the investigation outcome categories. As would be expected, cases with no suspect identified mostly involved strangers or people who had just met (within the previous 24 hours). In other such cases, either the victim generally would not name the offender or knew him only a little, or there was limited file information.

‘False complaint’ and ‘no offence’ cases had fairly similar profiles, with a low proportion of alleged offenders from intimate relationships. However, ‘false complaints’ were more likely to involve strangers than were ‘no offence’ cases.

Cases with a known or charged suspect had a more even distribution across the relationship types, with a lower proportion of stranger and just-met relationships than the other outcome categories and a greater proportion of intimate relationships. Strangers made up a higher proportion of cases in which the suspect was charged compared with cases where there was an identified suspect but the case was closed. The opposite was true for friends and acquaintances.

Complainants in ‘false complaint’ cases were significantly more likely to be aged 16 to 19 years than in other cases, whereas victims were less likely to be young in closed cases where there was a suspect but no charges laid.

Victims in ‘no offence’ cases were less likely to be Māori, as were victims in closed cases where there was a suspect but no charges laid, compared with all other ethnic groups combined. In contrast, Māori victims were over-represented in cases that resulted in a charge, as well as in cases that remained open with no suspect.

Cases recorded as ‘false complaints’ were more likely than other cases to involve a victim with a disability (mainly intellectual or psychiatric) or a victim who had made previous allegations of violent victimisation. The ‘no offence’ category also had a high proportion of victims with a disability, whereas the ‘no suspect’ categories had a low proportion.

In England, Kelly et al. (2005) also found over-representation within the ‘false complaints’ category of victims who were young, had mental health issues, had made previous allegations, and did not know the offender well.

The victim categories are not independent of each other. For example, fewer young people are in partner relationships. Therefore, in order to examine which of these and other factors are the best predictors of outcome, a multivariate modelling technique was used (chapter 6). The results largely confirmed the findings noted above, although there were no significant differences between ethnic groups once other factors were taken into account.

Table 6: Victim characteristics by investigation outcome (%)

Note: The ‘false complaint’ group (and to a lesser extent the ‘other no offence’ group) had a substantial proportion of missing data, especially for the victim-offender relationship (47 percent missing).

4.6    Investigation outcome by police district

Investigation outcomes are shown for each police district in Table 7. The significant differences relative to the average for other districts were:

• Auckland City had a lower rate of ‘no offence’ outcomes and a higher proportion of other outcomes that did not result in a charge, but no difference in the prosecution rate
• Canterbury had a higher rate of ‘false complaints’ and other ‘no offence’ outcomes and a lower prosecution rate
• Counties Manukau had a lower rate of ‘no offence’ outcomes
• Southern had a higher rate of ‘no offence’ outcomes and a lower prosecution rate and ‘no suspect’ rate
• Tasman had proportionally fewer cases with no suspect and more with a suspect where no charges were laid
• Waikato had more cases with no suspect
• Waitemata had a lower rate of ‘false complaints’ and other ‘no offence’ outcomes and a higher prosecution rate.

However, these differences were not significant when the effect of other factors was taken into account using multivariate modelling (chapter 6), with one exception – Auckland City had a significantly lower ‘no offence’ rate even after differences in case type between districts were taken into account.

Table 7: Investigation outcome by police district (%)

4.7    Suspect identified, but no charges laid

In 472 cases, or 24 percent of all cases, a suspect was identified but no charges were laid.

• Offender warned (39 cases or 2 percent of all cases). In almost half of these cases the warning reflected the victim’s wishes, either that she wanted the offender warned or trespassed rather than prosecuted (nine cases) or that she did not wish to proceed with the case (eight cases). A further 12 cases were not proceeded with due to insufficient or conflicting evidence or consent issues, and the other cases had no clear reason specified.
  

• Suspect identified but no charges laid and file still open (108 cases or 5 percent of all cases). In up to 92 cases charges might yet be laid, because the case was still being investigated (39 cases) or the offender had not yet been located (ten cases) or the investigation status was unknown (43 cases). Charges appeared to be unlikely in the remaining cases, due to lack of victim recall (three cases), insufficient evidence (seven cases) and the victim not wanting to proceed (six cases).
  

• Suspect identified but no charges laid and file closed (328 cases or 17 percent of all cases). The reasons for the closure of these files were not clear for 95 of these cases, with the remaining 233 cases reflecting a variety of reasons.

-    Victim withdrawal (117 cases). The victim indicated that she did not want to proceed or would not co-operate with the investigation (see section 4.10).

-    Insufficient or conflicting evidence (112 cases). In at least 17 of these cases the veracity of the complainant was questioned, in 11 cases the victim had little recall of the incident, and in ten cases the victim’s significant mental impairment was an issue. In the remaining 73 cases, there were issues of consent (at least 22 cases) or conflicting evidence or insufficient evidence to proceed.

-    Other reasons. The offender was not charged due to youth (two cases) or mental condition (one case), or the victim committed suicide (one case).

The validity of comparisons with overseas studies was limited by the use of different outcome categories, including the separation of victim withdrawal from other outcomes. In England, the withdrawal rate varied between about one in three and one in ten of all incidents and cases with insufficient evidence to proceed accounted for about a quarter to a third of all outcomes (Feist et al., 2007; HMIC and HMCPSI, 2007; MPS, 2005; Kelly et al., 2005). The most detailed Australian study, in Victoria, (OWP, 2006) found that 85 percent of cases did not proceed, including 21 percent that were still ongoing, 15 percent that had been withdrawn and 46 percent that resulted in ‘no further police action’.

4.8    No suspect identified

In 220 cases, or 11 percent of all cases, no suspect was identified. Almost two-thirds of these case files had been closed.

• No suspect identified and file still open (82 cases or 4 percent of all cases). The status of these cases was unknown for 15 cases, while another 19 cases were noted as still being investigated. Thus, charges might yet be laid in some of these 34 cases. Charges appeared to be unlikely in the remaining cases, due to lack of an identified suspect (38 cases), insufficient evidence that an offence occurred (five cases) and the victim withdrawing (five cases).

• No suspect identified and file closed (132 cases or 7 percent of all cases).

-    In 19 of these closed cases there appeared to be issues with conflicting or inconsistent evidence. The file notes indicate that seven of these cases were possible ‘false complaints’ and all had their validity questioned.

-    In 45 of the closed cases, the victim did not wish to proceed. Almost all these cases involved an offender who was a stranger or someone just met or an unknown relationship.

-    In 14 cases, the reason appeared to be that there was insufficient evidence to identify a suspect because the victim had limited recall about the incident, mostly due to a high level of intoxication.

-    For the remaining 60 closed cases, the reason the case had been closed was unclear. However, where there was any information recorded, it appeared there was insufficient evidence to identify a suspect. Where the offender’s relationship to the victim was known for these cases, almost all were recorded as strangers, with a few just-met relationships.

As noted above, the validity of comparisons with other studies was limited by the use of different outcome categories. In the earlier New Zealand study (Stace, 1983), 14 of 220 cases (6 percent) were considered to be genuine complaints with no suspect identified. Studies from England have found a ‘no suspect’ rate of about one in ten cases. Kelly et al. (2005) noted several cases where further investigation of case detail suggested this may not always have been the appropriate outcome (i.e. there appeared to be sufficient evidence to identify the offender but that information was not acted on).

4.9    ‘No offence’ and ‘false complaints’

4.9.1    Overview
The volume of cases that are classified ‘no offence’ (i.e. ‘no offence disclosed’ in
New Zealand or ‘no crime’ in England) has been the subject of a great deal of interest and research. The relatively high rate of cases classified as ‘no offence’ combined with confusion between the ‘no offence’ and ‘false complaint’ categories have contributed to the perception held by some people that many or most rapes are false allegations.

In the 1980s the ‘no offence’ category in England (Harris and Grace, 1999) and
New Zealand (Stace, 1983) accounted for almost half the total recorded cases. However, this was not due to a very high rate of valid ‘false complaints’, but rather was due to the use of the ‘no offence’ category as a ‘dust-bin category for the disposition of offences that do not fit in anywhere else’ (Stace, 1983: 5), including cases with insufficient evidence to proceed and cases of victim withdrawal.

Since then, police policy has aimed to restrict the use of the category to legitimate ‘no offence’ cases, which has resulted in a very significant drop in the ‘no offence’ rate in England, as discussed in the international comparisons section, below.

The ‘no offence’ category is appropriate for cases in which there is clear evidence that no sexual violation occurred and is made up of:

• false allegations or false complaints: incidents reported by the alleged victim that are fabricated

• offences recorded as a possible sexual violation that were not offences: for example, someone else reported what they perceived as an offence or the complainant was unsure whether violation had occurred

• offences not meeting the legal definition of sexual violation: for example, the reported incident was initially coded as a sexual violation, but after investigation was determined to be another offence (e.g. assault or abduction), the potential victim removed herself before any offence occurred or the complainant lacked an understanding of the offence due to intellectual disability

• offences recorded in error.

The ‘no offence’ category is not appropriate in circumstances where prosecution did not proceed because the victim withdrew, there was insufficient evidence, or no suspect was found.

The rate of ‘no offence’ categorising is also important because these cases are not recorded in official statistics, which can influence official estimates of prevalence and trends.

4.9.2    Results of this study
In the current study, just over a third of cases were classified by the police as ‘no offence’, including 8 percent of cases where the victim was charged or warned for making a false complaint and 4 percent of cases where the victim was suspected of making a false complaint.

Without seeing the full file details and talking to the victims and investigating officers, it is not possible to determine how many cases were inappropriately labelled as ‘no offence’ or ‘false complaint’. However, an attempt has been made to categorise cases based on what appeared to be considered the relevant circumstances for each case from the notes in the ‘comment’ fields.

4.9.3    ‘False complaints’, as charged or warned
The definition of a ‘false complaint’ in this study was that the alleged victim was officially sanctioned (i.e. charged or officially warned) for making a false complaint.

There were 157 such cases, or 8 percent of all cases. These could be further split based on what appeared to be considered the relevant circumstances.

• The victim admitted that the complaint was false (60 cases).
• The evidence did not support the complaint (35 cases), including cases where the incident could not have happened as stated (e.g. the offender did not exist), or there were discrepancies with witness, medical, scene or other evidence, or significant inconsistencies in the victim’s statement.
• Someone else made the complaint (nine cases) and the victim said it was not an offence or refused to co-operate with the police. Therefore, while these cases may be a ‘no offence’ or a false accusation to a third party, they are not false complaints to the police.
• The victim had a history of making allegations (ten cases), although this might not be the primary reason the current complaint was considered false (i.e. presumably the evidence did not support the complaint). In six of these cases, the file notes indicated the victim had an intellectual or a psychiatric condition.
• The victim had an intellectual or a psychiatric condition, with no other circumstances noted (11 cases). In total, victims had a disability in 42 of the ‘false complaint’ cases, but some of these are listed under other categories, such as ‘victim admission’.
• The reason for the case being documented as a ‘false complaint’ was unclear for the remaining 32 cases.

The perceived motives for making a ‘false complaint’ were generally not recorded in the database. Where there was an indication of motive in the file notes, the two most common perceived reasons were that the complaint was made to cover up consensual sex (18 cases) or the complainant wanted to get back at the offender or was jealous of, or upset with, the offender (19 cases).

The criterion of being warned or charged for a false complaint is a strict definition of the false complaint category, because it excludes cases that the police suspected to be false complaints. Therefore, it could be argued that the true false complaint rate could be somewhat higher than the estimated 8 percent. On the other hand, research based on victim interviews indicates that even some cases in which the alleged victim was warned or charged may be, in fact, real offences (Jordan, 2004a).

4.9.4    Suspected false complaints

A further 86 cases (4 percent of all cases) were officially classified as ‘no offence’ and the file information indicated that they were suspected false complaints, but they did not result in the victim being officially warned or charged.

• In nine cases, the complainant admitted the report was false or that sex was consensual.
• In 22 cases, the victim withdrew from the investigation. In half these cases the file information noted significant discrepancies in the complainant’s account or evidence not supporting the complaint. The remaining cases were retracted (four cases), the complainant refused to co-operate (three cases), someone else had reported (two cases) or the victim’s credibility was considered an issue due to mental health issues or repeated allegations (three cases).
• In the remaining 55 cases, the investigation concluded that no offence was committed. In 17 cases, no evidence was found that the alleged incident had occurred; in 12 cases, investigations indicated sex was consensual; in 16 cases, the evidence did not appear to support the complainant’s account; in three cases, the complainant had mental health issues; in four cases, the complainant was uncooperative (e.g. would not provide specific details or have a medical); and in three cases, the reason was unclear.

4.9.5    Other cases classified as ‘no offence’

Other than officially sanctioned false complaints and suspected false complaints, there were 424 cases (22 percent of all cases) that the police classified as ‘no offence’.

While these cases cannot be classified with any degree of certainty given the data limitations, they appeared to fit into the following groups.

• In 121 cases, the validity of the complaint could not be established because the complainant was unwilling or unable to co-operate (e.g. would not or could not provide details of the incident, or refused a medical examination or could not be contacted) or withdrew from the investigation.

-    In 43 cases, the alleged victim was uncooperative. In 21 of these cases, someone else reported the offence or the victim felt pressured to report.

-    In a further 13 cases, the victim withdrew and there was limited other information in the file notes.

-    In 65 cases, the complainant had limited or no recall of the incident and was uncertain whether an offence had occurred. This category includes eight cases where the victim took illegal drugs, five cases where the victim was on prescribed medication, 23 cases where the victim was intoxicated and seven where the victim thought she had been drugged or her drink spiked. Given that consent cannot be granted by a person under these circumstances, it is likely that at least some of these complainants were the victim of an offence.

• In 48 cases, intercourse occurred but the circumstances suggested lack of consent as it is legally defined (as outlined in section 1.5).

-    In 23 cases, the file notes indicate that sex was initially consensual, but at some stage the victim wanted to stop. These are offences, as the law recognises that consent does not extend to circumstances where the victim ‘allows the act because he or she is mistaken about its nature and quality’. These included cases where the sex became too rough, a condom was not used or broke, or the activities went beyond what was agreed or acceptable (e.g. the victim did not want anal sex).

-    In 25 cases, intercourse occurred and consent could not have been obtained due to the victim being intoxicated (ten cases), asleep when the assault started (12 cases), on medication (one case), too scared to say no (one case) or mistaken on identity (one case).

• In 84 cases, there appears to have been an offence committed, or at least no indication that the validity of the claim was doubted and no reason stated why the incident was classified as ‘no offence’.

-    In 56 cases, the victim did not wish to proceed with the investigation. For the 44 cases where the relationship was recorded, all but four involved offenders known to the victim, including 14 partners or ex-partners. Five of the alleged offenders were recorded as having gang associations.

-    In the other 28 cases, it appeared that the main issue was that consent was disputed or the evidence was insufficient to establish a case, with no clear indication that the police doubted the complainant.

• In 110 cases, the ‘no offence’ classification appeared appropriate.

-    In 42 cases, sex was found to be consensual, mostly based on the complainant’s admission.

-    In 16 cases, the complainant had limited intellectual capacity, but was either considered capable of consent (although in at least two cases the ability to consent appeared borderline) or no offence occurred.

-    In 11 cases, it appeared that someone else had reported an incident but no offence had occurred.

-    In 26 cases, the evidence did not support the complainant or there were significant inconsistencies in her account.

-    In four cases, the complainant had made several other unsubstantiated allegations.

-    In 11 cases, no offence occurred because the victim removed herself from a situation before an offence was committed (three cases), or the complainant suffered from delusions (two cases) or the incident did not constitute an offence (six cases).

• In the remaining 61 cases, there was insufficient file information to attempt a classification.

In summary, 667 cases, or 34 percent of all cases, were classified as ‘no offence’, including 157 officially sanctioned ‘false complaints’, 86 suspected ‘false complaints’ and 110 cases where the ‘no offence’ category appeared to be appropriate. Thus, approximately 18 percent of all cases appeared to be appropriately classified as ‘no offence’ based on the available file notes.

A further 132 cases (7 percent of all cases) appeared to be offences and thus misclassified as ‘no offence’. This group made up 20 percent of ‘no offence’ cases, a similar proportion to that considered by Jordan (2004a) to be probably genuine complaints in her study of 164 ‘no offence’ cases.

The validity of the remaining 9 percent of cases that were classified ‘no offence’ could not be judged from the available data, but it is not unreasonable to suppose that some of the cases classified as ‘victim withdrew’, ‘no recall’ or ‘unknown status’ were actually offences. Thus, the actual ‘no offence’ rate may be between about one in five cases and one in four cases.

Moreover, it is possible that some of the cases that appeared to be legitimate ‘false complaints’ or ‘no offence’ as judged by the investigation, could in reality be offences. In particular, previous research based on victim interviews indicates that sexual violation victims may respond to the trauma of the offence and intensive questioning in ways that harm their case from a police perspective:

• victim self-doubt or denial may be perceived as evidence that no offence occurred
• withdrawn or aggressive responses to trauma may seem uncooperative
• exhaustion, fear, shock or distrust of the police can lead to early withdrawal
• a desire to conceal aspects of the incident (such as excessive drinking or
  drug-taking) may lead to inconsistencies in the evidence
• victims may feel that persistent questioning indicates disbelief, leading to withdrawal or even retraction.

Jordan (2004b) provided examples of such cases in New Zealand.

4.9.6    Comparison with other New Zealand studies
The proportion of cases officially classified as ‘no offence’ in the current study (34 percent) was substantially lower than the 45 percent recorded by the earlier New Zealand study (Stace, 1983). The earlier study also noted that the ‘no offence’ category included some complaints withdrawn by the victim and some in which there was insufficient evidence to proceed. After examining the 220 files in the sample, the researchers concluded that 15 percent showed substantial evidence of an unfounded complaint, while 21 percent were possibly unfounded but with insufficient evidence to be sure (usually because consent was disputed and there was minimal corroborating evidence) and 15 percent were withdrawn by the victim – results not dissimilar to the present study.

Almost half the unfounded complaints (i.e. no offence had occurred) in the earlier study were due to someone else reporting the incident, when the ‘victim’ said no rape had occurred. Other reasons for unfounded complaints were to cover up consensual sex to avoid a parent’s or other person’s negative reaction (seven of 26 cases) and due to psychiatric illness (three cases).

The earlier New Zealand study concluded that it would be useful to have a wider range of resolution categories to limit the misuse of the ‘no offence’ category. Some overseas jurisdictions, for example, have categories for victim withdrawal or insufficient evidence to proceed.

More recent research in New Zealand included a study of 164 police files where the outcome had been coded ‘no offence’ (Jordan, 2004a). Of these cases:

• 21 percent (34 cases) were considered to be genuine offences, but did not proceed because the suspect was not found (five cases), the victim withdrew the complaint (13 cases), no charge was laid (nine cases) or the prosecution was not completed (13 cases)
• 38 percent were possibly genuine but lacking sufficient evidence or with some discrepancies in evidence
• 33 percent were considered by the police to be ‘false complaints’
• 8 percent were acknowledged by the victim as ‘false’ (although often the victim had not made the complaint in these cases).

Based on the information in police files, the police were more likely to have doubts about cases when the victim was intoxicated, reporting had been delayed, the victim had an intellectual impairment or psychiatric disturbance, the victim concealed aspects of her behaviour (such as excessive drinking), the victim had made a previous complaint of rape or the victim had a prior relationship with the offender.
A combination of factors was more important than a single factor.

Alcohol was a feature of many ‘no offence’ cases in both Jordan’s (2004a) sample and the current study. While excessive alcohol use by the victim can increase the likelihood of an incident being a ‘no offence’ (due to uncertainty about whether violation occurred), it can also influence perceptions of the victim’s credibility. Jordan (2004a: 38) notes:

A drunk woman tends to be viewed as responsible for what happens to her, while a drunk man may be absolved of responsibility for what he does while ‘under the influence’.

The issues of past allegations and mental health problems, which are over-represented in cases classified as ‘false complaints’ (section 4.9), are also vexed. Many victims had made past allegations of violent victimisation, but crime surveys and other research confirm that repeat victimisation is the reality for some women (Lievore, 2003; Davis et al., 2006). Even a history of several unsubstantiated allegations does not mean that the current allegation is false. Likewise, many victims had intellectual or psychiatric disorders, which may make them both more vulnerable to victimisation and potentially less likely to be believed.

4.9.7    Comparison with international studies

The ‘no offence’ category varied widely between studies in England, but tended to be lower than the officially classified rate in New Zealand, at around 15 percent to 25 percent of cases (Kelly et al., 2005; Feist et al., 2007). The rate of ‘false complaints’ was similar to New Zealand, at around 8 percent to 10 percent (Feist et al., 2007; HMIC and HMCPSI, 2007; Kelly et al., 2005; MPS, 2005).

Recent studies in England that have investigated ‘no offence’ incidents (and false complaints) have found that these categories still often include misclassified incidents, such as cases where there is insufficient evidence to proceed, an unverified suspicion that the complaint may be false, or where the victim has withdrawn a genuine complaint. For example, Feist et al. (2007) found that 15 percent of incidents were recorded as no-crime, but that 3 percent were incorrectly recorded, giving an actual no-crime rate of 12 percent. The HMIC and HMCPSI (2007) review included an attrition study of 752 cases, of which 24 percent were no-crimed, which dropped to 15 percent after adjustment for those incorrectly recorded.

Variation in ‘no offence’ rates between and within studies in England reflects some variation in case type, but also substantial differences in practice between areas (Feist et al., 2007; HMIC and HMCPSI, 2007; MPS, 2005). There have also been very significant changes over time, with a marked decrease in no-crime rates since the 1980s due to stricter reporting standards (Feist et al., 2007; Harris and Grace, 1999).

In Victoria, Australia, the minimum false report rate was 2 percent, corresponding to the percentage of cases where the alleged victim was charged with making a false complaint or warned that she would be charged (OWP, 2006). However, in a further 8 percent of cases the police were confident or reasonably confident that the report was false. Seven of 17 documented false reports involved a victim with a disability or mental health issue and ten of the 17 were known to the police from previous allegations or previous convictions.

4.10    Victim withdrawal

4.10.1    Results of the current study

‘Victim withdrawal’ is not an official police investigation outcome category and was not a variable explicitly included in the data collection. However, the file notes indicated this was a common event, in at least one in five cases (385 cases). Victim withdrawal is also of considerable interest with regard to the attrition of cases, so it is examined in more detail in this section.

‘Victim withdrawal’ was defined as those cases where the file notes indicated that the victim either did not want to proceed or did not co-operate with the investigation.

From the file notes, the victim appeared to be uncooperative in 30 percent of these cases – 49 percent of ‘no offence’ cases and 15 percent of other cases. This group includes victims who could not be contacted, refused to give incident details or refused to make a formal complaint after the initial report.

The remaining 70 percent of cases were withdrawn or not proceeded with at the victim’s request. Reasons for withdrawing in these cases were unknown for almost half the cases. For the other 139 cases, the summarised file notes give an indication of what the police perceived to be the reason the victim did not want to proceed:

• the victim did not want the offender charged or wanted the offender warned or trespassed (37 cases)
• someone else reported the incident or the victim felt pressured to report (22 cases)
• the victim had limited recall of the incident and was uncertain whether violation had occurred (18 cases)
• the victim wanted to report the offence but take no further action (17 cases)
• the victim retracted the allegation (ten cases)
• the victim was not yet ready to proceed (six cases), felt afraid or threatened (six cases), couldn’t face the process (six cases), was seeking help or advice (three cases), forgave the offender (three cases), was concerned about family reaction (four cases), was concerned about her career (one case), or wanted to move on (one case); or the family resolved the matter (three cases).

Based on the available information it was not possible to know whether the investigation process played a role in the victim’s decision to withdraw (i.e. withdrawal or refusal to co-operate may be due to the victim feeling as if she is not believed or that the case was unlikely to proceed on the available evidence).

The age and ethnic profiles of victims who withdrew were not significantly different from those of other victims. However, there were significant differences in the victim’s relationship with the offender (where this was known). The offender was a stranger in only 9 percent of cases where the victims withdrew compared with 17 percent of other cases. Conversely, the offender was more likely to be an ex-partner (13 percent compared with 9 percent) or a boyfriend (6 percent compared with 3 percent) where the victims withdrew, compared with other cases.

Compared with other victims, victims who withdrew were less likely to have a disability (10 percent compared with 16 percent) and more likely to have a criminal record (26 percent compared with 21 percent). Victims who withdrew did not differ significantly in their history of previous allegations of violent victimisation.

4.10.2    Comparisons with other research

In the previous New Zealand study (Stace, 1983) the rate of victim withdrawal was 14 percent, which is lower than the minimum estimate of 20 percent withdrawn in the present study. However, this may partly reflect differences between the samples (such as different profiles by age, offence and victim-offender relationship). Over half the withdrawn cases in the earlier study were reported by a person other than the victim. The victim’s reason for withdrawing, where known or implied, included reluctance to go through the investigation or trial process, a negative response from police (disbelief or being told of the low chances of conviction), concern about other people’s reactions, and fear of retribution.

In England, the withdrawal rate varied between about one in three and one in ten cases (Feist et al., 2007; HMIC and HMCPSI, 2007; MPS, 2005; Kelly et al., 2005; Lea et al., 2003). The extent of victim withdrawal varied widely between studies and between areas or police forces within studies. As in New Zealand, the vast majority of victim withdrawal took place during the investigation rather than after charges had been laid.

Estimates of the withdrawal rate varied depending on what was counted as a withdrawal – that is, an explicit withdrawal (the victim asks that the investigation be stopped) or an implicit withdrawal (the victim refuses to assist with inquiries or cannot be contacted). Inability to maintain contact with the victim may not be uncommon, in some areas at least. For example, police had difficulties contacting the victim in 37 percent of 677 cases in London (MPS, 2005).

The reasons for withdrawal also varied between studies, with studies based on victim interviews providing a different perspective from those relying on police files. The English studies cited above noted a variety of factors influencing a victim’s willingness to proceed:

• their experience with the investigating team, including the support and belief of the investigating officers, discussions around the likelihood of a conviction, and the availability of appropriate forensic services
• personal factors, such as fear of the offender, reconciliation with the offender, feelings of shame or wanting to move on, and family support (or lack of it)
• who initially reported the offence – for example, where family, friends or health workers reported against the wishes of the victim
• what the victim wanted to achieve – for example, to have the offender warned but not prosecuted or to clarify whether an offence took place in incidents in which recall was affected by drugs or alcohol.

Feist et al. (2007) found the two most commonly cited reasons were ‘not wishing to go through the investigative or court process’ and ‘wanting to move on’, followed by impact on the victim, friends or family, impact on the offender, and pressure from others to withdraw. Factors associated with withdrawal included quick reporting of the offence, victims who were uninjured, victims assaulted by partners or
ex-partners, and police force area. Feist et al. (2007) also noted that police forces with high withdrawal rates also tended to have low conviction rates.

In Victoria, Australia, 15 percent of cases were found to be withdrawn after inspection of case narratives, but around twice as many were recorded as ‘complaint withdrawn’ on the police database (OWP, 2006). The other cases should have been coded ‘no further police action’. Earlier studies in Victoria found that the withdrawal rate had increased from 14 percent in 1994/95 to 31 percent in 2002/03 (VLRC, 2003, as cited in OWP, 2006), based on the police data.

Cases involving current or former partners were more likely to be withdrawn (OWP, 2006). Other cases more likely to be withdrawn included cases involving female victims and older victims, cases without a medical examination, and cases in which the suspect was a known sexual offender.

4.11    Attrition before recording

This study, as for most attrition studies, was based on attrition of offences recorded by the police. However, it is worth reiterating that this ignores a significant point of attrition – whether the sexual violation offence is reported to (and recorded by) the police.

Based on various surveys across several countries, Daly and Bouhours (2008) calculated that an average of 14 percent of sexual violation incidents were reported, a rate not dissimilar to the estimated one in ten cases from the New Zealand Crime and Safety Survey 2006 (Mayhew and Reilly, 2007). In Australia, the reporting rate was estimated as 15 percent by the Women’s Safety Australia survey (ABS, 1996).

Reporting rates vary across surveys and estimation critically depends on how survey questions are asked, because victims’ perceptions of whether an incident meets the technical definition of rape may vary. Daly and Bouhours (2008) note that victimisation surveys in the United States and England and Wales show that more ‘non-traditional’ rapes (rape by an offender known to the victim) are now being reported to the police than in the past.

4.12    Summary of attrition during the police investigation

4.12.1    Prosecution rate

The prosecution rate was:

• 31 percent based on all recorded cases (i.e. charges were laid in 31 percent of cases)
• 49 percent based on all recorded offences (i.e. charges were laid for 49 percent of offences)
• 46 percent based on all cases classified as valid offences (excluding ‘no offence’)
• lower for rape than for unlawful sexual connection, attempted violation and other offences
• slightly lower than the rate from the earlier 1981 New Zealand study (Stace, 1983).

4.12.2    Other outcomes
Other outcomes of the police investigation stage accounted for 69 percent of outcomes for recorded cases (Figure 7), including ‘false complaints’ (8 percent), other ‘no offence’ (26 percent), no suspect identified (11 percent), and suspect identified but not charged (24 percent).
The two most common factors in cases with a known suspect that did not result in a prosecution were that the victim withdrew from the process or that there was insufficient or conflicting evidence. These cases were more likely to involve an offender who was a friend or an acquaintance compared with cases in which charges were laid.
The most common factors in cases with no identified suspect were that the victim withdrew from the process, there was insufficient evidence to identify a suspect who was a stranger to the victim, or the victim had limited recall due to intoxication.

Figure 7: Summary of investigation outcomes

4.12.3    ‘False complaints’ and ‘no offence’ outcomes

‘False complaints’ were defined as cases in which the complainant was charged or warned for making a false complaint. In about half the ‘false complaint’ cases for which further information was recorded, the file notes indicated that the complainant admitted the allegation was false. The remaining cases were identified as false based on evidence, lack of evidence, discrepancies and inconsistencies in the accounts given by the victim or other parties, or perceived victim credibility. ‘False complaints’ were more likely than other cases to be associated with offenders who were strangers and complainants who were teenagers or who had made previous allegations of violent victimisation or who had a psychiatric or an intellectual disability.

The ‘no offence’ category accounted for 34 percent of recorded cases (including the 8 percent designated ‘false complaints’), which was lower than the 45 percent ‘no offence’ rate from the 1981 New Zealand study (Stace, 1983). Research in England has also documented a decrease in ‘no offence’ cases, as police policy has directed this category be used only for valid ‘no offences’. However, recent research in England has still identified inappropriate use of the category for cases involving insufficient evidence or victim withdrawal.

Of 424 cases in the current New Zealand study that were documented as ‘no offence’ (excluding ‘false complaints’ and suspected ‘false complaints’):

• 132 cases (31 percent) appeared to be misclassified and were offences in that they involved intercourse but under circumstances that suggested lack of consent (48 cases) or the recorded file notes gave no indication that the validity of the claim was doubted (84 cases)
• 121 cases (29 percent) could not be established as either an offence or ‘no offence’ because the victim had limited recall of the incident (65 cases), withdrew (13 cases) or was uncooperative (43 cases)
• 110 cases (26 percent) appeared to be correctly classified as ‘no offence’, in that sex was consensual or no legal offence occurred
• 61 cases (14 percent) had insufficient information to attempt a classification.

Thus, overuse of the ‘no offence’ category still appears to be an issue in New Zealand. Based on the limited information available in the current data, the actual ‘no offence’ rate may be between about one in five cases and one in four cases.

Moreover, it is possible that some of the cases that appeared to be legitimate ‘false complaints’ or ‘no offence’ as judged by the investigation could in reality be offences. In particular, previous research based on victim interviews indicates that sexual violation victims may respond to the trauma of the offence and intensive questioning in ways that harm their case from a police perspective.

4.12.4    Victim withdrawal
At least one in five cases did not proceed due to victim withdrawal. That is, the investigating officer recorded that the victim did not want to proceed with the investigation, was uncooperative or could not be contacted. Withdrawn cases were more likely than other cases to involve an offender who was an ex-partner or a boyfriend. The most common reasons for not wanting to proceed, as recorded in the file notes, were:

• the victim wanted the offender warned or trespassed but not prosecuted
• someone else reported the incident or the victim was pressured to report
• the victim had limited recall of the incident
• the victim wanted to report the incident or seek advice but take no further action
• the victim did not feel able to proceed, was not ready to proceed or felt threatened
• the victim retracted the allegation.

5    Attrition rates through the court process

The focus of this chapter is the attrition rate as cases progress through the court process. Outcome definitions are presented in section 5.1 and a summary of the attrition pathway is presented in section 5.2. It is important that the overall results are not read in isolation from the broader context. In particular, there are several ways of calculating conviction rates and a variety of factors that influence conviction rates, as outlined in sections 5.3 to 5.6. Chapter 6 presents the results of a multivariate analysis of the factors influencing the attrition process.

5.1    Definitions

The unit of measure is the case, which may be a single offence or a group of related offences (at least one of which has been coded as sexual violation) against a victim documented in a single police file. For cases that relate to more than one offence, the outcome selected to represent the case is the one with the most serious outcome. Box 4 defines the outcome categories used in the analysis.

Box 4: Definitions: prosecution outcome categories-

Case flow refers to the series of events that progress a case to a resolution. The case can be discontinued at any stage by the prosecution, judge or defendant (by pleading guilty). Otherwise, the case will progress through the following stages (usually with several other hearings as well):

• charges laid by the police
• depositions hearing (a preliminary hearing to determine if there is a case to answer)
• committal – following depositions, the case is committed for trial and an indictment is laid by the Crown prosecutor
• trial (and sometimes retrial or appeals).

5.2    Overview of attrition rates during the court process

Figure 8 summarises the stages of attrition from recorded cases to disposition (outcome), focusing on attrition through the court process. The stage of the process, or case flow, is shown in the centre column. The outer columns summarise court outcomes. Detailed analysis of attrition at the police investigation stage is provided in chapter 4.

Figure 8: Attrition of sexual violation (SV) cases through the court process

Notes: All percentages are calculated as a percentage of total recorded cases. These results are estimates, because they include a correction for cases not yet finalised (see text). All boxes with single-line borders show the case flow and outcomes for the sexual violation offence with the most serious outcome in the case. Boxes with double-line borders show additional outcomes for other non-violation charges in the case. FGC = Family Group conference.

5.2.1    Correction for non-finalised cases
At the time of the data extraction (a minimum of six months after the offence was recorded by the police), 121 prosecutions were yet to be finalised. These cases could not be left out of the outcome calculation without biasing the overall results, as most of the cases in progress were awaiting trial and hence were more likely than the average case to be either acquitted or convicted at trial, and less likely to be withdrawn or discharged or resolved by an early guilty plea. Therefore, these cases were assigned an outcome based on the probability of outcomes for finalised court cases, correcting for the stage of the court process they had reached and the type of case.

5.2.2    Summary of case flow and outcomes
In summary, of the 1,955 cases in the study just under a third resulted in charges being laid (Figure 8). One youth offender was dealt with outside the formal court system at a Family Group Conference; otherwise, all charges were dealt with by the court system. Of the 600 cases dealt with by the Courts, 80 percent progressed at least as far as a depositions hearing and just over half went to trial.
Of the 600 cases in which an offender was prosecuted (i.e. charges were laid) the following outcomes occurred.

• In an estimated 42 percent of prosecutions (13 percent of all cases) at least one sexual violation charge within the case resulted in a conviction. A further 3 percent of cases resulted in a conviction for a related charge, bringing the total conviction rate to 16 percent. Four cases were successfully appealed (with more appeals yet to be heard). Further information on conviction rates is provided in section 5.3.
• In 26 percent of prosecutions (8 percent of all cases), the alleged offender was acquitted at trial. Half of cases that went to trial resulted in an acquittal, which is on a par with international studies (Daly and Bouhours, 2008).
• The case was withdrawn or discharged in 31 percent of prosecutions (9 percent of all cases). Non-conviction outcomes are discussed in more detail in section 5.4.
• Four cases resulted in other outcomes – three offenders were found unfit to stand trial and a fourth committed suicide, resulting in a stay of proceedings.

5.3    Conviction rates

5.3.1    Conviction rate based on cases

In an estimated 42 percent of prosecutions (13 percent of all cases), at least one sexual violation charge within the case resulted in a conviction. For cases that resulted in a conviction, the offender pled guilty pre-trial in 38 percent of cases and was found guilty at trial (or pled during the trial) in 62 percent of cases. Just over half of the guilty pleas were entered before depositions.

Four cases that were successfully appealed have been excluded from these convictions. However, the conviction is under appeal in several other cases (at least 12) and these have been included as convictions.

In addition, 62 cases, or 10 percent of prosecutions, resulted in a conviction on related charges of sexual offending (16 cases), violent offending (41 cases) or other offences (five cases). Thus, over half of all prosecutions (16 percent of all cases) resulted in a conviction. The overall conviction rate for any sex offences was 14 percent.

The additional convictions for sexual offences were all for indecent assault, except for two ‘assault with intent to commit rape’. The violent offence convictions were mainly for ‘male assaults female’ (26 cases), aggravated assault (seven cases) or other types of assault (six cases), sometimes in conjunction with other offences (threats, kidnapping and breach of a protection order), while the other offences were drugs (three cases) and dishonesty (two cases).

In at least ten cases, the file notes indicated that the convictions for a related sex offence were due to a charge substitution (i.e. the original charge was reduced to a lesser charge).

5.3.2    Comparisons with the earlier New Zealand study
In the 1981 New Zealand study, the conviction rate was 17 percent overall (or around 15 percent for adults), based on 220 cases of rape and attempted rape (Stace, 1983). In the current study the conviction rate for rape and attempted rape was lower (10 percent).

Stace (1983) reported that, of the prosecuted defendants:

• 47 percent were convicted on the rape or attempted rape charge (of whom 24 percent entered an guilty plea at or before depositions, 6 percent pled guilty at trial and 17 percent were found guilty)
• 17 percent were convicted on another charge
• 34 percent were not convicted (including 21 percent acquitted at trial)
• 2 percent were found unfit to plead.

5.3.3    Comparisons with international studies

Comparisons between studies were complicated by differences in the type of cases sampled and the definitions used. The conviction rate was usually defined as the percentage of incidents recorded by the police that result in a conviction. However, studies vary in their definition of ‘conviction’ and ‘recorded offence’. A conviction may include only a conviction for the violation offence, or a conviction for a lesser sexual offence or another offence. Official statistics of recorded crime usually exclude incidents classified as ‘no offence’, whereas most attrition studies include these in the base for assessing conviction rates.

Daly and Bouhours’ (2008) meta-analysis of 75 studies from five countries recalculated conviction rates with all recorded incidents as the base and distinguished between conviction types, where possible. Across all studies:

• the average overall conviction rate (including convictions for any sex offence) was 15 percent
• the conviction rate for the violation offence was 9 percent
• young victims had a higher overall conviction rate (18 percent) than adults (12 percent)
• the all-age overall conviction rate was 13 percent in the United States and Australia, 15 percent in England and Wales, 17 percent in Scotland and 21 percent in Canada
• the overall conviction rate decreased between 1970 and 1989 and between 1990 and 2007, from an average of 18 percent in the earlier period to 13 percent for the later period.

The most comparable figure to the present study was for adult victims, studies from 1990, and convictions for any sex offence within the sexual violation case, for which the average conviction rate was 11 percent (Daly and Bouhours, 2008).

The equivalent New Zealand rate appears to be comparable with or a little higher than this average, at 14 percent. However, some of the overseas studies in the meta-analysis were based on rape cases, which have a lower average conviction rate than other violation offences.

5.3.4    Conviction rate based on offences
The conviction rate estimates above were based on the total 1,955 cases recorded by the police (in which cases with multiple offences are represented by the offence with the most serious outcome). The rate can also be calculated based on all 2,888 individual offences recorded by the police (in which each offence is counted separately).

The case-based conviction rate for sexual violation of 13 percent was lower than the offence-based conviction rate of 17 percent (Table 8, conviction rates in shaded columns). That is, of all individual sexual violation offences recorded, 17 percent resulted in a conviction. This higher conviction rate was due to the higher prosecution rate for individual offences than for cases. As explained in section 4.3, the reason the offence-based prosecution rate (49 percent) was so much higher than the case-based prosecution rate (31 percent) was that cases involving multiple offences were very much more likely to result in charges being laid.

However, taken as a percentage of prosecutions, the case-based conviction rate (42 percent) was higher than the offence-based conviction rate (35 percent). That is, the probability of a conviction for any sexual violation offence within a prosecuted case was 42 percent, while individual offences had a 35 percent probability of conviction, once charges had been laid.

Table 8: Prosecution outcome by type of offence, for case-based and offence-based data (%)

Notes

1. In the case-based analysis, the case is represented by the offence with the most serious outcome.

2. Outcomes: Convicted for sexual violation; acquitted includes quashed; ‘other’ includes discharged, withdrawn and other outcomes of the prosecution process.

3. USC = unlawful sexual connection.

4. ‘Other’ offences include incest, inducing sexual connection and sexual exploitation of an impaired person.

5. In the offence-based analysis, all offences are represented, including multiple offences within a single case.

Rape offences had a lower conviction rate and a lower prosecution rate than other offences (Table 8). Rape offences had the highest acquittal rate as a percentage of prosecutions. ‘Other’ offences (inducing sexual connection, incest and sexual exploitation of a person with significant impairment) had the highest conviction rates and lowest acquittal rates.

5.3.5    Sentences imposed
Most convicted offenders received a prison sentence – either preventive detention (7 percent) or a standard prison sentence (88 percent – with an average imposed length of 5.7 years). A few offenders received home detention (4 percent) or supervision (1 percent).

5.4    Cases not convicted

5.4.1    Results of this study
The 349 cases in which an offender was prosecuted but not convicted accounted for 18 percent of all sexual violation outcomes or 58 percent of prosecution outcomes. These cases fall into four outcome categories:

• withdrawn: the police prosecutor may withdraw the charges at any stage up to committal, when the case is passed to the Crown prosecutor, who may also withdraw the charges (‘not proceeded with on indictment’)

• discharged/dismissed: the judge has the discretion to discharge the defendant at any point of the court process

• acquittal: a jury finds the defendant ‘not guilty’ at trial

• other outcomes: in this study four cases resulted in other outcomes (three offenders were found unfit to stand trial and a fourth committed suicide, resulting in a stay of proceedings).

It was not possible to identify the exact outcome for every case, but based on the available data on prosecutions an estimated

• 17 percent of cases were withdrawn
• 14 percent were discharged
• 26 percent were acquitted
• one percent resulted in ‘other’ outcomes (in addition to the 42 percent convicted).

The stage of the process reached was generally, although not always, recorded. Based on this information, the judge discharged or dismissed the case at trial in 39 percent of cases and pre-trial in 61 percent of cases where there was a discharge. The reasons for these discharges were not recorded.

An estimated 56 percent of withdrawn cases were withdrawn before depositions, 22 percent at depositions and 22 percent after depositions. The reasons for the charges being withdrawn were not stated for two-thirds of withdrawn cases. Where reasons were given, these related to:

• victim issues (17 cases): in three cases the victim admitted lying or retracted her statement, five refused to give evidence or would not sign the depositions, two were reluctant to proceed, one victim disappeared, one victim went overseas, and the rest did not want to proceed
• pleas/charge changes (ten cases) – seven defendants pled guilty to another charge in the case and three had the violation charge reduced to a lesser charge
• other reasons: new contrary evidence (three cases), case withdrawn by the Youth Court and resolved at a Family Group Conference (one case), case withdrawn following the suicide of the defendant (one case).

5.4.2    International comparisons
International comparisons were complicated by differences in the prosecution process between countries (and between states within countries). In many jurisdictions, police pass the file to the prosecuting agency at an earlier stage than in New Zealand, and some jurisdictions have no committal hearing. Therefore, the prosecution agency may have a greater role in attrition via its decision on whether there is sufficient evidence to go to trial.

Averaged across all relevant studies, Daly and Bouhours (2008) found that a third of cases in which charges were laid did not proceed past the prosecutor to Court, and over half of those that did proceed did not result in a conviction for a sexual violation offence. Thus, over two-thirds of prosecutions did not result in a conviction.

In England, estimates of attrition at the prosecutor stage vary from 2 percent to 18 percent of cases (Harris and Grace 1999; Kelly and Regan, 2001; Kelly et al., 2005; Feist et al., 2007). Feist et al. (2007) found that cases that the Crown Prosecution Service did not proceed with were more likely to involve partners or
ex-partners, adult victims, and victims who reported the crime themselves, and less likely to involve cases where a medical history was taken or the offender was linked to other offences.

Across five Australian states (Lievore, 2004), over a third of cases referred to the Director of Public Prosecutions (DPP) were withdrawn, either due to insufficient prospects of conviction (55 percent) or because the victim did not wish to proceed (45 percent). Most cases were discontinued before indictment. Cases were more likely to proceed where the victim was injured or the assault was severe, where the victim expressed non-consent and where there was additional evidence linking the defendant to the assault.

5.5    Victim characteristics by prosecution outcome

Table 9 profiles the characteristics of victims associated with each prosecution outcome category for sexual violation cases.

The relationship between the victim and offender showed substantial variation among the prosecution outcome categories. In cases that resulted in a conviction for a sexual violation offence, the defendant was more likely to be family or a stranger and less likely to be a current partner or a boyfriend, an ex-boyfriend, or a date than for other outcomes. Current partners were over-represented in the cases that were discharged or withdrawn, while strangers and family made up only a small proportion of this group.

Table 9: Victim characteristics by prosecution outcome (%)

Note

1. Outcomes: Convicted for sexual violation; acquitted includes quashed; ‘other’ includes discharged, withdrawn and other outcomes of the prosecution process.

The direction of these effects showed some critical differences from the investigation outcome results (Table 6). In particular, current partners were more likely to be prosecuted, but, if prosecuted, were less likely to be convicted. Conversely, strangers were less likely to be prosecuted (reflecting the lower likelihood of suspect identification), but, if prosecuted, were more likely to be convicted. Offenders who were family members had high rates of both prosecution and conviction.

Victims in cases that resulted in a conviction for sexual violation were significantly more likely to be aged 16 to 19 years than in other cases, whereas victims were less likely to be young in discharged or withdrawn cases. Victims aged 25 to 29 years were under represented in convictions. There were no significant differences between outcome profiles by ethnicity.

Cases resulting in a conviction were more likely than other cases to involve a victim with a disability (mainly intellectual or psychiatric), but less likely to involve a victim who had made previous allegations of violent victimisation. The prosecution rate was also lower for victims who had made previous allegations, giving this group a low overall conviction rate. In contrast, the prosecution rate was lower in cases where the victim had a disability, but if prosecuted, these cases were more likely to result in a conviction.

The victim categories were not independent of each other. For example, fewer young people were in partner relationships than were victims in the middle age groups. Therefore, in order to examine which of these and other factors were the best predictors of outcome a multivariate modelling technique was used (chapter 6). The model results confirmed that young victims and offenders who were strangers (alone and in combination) were associated with an increased conviction rate, while current partners and boyfriends had a lower conviction rate (but a high rate of associated convictions for assault). Victims with a disability also had a higher conviction rate.

5.6    Prosecution outcome by police district

Prosecution outcomes are shown for each police district in Table 10, as a percentage of cases in which an offender was prosecuted. Trials do not necessarily take place within the same police district, especially in the case of High Court trials.

The sample sizes were not large for many districts. The only statistically significant difference was that Tasman had a lower conviction rate than other areas, based on a small number of cases. When taken as a percentage of all recorded cases, the conviction rate was significantly higher in Bay of Plenty and lower in Tasman compared with other districts.

Table 10: Prosecution outcome by police district (%)

Note

1. Outcomes: Convicted for sexual violation; acquitted includes quashed; ‘other’ includes discharged, withdrawn and other outcomes of the prosecution process.

5.7    Summary of attrition during the court process

The conviction rate for sexual violation was estimated to be:

• 13 percent based on all recorded cases (or 14 percent if all convictions for any sex offence were included or 16 percent including convictions for any offence)
• 20 percent based on all cases recorded as valid offences (i.e. excluding ‘no offence’)
• 17 percent based on all recorded offences
• 42 percent based on cases in which an offender was prosecuted (i.e. cases in which charges are laid).
  The conviction rate was:
• lower for rape than for unlawful sexual connection, attempted violation and other offences
• lower than in the 1981 New Zealand study (Stace, 1983).

Other outcomes accounted for 58 percent of outcomes for prosecutions (Figure 9), including acquitted at trial (27 percent), discharged by the judge (14 percent), withdrawn by the prosecution (17 percent) and other outcomes (less than one percent).

Figure 9: Summary of prosecution outcomes

Current partners were more likely to be prosecuted, but, if prosecuted, were less likely to be convicted. Conversely, strangers were less likely to be prosecuted, but, if prosecuted, were more likely to be convicted. Offenders who were family members had high rates of both prosecution and conviction.

Victims in cases that resulted in a conviction for sexual violation were more likely to be aged 16 to 19 years or to have a disability, but were less likely to have made previous allegations of violent victimisation.

 6    Factors influencing outcomes

6.1    Overview

The aim of this analysis was to assess which factors were associated with or predicted the likelihood of conviction and other outcomes.

The analysis presented in previous chapters examined some associations between victim characteristics and outcomes, looking at each factor in isolation. However, victim characteristics are neither independent of each other nor independent of a range of other factors. For example, incidents of sexual violation in which the victim was aged under 20 years were also more likely to be committed by strangers. Therefore, the results presented in this chapter were based on a multivariate modelling technique (logistic regression), because this approach allows the major predictors to be identified from a range of possible factors within a single analysis.

Furthermore, for some attrition points in this analysis, especially ‘false complaints’ and other ‘no offence’ outcomes, the range of available data was restricted to victim and offence factors, with very limited information on alleged offenders and incident-related factors. Thus, victim characteristics will necessarily be major predictors in these models, raising significant issues of bias and potentially exacerbating stereotypes by limiting the focus to victim characteristics.

6.2    Approach

Instead of constructing a single model of conviction rates based on all data, a staged approach was used, with five models assessing the relevant factors at various attrition points (Figure 10). This approach was used because it better reflects the reality of the process. That is, to achieve a conviction, the case must pass through a series of potential attrition points: is it a false complaint? Is it an offence? Can a suspect be identified? Can charges be laid? Is the evidence strong enough to convict the defendant?

Figure 10: Diagram of attrition points that formed the basis of models

The staged approach also had the advantage of highlighting the differences between key factors found at different stages of the process. This was particularly important for factors that had varying effects at different points, such as the lower likelihood of suspect identification, but higher likelihood of conviction, for offenders who were strangers.

The staged approach was also useful for a more pragmatic reason, because data collection varied between stages. For example, no offender details were recorded for ‘false complaints’ or for many cases where the offender was not identified.

6.3    Factors tested

The factors tested as predictors included:

• offence – type and number of offences
• location/time – scene, time, day
• police district
• offender – gender, age, ethnicity and origin (New Zealand born, overseas born), criminal history, disability, alcohol and drug use

• victim – gender, age, ethnicity and origin (New Zealand born, overseas born), disability, criminal history, previous allegations, protection orders, alcohol and drug use, sex-worker
• circumstances – injury, force, threats, weapons, reporting time
• evidence – medical, witness, forensic, surveillance
• interaction terms – added as needed.

One of Jordan’s (2004a) conclusions was that a combination of negative or positive factors, rather than any one factor in isolation, might ‘tip the balance’ for or against the complainant. To test this, two derived scale variables were tested in the models:

• likely aggravating factors in favour of the victim (multiple offences, multiple offenders, multiple victims, offender with any criminal history, offender with previous sex convictions, offender with previous violent convictions, any type of force, injury or threat to victim, scale of injury)
• perceived negative factors against the victim (use of alcohol and drugs, intoxication, victim has a criminal history, victim has made previous allegations, reported in following month or later, victim refused medical, victim a sex-worker).

These variables were rather crude, in that they represent a simple unweighted count of the available factors. For example, a case in which the victim sustained a minor injury and had none of the other factors would score 1, whereas a seriously injured victim attacked by two offenders with previous violent convictions would score at least 9.

6.4    ‘False complaints’

The first model assessed which factors best predicted which complaints would be classified as ‘false complaints’, compared with all other cases. ‘False complaints’ were defined as cases in which the victim was charged or warned for making a false complaint.

Only a very limited range of factors could be included in this model because most variables were not recorded in the database for the ‘false complaint’ group. Thus, alleged offender and incident characteristics had to be excluded from the model. The factors that were tested included offence type, number of offences, police district and victim characteristics.

Two key variables (victim age and victim-offender relationship) were missing from 34 percent and 47 percent of ‘false complaint’ cases respectively compared with 5 percent and 27 percent of other cases. Therefore, two models were tested: Model A, based on only those cases where age and relationship were recorded, and Model B, based on all cases and excluding these two factors (Table 11).

Both models were weak overall, with only 11 to 13 percent of the total variation explained and poor predictive ability. The models were similar in that the number and type of offences and previous allegations of sexual victimisation were all significant predictors in both models. The odds ratio shows the relative effect and the direction of the effect, with an odds ratio of greater than one meaning that the factor is positively associated with the outcome and vice versa. Thus, rape was more often, and unlawful sexual connection was less often, associated with ‘false complaints’. Cases involving more than one sexual violation offence were much less likely to be classified as ‘false complaints’. Cases in which the complainant had made previous allegations of sexual victimisation were much more likely to be classified as ‘false complaints’.

Where the two models differed was that age and relationship were significant predictors when the relevant variables were included and victim disability was significant when age and relationship were not included in the model. The factor most significantly associated with ‘false complaints’ was age and relationship combined, rather than any age or relationship factor separately. That is, complainants aged 16 to 19 years who alleged violation by a stranger were very much more likely to be in the ‘false complaint’ category than other complainants. In the absence of the age and relationship factors, disability (mainly psychiatric or intellectual) was significant.

Overseas studies have noted similar results. For example, Kelly et al. (2005) found that ‘false complaints’ were more common among young adults, complainants with mental health issues, those who had made previous allegations, and incidents in which the alleged offender was not well known to the victim.

However, while these groups were over-represented among ‘false complaints’ in this study as well, it is important to emphasise that none was a strong predictor of a ‘false complaint’. For example, 80 percent of young victim/stranger cases and
two-thirds of disability/previous allegation cases were not classified as ‘false complaints’ and these groups made up only a quarter and a tenth of ‘false complaints’ respectively. It may be that factors other than those measured would be better predictors of ‘false complaints’, or simply that ‘false complaints’ are not easily predictable.

The models were also useful as an indication of what factors were not relevant. For example, once other factors were taken into account, the rate of ‘false complaints’ did not differ between police districts or by victim characteristics such as gender, ethnicity, and criminal history.

Table 11: Predictors of the probability of a case being documented as a false complaint

Model specifications:
Dependent variable: cases in which the complainant was charged or warned for making a ‘false complaint’ (coded 1), all other cases (coded 0).

Excludes cases not yet finalised. Model with age/relationship also excludes cases with missing data for victim age or victim-offender relationship.

An odds ratio of greater than one means the factor is positively associated with the outcome and vice versa.

Model A: N = 1,074, constant = -2.9, Nagelkerke R2 = 0.13.

Model B: N = 1,614, constant = -3.1, Nagelkerke R2 = 0.11.

6.5    No offence disclosed

The second model assessed which factors best predicted that a case would be classified as ‘no offence’, excluding ‘false complaints’. The analysis of ‘no offence’ cases in section 4.9 indicated that some of these cases might have been incorrectly classified. For the modelling, cases that were probably an offence were excluded from the ‘no offence’ group, while cases that were definitely, probably or possibly a ‘no offence’ were included in the ‘no offence’ group.

The ‘no offence’ category had a high proportion of missing data in relation to alleged offenders, so offender-related factors, other than offender-victim relationship, were excluded from the model. All other factors were included in the model but with indicator variables to test for missing data effects.
The resulting best-fit model explained a quarter of the total variation (Table 12), with five strongly significant predictors. Incidents involving any threat, force or injury had a low likelihood of being classified as ‘no offence’, whereas incidents in which the victim was uncertain as to whether violation had occurred were over-represented in the ‘no offence’ group. Cases that involved more than one offence were very unlikely to be classified ‘no offence’. Auckland City Police District had a significantly lower rate of ‘no offence’ than other districts.

The final predictor was a combination of victim-offender relationship and incident location. Incidents that occurred in a place other than a dwelling (most commonly a public place) and in which the offender was not an intimate (partner, ex-partner or family) had a higher probability of being a ‘no offence’ than other cases.

Table 12: Predictors of the probability a ‘no offence’ classification

Model specifications:

Dependent variable: cases classified as ‘no offence’ (coded 1), all other cases (0).

Excludes cases not yet finalised and ‘false complaints’.

An odds ratio of greater than one means that the factor is positively associated with the outcome and vice versa.

N = 1,462, constant = -0.87, Nagelkerke R2 = 0.24.

The interactions between relationship and location were complex. The results partly reflected the generally lower likelihood of a ‘no offence’ outcome where there was an intimate relationship compared with non-intimate relationships. In fact, the ‘intimate relationship’ variable was a significant but minor negative predictor in the model. But also, offences committed by non-intimates (especially strangers, ‘other known’ offenders and offenders with an unknown relationship to the victim) were less likely to be classified ‘no offence’ if the incident occurred in a dwelling than elsewhere. Conversely, offences committed by partners in a dwelling appeared more likely to be in the ‘no offence’ category, although this was based on a small number of non-dwelling cases and was not significant in the model.

There were also a number of minor predictors that explained a small amount of variation but were not consistently present in the various test models based on random subsamples of the data. The disability variable, either alone or in combination with previous sexual allegations, was fairly consistently present as a minor predictor, associated with higher ‘no offence’ rates.

6.6    Suspect identification rate

The third model assessed which factors best predicted that a suspect was identified, for cases that were classified as offences. Cases for which no suspect was identified had a high proportion of missing data in relation to alleged offenders, so offender-related factors were excluded from the model except for victim-offender relationship. All other factors were included in the model, with indicator variables to test for missing data effects.

The resulting best-fit model explained 40 percent of the total variation (Table 13). The three strongest effects were all related to victim-offender relationship. By far the strongest predictor was that the offender was a stranger or someone whom the victim had just met within 24 hours, which was associated with a lower rate of suspect identification. This effect was particularly strong if the offence was committed in a public place.

Cases for which no victim-offender relationship was recorded were also less likely to have a suspect identified. (‘Offender-victim relationship’ was missing for 35 percent of no-suspect cases and 21 percent of cases with a suspect.) Half of these
missing-relationship cases were either withdrawn by the victim or suspected ‘false complaints’.

Cases involving more than one offence were more likely to have an identified suspect, as were cases in which there was supporting evidence (witnesses, DNA or other forensic evidence, or surveillance footage). Cases in which the victim refused a medical examination were less likely to have an identified suspect, as were rape offences.

Table 13: Predictors of the probability that a suspect was identified, of cases identified as offences

Model specifications:

Dependent variable: cases with suspect identified (coded 1), other cases (0).

Excludes cases not yet finalised, ‘no offence’ cases and ‘false complaints’.

An odds ratio of greater than one means that the factor is positively associated with the outcome and vice versa.

N = 1,088, constant = 3.68, Nagelkerke R2 = 0.40.

6.7    Suspect prosecution rate

The fourth model assessed which factors best predicted the probability of prosecution (i.e. whether charges would be laid), given an identified suspect.
All factors, including offender-related data, were included in this model, with indicator variables to test for missing data effects.

The resulting best-fit model explained 41 percent of the total variation (Table 14). Cases involving more than one offence were much more likely to result in a prosecution than single-offence cases. However, within the multiple-offence group, cases involving multiple offenders were less likely to result in a prosecution than cases involving multiple offences by one offender. The multiple-offender group was quite small and contained a higher proportion of withdrawn cases and cases involving consent issues compared with the single-offender/multiple-offence cases.

Table 14: Predictors of the probability that a suspect was prosecuted

Model specifications:

Dependent variable: cases with charges laid (coded 1), other cases (0).

Excludes cases not yet finalised, ‘no suspect’ cases, ‘no offence’ cases and ‘false complaints’.

An odds ratio of greater than one means that the factor is positively associated with the outcome and vice versa.

N = 923, constant = -0.39, Nagelkerke R2 = 0.41.

The ‘aggravating factors’ scale variable (described in section 6.3) was also a strong predictor of a case involving a prosecution. Cases with none of the aggravating factors (i.e. cases with just one victim, offender and offence, an offender with no criminal history and no force, injury or threat to the victim) had a suspect prosecution rate of 17 percent, compared with rates of 44 percent, 63 percent and 84 percent respectively for cases with scores of 1–2, 3–4 or 5+ aggravating factors. Each of the components was important, with multiple offences and threats also being in the model in addition to the scale variable.

Conversely, the scale variable that attempted to measure the perceived negative attributes of the victim had an odds ratio under one, indicating that these factors significantly decreased the probability of prosecution. Cases in which the victim had none of the factors, had a suspect prosecution rate of 61 percent compared with rates of 49 percent and 35 percent respectively for cases with scores of 1–2 or 3+ negative factors. The following factors within the scale variable were all associated with lower prosecution rates: the victim refused a medical, had a psychiatric condition, had made previous sexual allegations, was intoxicated or did not report promptly. Two other victim factors were not associated with lower prosecution rates: victims who had a previous criminal history or who were sex-workers.

Rape had a lower suspect prosecution rate than other offences.

Overseas studies have generally examined the probability of prosecution based on all cases. Nevertheless, the main findings are not dissimilar. In Victoria, Australia (OWP, 2006), charges were more likely to be laid in cases involving family members, young victims, male victims (based on a small sample), victims subject to multiple assaults or who were injured, and victims who had a medical examination. A strong relationship was found between charging and the offender’s prior history of sexual offending.

In England, Feist et al. (2007) found that the predictors of getting a case to Court and getting a conviction were that the assault was linked to a sexual offence against another victim, the victim’s medical history was obtained, the offender threatened the victim, forensic evidence was recovered, witnesses were present and the offence was reported promptly.

6.8    Conviction rate

The fifth model assessed which factors best predicted the probability of conviction for sexual violation, based on cases in which an offender was prosecuted.
All factors, including offender-related data, were included in this model, with indicator variables to test for missing data effects.

The resulting best-fit model explained 23 percent of the total variation (Table 15). Cases in which charges were laid against a current partner (married or de facto) or boyfriend had a much lower probability of conviction for sexual violation than other cases. The conviction rate was higher if the offender was a stranger or the victim was aged under 20, and particularly high for a combination of these two factors.

Of cases that were not convicted, half were withdrawn for cases in which the offender was a current partner or boyfriend, whereas the majority of non-convicted offenders who were strangers were acquitted. However, of current partners who were prosecuted but not convicted for sexual violation, almost half were convicted for another violent offence, especially ‘male assaults female’.

Table 15: Predictors of the probability of conviction for cases in which the offender was prosecuted

Notes

1. Set at 0 for neither factor, 1 for either factor alone, 2 if the offender was a stranger and the victim was aged under 20.

Model specifications:

Dependent variable: convicted for sexual violation (coded 1), other cases (0).

Excludes cases not yet finalised or not prosecuted.

An odds ratio of greater than one means that the factor is positively associated with the outcome and vice versa.

N = 465, constant = -0.86, Nagelkerke R2 = 0.23.

The ‘aggravating factors’ scale variable (described in section 6.3) was also a strong predictor of a case being convicted. Cases with none of the aggravating factors (i.e. cases with just one victim, offender and offence, an offender with no criminal history and no force, injury or threat to the victim) had a conviction rate of 30 percent, as a percentage of finalised prosecutions, compared with rates of 39 percent, 44 percent and 74 percent respectively for cases with scores of 1–2, 3–5 and 6+ aggravating factors. All aspects of the ‘aggravating factor’ composite variable were important, as can be seen from the conviction rates for each individual factor in Table 16. Weapon use was also present independently in the model, increasing the probability of conviction.

Conversely, the scale variable that attempted to measure the perceived negative attributes of the victim (alcohol and drug use, criminal history, previous allegations etc.) was not significant in the conviction rate model.

Victims who had an intellectual disability also had a higher conviction rate. Rape had a lower conviction rate than other offences.

Of the police districts only Tasman District appeared in any model. This district was significant in the full-data model (associated with a reduced conviction rate). However, as the effect was minor and not consistent in test models based on random subsamples, this variable was excluded from the best-fit model shown in the results table.

International studies have generally found similar results, although the key predictors vary somewhat depending on which outcomes are compared and which factors are included in the models. The factors most often associated with higher conviction rates in a range of international studies included cases involving violence or injury, suspects with a criminal history, offending by strangers, witness and forensic evidence, younger victims and victims of ‘good character’.

For example, Daly and Bouhours (2008) analysed 33 studies of adult or mixed-age victims, which examined the associations between various outcomes and a range of factors. The most frequently examined factors were victim-offender relations, victim injury and victim age, and the majority of studies also examined physical evidence, use of force and victim character. Their results indicate that:

In the early period, the factor evincing the most consistent effect on police and court decisions is the victim’s ‘good character’ (89%). Next is the presence of injuries (67%) and the suspect’s criminal history (58%). Close to half of observations show effects for stranger victim-offender relations (48%), use of force/weapon (47%), and witness or physical evidence (45%). ... In the later period, the picture changes significantly for two factors: the effect of victim’s ‘good character’ has decreased significantly (38%) as has stranger victim-offender relations (25%). The witness or physical evidence, injuries, and weapons factors increase somewhat (by 5 to 9 percentage points). ... Across the two time periods, the suspect’s having a criminal history has a positive effect on police proceeding and on court conviction. ... Of 43 observations, over four in ten (42%) show a positive relationship between younger aged victims and police proceeding or court conviction. [Emphasis added]

6.9    Summary of factors influencing attrition

In summary, a few key factors were the major predictors of outcomes, with the predictors varying between stages of the investigation and prosecution. Table 16 provides a summary of conviction and prosecution rates for these key predictors.
The main findings were as follows.

• Cases involving more than one offence were much more likely to proceed through all stages, while rape cases were less likely to proceed at almost every stage.
• Victim-offender relationship was a significant factor at almost all stages, although the analysis was complicated by missing relationship data for many cases. The significant predictors were strangers (or strangers in combination with other non-intimates) and current partners or boyfriends.

-    Attacks by a stranger were more often associated with ‘false complaints’ and ‘no offence’ cases and had a high attrition rate due to their association with unidentified suspects. However, if prosecuted, stranger attacks were much more likely to result in a conviction, giving these cases a relatively high overall conviction rate.

-    Current partners and boyfriends had a high prosecution rate but a very low conviction rate for sexual violation (although many were convicted of other violent offences).

-    Other non-intimate offenders had relatively high attrition rate at most stages and therefore a low overall conviction rate.

-    Conversely, family members had high prosecution and conviction rates.

• Young victims had a higher rate of ‘false complaints’ (for stranger attacks), but also a higher rate of conviction for cases in which an offender was prosecuted.
• Cases involving force, threats or injury were less likely to be classified as ‘no offence’ and more likely to result in a prosecution and conviction, especially if the injuries were serious.
• Offenders with previous sex or violence convictions were much more likely to be prosecuted and convicted.
• Suspects were less likely to be prosecuted if the victim had one or more of the following attributes: refused a medical, had a psychiatric condition, had made previous sexual allegations, was intoxicated or did not report promptly.
• A combination of several aggravating or negative factors in the case had more impact than single factors alone.
• Victims with a psychiatric condition or intellectual disability had a higher rate of ‘false complaints’, as did victims who had made previous allegations of sexual victimisation. However, prosecutions were more likely to result in a conviction if the victims had an intellectual disability.
• Cases were more likely to be classed as ‘no offence’ if the victim was uncertain violation had occurred.
• Suspect identification was more likely when there was witness or forensic evidence, but less likely if the victim refused a medical examination.

No significant differences in outcomes were found between police districts, once other factors had been taken into account in the modelling, with one exception – Auckland City had a low rate of ‘no offence’ cases.

The following factors were not predictors of any outcome:

• victim gender, ethnicity and origin (New Zealand born, overseas born), victim criminal history, victim a sex-worker
• offender gender, age, ethnicity and origin (New Zealand, overseas born)
• incident timing (day, time of day, year).

Table 16: Summary of conviction and prosecution rates for key factors

Notes: All rates are based on finalised cases, so figures may differ from those in chapters 4 and 5, which include non-finalised cases and have an adjustment for non-finalised court cases. Conviction rates are for sexual violation offences.

6.10    Conclusion

In summary, this study examined the attrition of sexual violation cases from recording of the offence by the police to final outcome. The study was based on all 1,955 cases recorded as sexual violation by the police over a 30-month period, making it the largest study in New Zealand and one of the larger studies undertaken anywhere. This study was also distinctive in that it was a nationally representative sample and in that it examined the factors associated with each stage of attrition.

Generally, the results appeared consistent with international studies, although valid comparisons were complicated by differences in definition, methodology and criminal justice processes.

The key results in relation to attrition were that:

• attrition rates were high – only 13% of cases resulted in a conviction for sexual violation
• the majority of attrition occurred during the investigation phase – charges were laid in just under a third of cases
• a third of cases were classified by the police as ‘no offence’ (including 8 percent where the complainant was charged or warned for making a false complaint), although at least some of these did appear to be offences based on the available information
• at least one in five cases did not proceed due to victim withdrawal
• attrition rates varied between cases, depending on factors such as the number and type of offences, the relationship between the victim and offender, the age and mental state of the victim, the criminal history of the offender and aggravating circumstances of the offending
• the effect of some factors, especially the victim-offender relationship, differed between stages of the justice process.

A significant limitation of the analysis was the lack of data on some potentially important variables and substantial levels of missing data for other key factors.

All data were derived from a data set that summarised the available file information stored on the police NIA database, which meant the available information provided only a narrow perspective on the complex set of circumstances that make up each incident.

Appendix A: Sexual offences included in this study

The study used a broad definition of sexual violation, including sexual violation as it is defined under section 128 of the Crimes Act 1961 (rape and unlawful sexual connection), attempted sexual violation (section 129) and other offences involving sexual violation (incest, inducing sexual connection, and sexual exploitation of a person with significant impairment). The following police offence codes were included.

Offence code and offence description

2653    Male Rapes Female over 16

2654    Husband Rapes Wife

2657    Unlawful Sexual Connection Female over 16

2658    Unlawful Sexual Connection with Spouse

2663    Attempt to Rape – Female over 16

2664    Attempted to Rape – Spouse

2667    Attempted Unlawful Sexual Connection – Female over 16

2668    Attempted Unlawful Sexual Connection – Spouse

2669    Other Attempt Commit Sexual Violation offences

2695    Unlawful Sexual Connection Male over 16

2698    Attempted Unlawful Sexual Connection Male over 16

2699    Other Sexual Offences against Male Victim

2713    Parent Incest Child – over 16

2716    Brother Incest Sister – over 16

2719    Other Incest

2829    Other Unlawful Sexual Intercourse

2851    Sex Exploitation Significant Impairment

2852    Attempt Sex Exploitation Significant Impairment

2853    Does Indecent Act – Person Significant Impairment

2645    Induce Sexual Connection

2649    Other Inducing Sexual Connection Offences

2972    Induce/Compel to provide sex services

Appendix B: Literature reviewed

Research in New Zealand

The largest research project on rape in New Zealand (the Rape Study project) was undertaken in the early 1980s, as part of a programme to reform the laws relating to rape. The revised legislation was enacted in 1985, as outlined in section 1.5.

The Rape Study project included:

• a review of legislation and practice (Young, 1983)
• a survey of the experiences of 50 victims (Stone et al., 1983)
• a study of the processing of 220 cases by the police and interviews with 20 officers (Stace, 1983)
• a study of court processes and outcomes for 83 cases (Lee, 1983)
• a survey of judges and lawyers (Oxley, 1983).

A decade after the 1985 legislative changes, the Institute of Criminology began a follow-up research programme. This research generated published papers, books and a thesis on the subject of sexual violence against women, with a particular focus on the victim’s experience of reporting to the police (Jordan, 1998, 2001a, 2001b, 2004a, 2004b).

International research

Various aspects of attrition in sexual violation cases have been studied in a large number of international studies. The most comprehensive comparison of attrition is provided by Daly and Bouhours’ (2008) meta-analysis of 75 studies from five English-speaking countries.

The literature review undertaken for the present study combined key results from this meta-analysis with a more focused examination of recent studies from the two jurisdictions most similar to New Zealand – England and Australia. Full citations for these and other studies referred to are at the end of the report.

Some caution is needed in interpreting the results, given that the studies are based on a wide variety of locations, jurisdictions, sampling procedures and so on. All the reviewed studies were based on a sample of cases drawn from one or more specific locations; none was a nationally representative sample. Most used data from police databases, often supplemented by interviews with investigating officers and victims. The source of the data influences the breadth, depth, accuracy and perspectives of findings. For example, victim interviews give a different perspective to police records.

Comparisons between studies are also significantly complicated by differences in scope between studies. In particular, the age range of victims varied between studies, as did the type of offence included (i.e. rape, all penetrative offences or all sexual offences). For many studies it was difficult to determine exactly which offences were included. In addition, some studies base attrition calculations on all recorded cases, while others exclude ‘no-crimed’ offences.

Daly and Bouhours’ (2008) meta-analysis adjusted rates to include ‘no-crime’ outcomes and used a definition of rape that included all sexual violation (rape and unlawful sexual connection) and attempted violation as defined in New Zealand: ‘unwanted oral, anal, or vaginal penetration against consent through force, threat of force or when incapacitated’. However, studies within their sample varied in the offences included and, in particular, whether offences involving non-penile penetration and attempted violation were included.

In Australia, sexual assault legislation varies between states (Lievore, 2004: Appendix  B), but the offences defined are effectively equivalent to the New Zealand definitions. Lievore (2004) did not include attempted sexual assault, and this was probably also true for the Victorian study (OWP, 2006) as no mention was made of attempted rape.
The research based in England generally related to rape (with no mention of attempted rape being included), and hence was presumably restricted to penile penetration as defined under the Sexual Offences Act 2003 (HMIC and HMCPSI, 2007) and previous Acts.
Table 17 summarises the scope of the key comparison studies.
 

Table 17: Samples and scope of focus studies

References

ABS (1996) Women’s Safety Australia. Catalogue No. 4129.0. Canberra, Australia: Australian Bureau of Statistics.

Bazley, M. (2007) Report of the Commission of Inquiry into Police Conduct: volume 1. Wellington: Commission of Inquiry into Police Conduct.

Daly, K., and Bouhours, B. (2008) ‘Rape and attrition in the legal process: a comparative analysis of five countries’, unpublished, Griffith University, Australia. Available from http://www.griffith.edu.au/professional-page/professor-kathleen-daly/publications.

Davis, R., Guthrie, P., Ross, T., and O’Sullivan, C. (2006) Reducing Sexual Revictimization: a field test with an urban sample. New York: Vera Institute of Justice.

Feist, A., Ashe, J., Lawrence, J., McPhee, D., and Wilson, R. (2007) Investigating and Detecting Recorded Offences of Rape. Home Office Online Report 18/07. London: Home Office.

Fitzgerald, J., (2006) The Attrition of Sexual Offences from the New South Wales Criminal Justice System. Crime and Justice Bulletin 92. Australia: New South Wales Bureau of Crime Statistics and Research.

Harris, J., and Grace, S. (1999) A Question of Evidence? Investigating and prosecuting rape in the 1990s. Home Office Research Study 196. London: Home Office.

HMIC and HMCPSI (2002) Report on the Joint Inspection into the Investigation and Prosecution of Cases Involving Allegations of Rape. UK: Her Majesty’s Inspectorate of Constabulary and Her Majesty’s Crown Prosecution Service Inspectorate.

HMIC and HMCPSI (2007) Without Consent: a report on the Joint Review of the Investigation and Prosecution of Rape Offences. London: Her Majesty’s Inspectorate of Constabulary and Her Majesty’s Crown Prosecution Service Inspectorate.

Jordan, J. (1998) Reporting Rape: women’s experiences with the police, doctors and support agencies. Wellington: Institute of Criminology.

Jordan, J. (2001a) ‘True ‘lies’ and false ‘truths’: women, rape and the police’. PhD thesis, Victoria University of Wellington.

Jordan, J. (2001b) ‘Worlds apart? Women, rape and the police reporting process’, British Journal of Criminology, 41: 679–706.

Jordan, J. (2004a) ‘Beyond belief? Police, rape and women’s credibility’, Criminal Justice, 4(1): 29–59.

Jordan, J. (2004b) The Word of a Woman? Police, rape and belief. UK: Palgrave McMillan.

Kelly, L. (2001) Routes to (In)justice: a research review on the reporting, investigation and prosecution of rape cases. London: Child and Women Abuse Studies Unit, University of North London.

Kelly, L., Lovett, J., and Regan, L. (2005) A Gap or a Chasm? Attrition in reported rape cases. Home Office Research Study 293. London: Home Office.

Kelly, L., and Regan, L. (2001) Rape: the forgotten issue? A European research and networking project. London: Child and Women Abuse Studies Unit, University of North London.

Kingi, V., and Jordan, J., (with Moeke-Maxwell, T., and Fairbairn-Dunlop, P.) (2009) Responding to Sexual Violence: pathways to recovery. Wellington: Ministry of Women’s Affairs.

Lea, S.J., Lanvers, U., and Shaw, S. (2003) ‘Attrition in rape cases’, British Journal of Criminology, 43(3): 583–599.

Lee, A.J. (1983) ‘1980 and 1981 court files’, in Department of Justice (Ed.), Rape Study: research reports. Volume 2. Wellington: Department of Justice.

Lievore, D. (2003) Non-Reporting and Hidden Recording of Sexual Assault: an international literature review. Australia: Commonwealth of Australia.

Lievore, D. (2004) Prosecutorial Decisions in Adult Sexual Assault Cases: an Australian study. Australia: Office of the Status of Women.

Mayhew, P., and Reilly, J. (2007) New Zealand Crime and Safety Survey 2006: key findings. Wellington: Ministry of Justice.

Ministry of Justice (2008) Improvements to Sexual Violence Legislation in New Zealand: public discussion document. Wellington: Ministry of Justice

Mossman, E., Jordan, J., MacGibbon, L., Kingi, V., and Moore, L. (2009a) Responding to Sexual Violence: a review of literature on good practice. Wellington: Ministry of Women’s Affairs.

Mossman, E., MacGibbon, L., Kingi, V., and Jordan, J. (2009b) Responding to Sexual Violence: environmental scan of New Zealand agencies. Wellington: Ministry of Women’s Affairs.

MPS (2005) A Review of Rape Investigations in the MPS. London: Directorate of Strategic Development and Territorial Policing, Project Sapphire, Metropolitan Police Service.

OWP (2006) Study of Reported Rapes in Victoria 2000–2003: summary research report. Australia: Office of Women’s Policy, Department for Victorian Communities.

Oxley, P. (1983) ‘Results of questionnaires to the judiciary and lawyers’, in Department of Justice (Ed.), Rape Study: research reports. Volume 2. Wellington: Department of Justice.

Stace, M. (1983) ‘Rape complaints and the police’, in Department of Justice (Ed.), Rape Study: research reports. Volume 2. Wellington: Department of Justice.

Stone, J., Barrington, R., and Bevan, C. (1983) ‘The victim survey’, in Department of Justice (Ed.), Rape Study: research reports. Volume 2. Wellington: Department of Justice.

Taylor, N. (2007) Juror Attitudes and Biases in Sexual Assault Cases. Trends and issues in crime and criminal justice no. 344. Australia: Australian Institute of Criminology.

Triggs, S. (1997) Interpreting Trends in Recorded Crime in New Zealand. Wellington: Ministry of Justice.

VLRC (2003) Sexual Offences: interim report. Melbourne, Australia: Victoria Law Reform Commission.

Walby, S., and Allen, J. (2004) Domestic Violence, Sexual Assault and Stalking: findings from the British Crime Survey. Home Office Research Study 276. London: Home Office.

Young, W. (1983) Rape Study: a discussion of law and practice. Volume 1. Wellington: Department of Justice.

Disclaimer

This report was commissioned by the Ministry of Women’s Affairs. The views, opinions and conclusions expressed in the report are intended to inform and stimulate wider debate. They do not represent government policy.
 

Published in September 2009
by the Ministry of Women’s Affairs | Minitatanga Mō Ngā Wāhine
PO Box 10 049, Wellington, 6143, New Zealand
Phone: 0064 4 915 7112
Fax: 0064 4 916 1604
Email: mwa@mwa.govt.nz
Website: www.mwa.govt.nz

ISBN 978-0-478252-48-4 (Print)
ISBN 978-0-478252-49-1 (Digital)