Who uses domestic, family, and sexual violence, how, and why?

The State of Knowledge Report on Violence Perpetration reviews what is known about who perpetrates domestic, family, and sexual violence, how, and why, in order to enhance national efforts to end this violence.


The report's implications
Data collection on domestic, family, and sexual violence should include deliberate attention to perpetration -to the prevalence and character of violence perpetration, including its gendered and intersectional dynamics. Little is known about violence perpetration in Australia. National, population-based data on perpetration are a vital tool for violence prevention and reduction. Without consistent, comparable, and regularly captured data on perpetration, we are unable to be guided by evidence, to target interventions effectively to prevent and reduce perpetration, or to benchmark and measure the efficacy of our efforts.
Gathering data on the perpetration of domestic, family, and sexual violence is feasible. There is growing experience and expertise in how to collect robust and valid data on people's use of violence, including through largescale self-report surveys.
We need more and better data not only on the extent of violence perpetration in Australia, but on the perpetration's dynamics and drivers. This includes research on the factors that intensify the risks of perpetration or protect against it, perpetration among different populations and settings, and people's pathways both into and out of perpetration.
Significant proportions of the population have perpetrated domestic, family, or sexual violence. Looking at intimate partner violence, given that about 1.6 million women (17%) and 548,000 men (6.1%) in Australia have experienced physical or sexual violence from a current or previous cohabiting partner since the age of 15, 1 then in turn, large numbers of people are the perpetrators of this violence. Looking at sexual violence, studies in countries similar to Australia find that anywhere from one fifth to one quarter of young men have perpetrated sexual violence.
Most perpetration happens without ever coming to the attention of police or legal systems.
The vast majority of domestic, family, and sexual violence is committed by individuals who are not -and probably never will be -identified or sanctioned by the authorities. 2 Police and legal system responses must be safe for all victim-survivors, and there must be accountability for perpetrators. However, the focus and investment of efforts to reduce and prevent perpetration, hold perpetrators accountable, and provide justice for victimsurvivors must fall outside these systems to have lasting effects.
Domestic, family, and sexual violence have their roots in factors at multiple levels of society, including social and gender inequalities, social norms and the character of settings such as neighbourhoods, workplaces, and informal social networks.
Individuals use violence because, for example, they have learnt that behaving in coercive or abusive ways is normal or acceptable; they believe that such behaviour is expected in their social circles and settings; they have become invested in domination, control, and entitlement over their intimate partners or others; they have experienced violence themselves as children with impacts on their emotional and social development and attitudes; they have grown up in communities and contexts that normalise their use of coercion and abuse as part of sexuality or relationships; they are enabled by wider gender inequalities and other social inequalities; and/or, they expect to face few, if any, negative consequences for their actions.
Systemic and cultural change is necessary to lessen the risks of violence perpetration and victimisation. Violence prevention efforts to date have often focused on individual-level strategies and community-level strategies are a vital next step in prevention. 3 Changing individuals' attitudes and behaviours is unlikely to make widespread and lasting change to rates of violence when community and societal factors continue to reinforce violence. We must see systematic, multi-level, ecological approaches to the prevention of domestic, family, and sexual violence. 4 Most sexual violence perpetration starts young, in boys' and young men's teenage years, and then persists. We must address sexual coercion and other forms of violence, and the risk markers for these, early in young people's lives, including through strategies in primary and secondary schools. These include the universal provision of respectful relationships education and comprehensive sexuality education, complemented by interventions aimed at secondary and tertiary prevention and working with those who are already violent or who are at risk of using violence. We must develop strategies to respond to the large numbers of people, including young people, who are already perpetrating domestic, family, and sexual violence.
Prevention efforts should include intensified attention to the drivers of domestic, family, and sexual violence associated with masculinity, including violence-supportive and sexist masculine norms, peer cultures and settings. Prevention efforts should address cooccurring and overlapping patterns of violence perpetration and shared risk factors for them, rather than assuming discrete and siloed types of violence, [5][6][7][8] while also addressing the distinct risk and protective factors for specific forms of violence.
Many adult perpetrators of domestic and family violence were themselves victims of such violence as children. Childhood exposure to domestic, family, and sexual violence is a significant risk factor for later perpetration of violence. Therefore, it is vital to provide support and trauma recovery for children affected by domestic, family, and sexual violence to contribute to violence prevention.
It is possible to reduce the likelihood that individuals will start using domestic, family or sexual violence. While we need to know much more about the character and drivers of violence perpetration, we also need to know much more about the factors that protect against it. Protective factors are traits, behaviours and/or conditions that reduce or buffer against the risk of violence perpetration (and victimisation). Prevention efforts should identify and seek to increase the factors at the individual, relationship, community and societal levels that protect against perpetration.
It is also possible to reduce the likelihood that individuals at risk of or already perpetrating violence will start using, persist in or intensify their use of violence. The risk factors for perpetration include factors that are relatively static (such as childhood trauma histories and personality traits) and others that are relatively malleable (such as peer influences). Changing the more malleable risk factors, mitigating the negative influence of more static risk factors, and boosting protective factors can discourage individuals' use of violence or steer them away from the ongoing use of violence. 9 Prevention and response efforts should include attention to the personal and systemic changes that will encourage desistance from and the cessation of violence perpetration.

Domestic, family, and sexual violence are, fundamentally problems of perpetration.
Every act of violence involves a perpetrator and a victim, and it is time to increase our attention on perpetrators and perpetration. Perpetration, ultimately, is the problem we must solve if we are to end domestic, family, and sexual violence.
It is time to reframe the problems of domestic, family, and sexual violence to make the perpetrators more visible and more accountable. It is time to know much more about the extent and character of people's use of violence and about the social conditions that make this more or less likely. And it is time to use this knowledge to guide efforts to prevent and reduce domestic, family, and sexual violence.

The report's key findings
Data and knowledge on violence perpetration are limited and disparate. There should be far more scholarly attention given to the prevalence, dynamics, pathways and drivers of violence perpetration.
Nevertheless, existing research finds that the use of sexual and domestic violence is common. For example, sexual violence against women has been perpetrated by large minorities of men, including young men. Sexual violence has been perpetrated by around one in four men in studies among US university students and in community populations in other countries, by between one in ten and one in eight men in various European and other countries, and by close to half of men in some countries. In sexual violence, the use of verbal coercion and alcohol or drug-facilitated or incapacitated coercion is more common than the use of physical force. For example, North American studies of male university students find that about one in five have coerced someone into sex using verbal tactics (verbal pressuring, anger, threats to the relationship, etc.), while about one in 15 have used alcohol or drug incapacitation, physical force, or threats of physical force. 10 Studies of domestic violence, particularly those relying on self-reports of acts of physical aggression in relationships, also find that substantial proportions of men and women and young people have used aggression against an intimate or dating partner. The figures vary markedly, depending on the study and the kinds of behaviours measured, showing that anywhere from one in ten to one in three people has used physical aggression against an intimate partner. If studies of domestic and dating violence only ask people whether they have ever used various physically aggressive behaviours, they often find gender symmetry in the use of violence. If, on the other hand, they also ask about injury, fear, control, the context for the use of violence, and its history, they find gender asymmetries or contrasts, with males' perpetration of domestic and dating violence being more frequent, severe, injurious and harmful. Studies focused on self-reported perpetration of sexual violence typically find strong gender asymmetries, with far more perpetration by males than females.
Perpetration is driven by risk factors at the individual, relationship, community and societal levels. Although the use of domestic or sexual violence is influenced by a wide variety of factors at multiple levels of the social ecology, factors from macro to micro that receive consistent emphasis in explaining men's use of violence include: • features of societies, communities and neighbourhoods, particularly patriarchal structures and norms; • violence-supportive settings and contexts; • situational variables, such as separation and firearm availability; • sexist and violence-supportive peers; • violence-supportive and hostile masculine attitudes; and • witnessing or experiencing childhood violence.
There is considerable diversity among perpetrators and in perpetration. Different types of perpetrators and perpetration have different risk factors. Among perpetrators of the one type of violence (such as sexual violence), there are differences related to the severity, frequency, and form of their use of violence.
At the same time, there are also patterns of co-occurrence and overlap. Individuals who perpetrate one form of violence may also perpetrate others. In addition, the risk factors for different forms of violence perpetration overlap.
Perpetrators of sexual violence, and particularly male perpetrators, start young; typically committing their first acts of sexual violence as teenagers. Most individuals who commit sexual violence as young adults will continue to do so, especially if, like the vast majority of perpetrators, they avoid criminal detection and sanction. There is diversity in the trajectories of perpetration. Many young people's patterns of perpetration and nonperpetration of violence persist over time.
While a substantial proportion of people engage in violent and abusive behaviour, including acts that meet legal definitions of crime, very few encounter the legal system and fewer still receive any kind of formal sanction.

Introduction
Efforts to understand, reduce and prevent domestic, family, and sexual violence have often focused on victims and victimisation.
Assessments of the prevalence of domestic, family, and sexual violence typically focus on the extent of victimisation, much scholarship centres on victims and victimisation, and media and community accounts of domestic and sexual violence often focus on victim-survivors and the impacts of the violence they experience. Particularly in the context of the widespread silencing of victim-survivors' voices, there are good reasons to centre victim-survivors' lived experience and voices in research, policy and programming. 11 At the same time, it is vital to examine perpetrators and perpetration. As noted elsewhere: "It is time to reframe the problems of domestic and sexual violence in Australia: to put perpetrators in the picture and to focus more on preventing and reducing the perpetration of abusive behaviours". 12 Three tasks are urgent if we are to make progress in lessening domestic, family, and sexual violence.
This report contributes to the first of these tasks; identifying the state of knowledge about violence perpetration. It thus contributes to the broader goal of 'putting perpetrators in the picture'. This is vital to properly understand domestic, family, and sexual violence, lessen victim-blaming, hold perpetrators accountable, identify the social conditions that enable and encourage perpetration, and change these social conditions to make progress in ending domestic, family, and sexual violence. 12 This report maps existing research and knowledge on who uses domestic, family, and sexual violence, how and why. It concentrates on three major forms of interpersonal violence: intimate partner and family violence, sexual violence, and sexual harassment. The report also touches on a fourth form of violence: child sexual abuse. The report includes violence by men and women and others; violence in heterosexual, gay, lesbian and other intimate relationships; and sexual violence and sexual harassment. It focuses particularly on the extent and character of violence perpetration, and leaves aside such issues as its impact on victim-survivors. The report relies on a narrative review of Australian and international scholarly works on perpetration that have been published since 2010. The report is intended for stakeholders involved in violence prevention, including individuals and organisations in the violence prevention and health sectors, policy makers, researchers and advocates.

The Perpetration Project
This report is a product of the Perpetration Project, a national research project on the perpetration of violence in intimate, domestic and family settings in Australia. The Perpetration Project is intended to contribute to the reduction and prevention of domestic, family, and sexual violence in Australia. It is a collaboration among researchers and advocates from Queensland University of Technology (QUT), the Equality Institute and the Accountability Matters Project.

Terminology
The term "domestic violence" in this paper is used to refer to intimate partner violence; "family violence" refers to violence that occurs between immediate and extended family and kinship groups; and "sexual violence" is encompassing of sexual assault, rape, sexual harassment, sexual coercion and sexual abuse.
The term 'perpetrator' is used in this report to refer to any individual who uses violence, whether or not this violence is legally defined as criminal or has been subject to legal sanction. The term 'perpetration' is used for any behaviour that meets common definitions of domestic or sexual violence, rather than for specifically criminal acts. 13 This fits with much of the scholarship on the perpetration of domestic, family, and sexual violence, in that it examines the use of behaviours that are both within and outside legal definitions. The terms 'victim' and 'victim-survivor' are used interchangeably to refer to any person who has been subjected to violence.
Before spelling out what is known about patterns of violence perpetration, this report briefly outlines the state of research on perpetration.
We must: 1. generate data and knowledge on the perpetration of violence; 2. reorient prevention efforts to have a greater focus on preventing initial and ongoing perpetration; and 3. reframe domestic, family, and sexual violence in community, media and policy discourse as problems above all of perpetrators and perpetration.

Research on perpetration
There is a growing, although small, body of scholarship on perpetrators and perpetration.
This comes from diverse disciplinary and institutional locations, including community and population-based research, forensic criminology, psychiatry and psychology.
There are significant limitations to the volume, breadth and depth of existing research on violence perpetration. There has been far less research on perpetration than victimisation.
In Australia, while national data on victimisation are generated every four years from the Personal Safety Survey, there are no national data on perpetration. The same is true in other countries. In the USA, for example, while there have been a series of large-scale national or epidemiological studies on sexual victimisation, including six among university women, only one similarly large-scale study has been conducted on sexual perpetration in university men -35 years ago. 10 Overlapping with this, there has been less research on risk factors for perpetration than for victimisation. This is true for broad forms of violence such as domestic violence and sexual violence, and for these forms of violence among particular populations, for instance LGB populations. 14 The level of research on violence perpetration is highly uneven across different forms of violence, populations, institutional settings and geography: • Forms of violence: There has been far more research on the perpetration of domestic violence and sexual violence than the perpetration of sexual harassment. As the report notes further below, the scholarship on domestic, family, and sexual violence perpetration is substantial enough to allow a series of systematic reviews and meta-analyses, but this is not the case for sexual harassment.  19 For example, much of the research on sexual violence perpetration is based on samples from the USA. There has been very little research on perpetrators and perpetration in Australia. • Detected perpetrators: Research on perpetrators often comes from data related to the legal system, such as police statistics, criminal court data and prison census data. However, perpetrators who have pleaded or been found guilty represent only a tiny proportion of the population of perpetrators. 13 These 'detected perpetrators' are unlikely to be representative of the much larger pool of 'undetected perpetrators' in the community. Criminal legal system samples and data based on convicted offenders provide only a limited basis for understanding the beliefs, attitudes and behaviours of perpetrators, or the drivers of their perpetration. 13 There is far more research on risk factors for perpetration than on protective factors that make violence perpetration less likely. For example, in a systematic review intended to explore risk and protective factors for men's sexual violence against women at higher education institutions, none of the articles explicitly discussed protective factors and no two articles measured the same protective factors. 15 There are also significant limitations to the methodologies used in existing research on perpetration.
• Some widely used measures of violence focus only on counting violent acts and do not gather information on the intensity, frequency, context, consequences or meaning of these behaviours. The Conflict Tactics Scale (CTS), often used in research on domestic or intimate partner violence, typically neglects issues of injury and fear, omits sexual violence, ignores the history or context of the violence, and draws on samples potentially shaped by high rates of refusal, particularly among individuals either practising or suffering severe and controlling forms of violence. 20,22 The report returns to this assessment in the Appendix. Studies using the CTS usually find gender symmetries in the perpetration of domestic violence, but their omission of key dimensions of violence means these findings are questionable. Meanwhile, self-report studies using different methods for measuring domestic violence and other data sources, such as police and hospital data, often find strong gender asymmetries in the perpetration of domestic violence. • The methodologies used in research on violence perpetration are overwhelmingly quantitative and based on cross-sectional data. Very few studies use longitudinal data, as Duval et al. 16 note, specifically referencing studies conducted among university students. The absence of longitudinal data means that it is harder to identify the drivers of violence perpetration or the processes through which particular factors feed into an increased or decreased likelihood of perpetration. • Administrative data collected by police, hospitals and other services, provide an alternative source of information on violence. However, these data typically focuses on victimisation rather than perpetration and victimisation. Moreover, these data are limited by under-reporting of domestic, family, and sexual violence, professionals' poor understanding of domestic violence and poor responses to victim-survivors and perpetrators, and unjust and discriminatory systems and cultures. These limitations in existing data should be borne in mind when reviewing the patterns of perpetration described in the following section.
The report makes recommendations for future research on violence perpetration in Section 4, and provides a further examination of methodological issues in defining, measuring and analysing violence perpetration in the Appendix.
We need more research that identifies and understands the protective factors that help prevent and limit perpetration of violence to inform early interventions and work with perpetrators to desist using violence. Insights on protective factors can assist prevention efforts to reinforce the individual, community and societal characteristics that make perpetration less likely.

Patterns of perpetration
Who uses violence, and against whom? How do patterns of violence perpetration vary among different populations?
This section explores the extent to which different population groups use violence and the forms of violence perpetration most common among different groups. Scholarship on patterns of perpetration is uneven, with far more research on some groups and relationship dynamics than others, as this report noted in Section Two. This discussion does not reflect a systematic review representative of all groups and populations, but does highlight key findings from the scholarship.

Violence by men and by women women
The most significant and consistent finding in scholarship on violence perpetration is that most violence is perpetrated by men. This finding is evident from victimisation data, legal system data (from police and other institutions), and often from self-report data. This pattern is true for violence overall, and true for specific forms of violence including sexual and domestic violence. Some self-report studies, particularly those focused on domestic violence and using measures focused on physically aggressive acts, find gender symmetry with similar proportions of men and women using violence. As we note below, however, closer examination reveals gender contrasts in the severity of violence, the presence of coercive control, motivations for perpetration, and the impacts of victimisation.
We begin with victimisation data, that is, on victims' reports of who assaulted them.

Violence overall
Most violence and abuse are perpetrated by men, as illustrated by Australian data on violence overall:

Homicides
Close to nine in ten perpetrators of homicide in Australia are male. Men commit homicides at six to seven times the rate of women. 25 • 87% of homicide offenders from 2019 to 2020 were male 25 and 87% of homicide offenders from 1989 to 2020 were male. 25

Domestic violence
National victimisation data find that threequarters (75%) of all victims of domestic violence reported the perpetrator as male, while one-quarter (25%) reported the perpetrator as female. 26 Among specific populations, such as university students, victim data again document that most sexual assault is perpetrated by males.
• In a 2017 national survey of university students in Australia, more than four out of five students (83%) who were sexually assaulted in a university setting in 2015 or 2016 said that the perpetrator of the most recent incident was male. 30 › 92% of women reported that the perpetrators of the most recent incident of sexual assault were men only. › Men were most likely to have been sexually assaulted by men only (41%), and also reported incidents of sexual assault by women only (26%) or by both men and women (24%). 30

Sexual harassment
The majority of sexual violence and harassment is perpetrated by men. Women overwhelmingly experience sexual violence and harassment from men, whereas men (who experience a lesser rate of sexual violence and harassment than women overall) experience a more even split of male and female perpetrators. Among specific populations such as university students and school students, again, most sexual harassment is perpetrated by males rather than females.
• In a 2017 Australian national survey of university students, the majority of students (71%) who were sexually harassed in a university setting in 2015 or 2016 said that the perpetrator of the most recent incident was male. 30 › 86% of women reported that the perpetrators of the most recent incident of sexual harassment were men only. › Men were also more likely to have been sexually harassed by men only (37%), however some also reported experiencing sexual harassment by women only (30%) or by both men and women (22%). 30 Among school students in Australia: • A national study among 408 schools across Australia, both primary and secondary, found that, 'Boys are the usual perpetrators of sex-based harassment of their own and of the other sex, although some girls also harass both sexes.'34 (Unfortunately, a more recent study of sexual harassment among NSW school students did not include data on the sex of perpetrators.) 35

Self-report data on the use of violence
Self-report studies on the use of violence in general (rather than domestic, family or sexual violence in particular) are rare, but they find that greater proportions of males than females use violence. This is true, for example, in US studies of university students and adolescents: •

Domestic violence: self-reported perpetration
What about self-report data on the perpetration of domestic violence in particular? It should be noted, first, that many studies generating data on self-reported perpetration (and victimisation) only ask participants if they or their partners have ever committed any violent acts from a specified list (slapping, kicking, punching and so on) over the past 12 months. This is the case for the Conflict Tactics Scale (CTS), one of the most widely used measures for intimate partner violence. Many studies using the CTS do not ask about the impact of the violent acts (to do with fear or injury for example), do not examine the frequency or severity of the violence, do not ask about the intent of the violence, and neglect whether the violent acts were in self-defence and what the history of violence was. The CTS omits important forms of violence by intimate partners including sexual violence and stalking and omits incidents after separation and divorce. A 1996 revision to the Conflict Tactics Scale, the CTS-2, added items on injury and sexual coercion, but they are still limited and many researchers do not use them. 40 The CTS is also particularly vulnerable to 'false positives', that is, to the inaccurate inclusion in perpetration and victimisation rates of behaviours that in fact involve joking and horseplay rather than violence or abuse. 41 (We return to methodological issues in assessing violence perpetration in the Appendix.) This means that many studies purporting to measure domestic violence, intimate partner violence or dating violence are primarily measuring individuals' use of a range of physically aggressive acts. They do far less, however, to measure what many researchers and practitioners consider to be domestic violence in the 'proper' sense, that is, a pattern of power and control exerted by an individual over their intimate partner or former partner. In this latter view, domestic violence is seen to involve a range of physical and/or non-physical strategies of violence and abuse, taking place within an intimate or familial relationship and forming a pattern of coercive and controlling behaviour. 42 It is in this sense that researchers and others describe domestic violence in terms of 'intimate terrorism', 43 'coercive controlling violence' 22 or simply 'coercive control'. 44 Key differences between this view of domestic violence and the view implicit in the CTS are that in the former: • a wider range of physical and nonphysical behaviours are included; • these are seen often as intended to exert power or control and/or to do harm; • these behaviours are seen to take place within and to sustain a pattern of domination and control, a 'patterned subjugation of one partner by the other'; 44 and • the impact of domestic violence is seen less in terms of discrete acts or episodes of violence and more in terms of the ongoing, chronic and cumulative impact of a regime of domination and control.
Studies of domestic violence that use the Conflict Tactics Scales (CTS and CTS-2) or other acts-based measures of aggression tend to find that males and females perpetrate domestic violence (as defined by these measures) at similar rates or, in some instances, that women report higher rates of perpetration than men. Findings of 'gender symmetry' tell us only that roughly similar numbers of men and women report that at least once in some specified time period they have engaged in at least one of the violent behaviours listed in the CTS or other survey instrument. 22 However, even in these studies it is clear that men's use of violence produces more physical injuries, more negative psychological consequences and more fear than women's use of violence. 22  • In a US study among adolescents, girls aged 15-18 reported perpetrating moderate threats/physical violence at three times the rate, and serious psychological abuse at more than four times the rate, than boys aged 15-18. 59 • A recent meta-analytic review on teen dating violence, based on meta-analysis of 31 studies, found that girls had higher rates of perpetration than boys, and that boys and girls had similar rates of teen dating violence victimisation. However, the authors also question this, noting the inconsistency between reported rates of perpetration and victimisation and methodological issues in these studies. 60 This pattern of gender symmetry or greater female perpetration is also evident in at least some studies of adolescent family violence: • In a recent Australian study, one in five young people (20%) self-reported that they had ever used violence against a family member, including 23% of females and 14% of males. The most common forms of family violence perpetrated were verbal abuse (17% of females and 9% of males), physical violence (11% and 7%, respectively), and emotional/psychological abuse (6% and 2%, respectively). 61 However, for several reasons, apparent findings that girls and young women are as likely, or more likely, than boys and young men to perpetrate dating violence must be interpreted with caution. First, social norms may lead to boys and young men underreporting their violent behaviour. Girls may be more likely to report their perpetration of violence than boys are to report theirs. Second, girls and young women's use of physical aggression may represent efforts to defend themselves against their intimate partners' violence rather than a means of power and control. 62 66 Most of the men had been in relationships in the 12 months preceding the survey, and of these, 23% had perpetrated sexual or physical intimate partner violence in that time.
Some studies among LGBTQ populations also document significant rates of perpetration. For example: • In a study among lesbian, gay, bisexual, transgender, queer, and questioning (LGBTQ) university students, using the CTS-2, nearly one-third (29.7%) of the sample reported engaging in any type of partner violence. One in five (19.9%) had perpetrated physical partner violence, while 12.5% and 10.5% reported perpetrating psychological and sexual partner violence, respectively. 67 (This study did not include gender-specific analyses.)

Gender contrasts in the perpetration of domestic violence
There are significant gender contrasts in the perpetration of domestic violence, including in: • the severity of violence used; • the kinds of violence used, particularly in relation to sexual violence; • the presence or absence of a wider pattern of coercive control; and • the motivations for perpetration, including self-defence.
Some of the studies described thus far, and particularly those relying on acts-based measures of domestic violence, find apparent symmetries in men's and women's use of and subjection to domestic violence, with men and women using domestic violence at similar rates. They do so, in part, because of how actsbased measures work, as this report explores in greater detail in the Appendix. However, closer and more comprehensive examinations of domestic violence perpetration and victimisation find gender asymmetries.

Severity:
Men are more likely than women to physically assault, sexually assault and/or murder an intimate partner. 68 While studies relying on clinical samples (e.g., from criminal legal and service settings) have mixed findings on severity, studies with the largest samples show that, on average, men perpetrate more severe violence than women. 69 Although both men and women may use severe violence against intimate partners, men are more likely than women to do so, including such behaviours as nonfatal strangulation and intimate partner homicide. 70 Intimate partner sexual violence: Domestic violence may be accompanied by sexual assault and coercion and, indeed, some definitions of domestic violence include sexual coercion and sexual violence. Comparing females' and males' experiences of intimate partner violence, women are far more likely than men to be sexually assaulted by an intimate partner or ex-partner. Australian Bureau of Statistics (ABS) data finds that among victims of violence by an intimate partner, far more women than men -12 times as many -had suffered sexual assault. This suggests that male perpetrators of domestic violence against female partners are far more likely to sexually assault their partners than the reverse. Women's victimisation by current or former intimate partners is more likely than men's to include sexual assault. Sexual assault was part of the experience of intimate partner violence for a far higher proportion of female victims (37%) than male victims (10%). 71 Six further studies corroborate the finding that men's sexual violence and coercion against female partners is more common than women's sexual violence and coercion against male partners, even in samples comprised entirely of perpetrators. 69 Coercive control: There is growing recognition that domestic violence often involves, or can be defined by, a systematic pattern of power and control exerted by an individual against their intimate partner. The term 'coercive control' refers to 'the patterned subjugation of one partner by the other', 44 as this report explores in greater detail below. Examining men's and women's perpetration of domestic violence, male perpetrators are more likely than female perpetrators to use coercive and controlling strategies. 72,73 As a recent systematic review establishes, most coercive control or 'intimate terrorism' is by men against women. 74

Motivations for perpetration, including self-defence:
There are gender differences in the motivations for domestic violence perpetration. A review of studies on men's and women's motivations for intimate partner violence finds some support for the point that men's intimate partner violence is more likely than women's to be motivated by power and control and that women's intimate partner violence is more likely than men's to be motivated by self-defence (and to take place in the context of their partners' violence). 75 Men's perpetration of intimate partner violence is more likely than women's to be motivated by control or other instrumental reasons and to be unilateral rather than bilateral (two-way) or reactive.
In a systematic review of evidence regarding women's motivations for the use of physical intimate partner violence in heterosexual relationships, none of the 14 studies that ranked women's motivations for perpetrating violence found that control was the primary motivation. 76 Instead, four found that self-defence was women's primary motivation, and in another it was the second most common, while other common motivations included anger, desiring one's partner's attention and retaliation.
Various other studies find gender contrasts in motivations for perpetrating relationship aggression. Women are more likely to identify emotional expression, self-defence or retaliation as reasons for their aggression, while men are more likely to identify instrumental reasons directed towards particular goals (e.g., 'to get her to stop nagging and leave me alone'). 77 On the other hand, women's intimate violence can also be motivated by efforts to show anger and other feelings, desire for attention, retaliation for emotional hurt, jealousy and control, 78 and CTS-based studies find significant proportions of couples characterised by female-only violence. 79

Women's physical violence towards intimate male partners is more likely than men's to involve self-defence.
This is demonstrated in studies among female perpetrators, 78, 80, 81 men presenting to hospital emergency departments with injuries inflicted by their female partners 82 and heterosexual couples. 20 In other words, when a woman is violent to her male partner, it is often in the context of his violence to her. It is largely reactive and self-protective. 20 For example, in a representative sample of American university undergraduates, women's use of physical violence was more likely than men's to occur in the context of a partner's violence towards them. This suggests that women were unlikely to be deliberately using violence to exert control over their partners and that their violence is more likely to be bilateral (two-way) or defensive. 83 These contrasts hold even in studies conducted among male and female perpetrators of intimate partner violence (rather than general population samples), where one might expect to see a greater representation of seriously violent female perpetrators. Still, male perpetrators are more likely than female perpetrators to be initiating violence and are less likely to be using violence in response to ongoing violence by their partners. [84][85][86] We return now to examining patterns of self-reported perpetration, turning to sexual violence.

Sexual violence: self-reported perpetration
Relatively few studies have used general population samples in asking about sexual violence perpetration, with most studies asking only males. Nevertheless, there are some studies that do include samples of both males and females, particularly among school and university populations. These consistently find significantly higher rates of sexual violence perpetration by males than females. To note some examples published since 2010: • In a US study of high school students, 10.9% of males and 5.2% of females had perpetrated sexual violence in the last year. 87 • In a large-scale US study of high school students, 8% reported that they had perpetrated at least one of three unwanted sexual activities, comprising 10.6% of males and 5.8% of females. 88 • In a study among university students in Botswana, males were more likely than females to report having used force or threats to make a partner have sex. 46 • In a study among university students in Nigeria, 22% had perpetrated a form of non-consensual sex, including 28% of men and 19% of women. 89 • In a US study of university students, 6% of students had perpetrated sexual violence, comprising 14.5% of men and 3.8% of women. 90 • In a US study of university students, 1.3% of the sample perpetrated unwanted sexual contact and 0.9% perpetrated sexual violence while attending university. Before attending university, 2% of the sample perpetrated either unwanted sexual contact or sexual violence. Male undergraduates were more likely than female undergraduates to report perpetration of unwanted sexual contact (3.4% versus 1.0%) and overall sexual violence (2.1% versus 0.7%). 91 • In a US study of youth aged 10-21, more males than females perpetrated four out of five forms of sexual violence, but one form was perpetrated more by females than males. Sexual harassment perpetration was reported by 23% of males and 17% of females, sexual assault by 10% of males and 12% of females, attempted rape by 8% of males and 3% of females, and rape by 4% of males and 2% of females. 92 • In another US study of university students, 2.1% reported any perpetration of sexual assault (sexualised touching; attempted oral, anal or vaginal penetration; or completed penetration without consent) since starting university. 93 About 3.3% of male students, 1.4% of female students and 0% of gender-non-conforming students (1.6% of the sample) reported any sexual assault perpetration. • In a Chilean study among university students, 26.8% of men and 16.5% of women reported at least one incident of sexual aggression perpetration since the age of 14.94 More men than women reported sexual aggression perpetration towards a friend or acquaintance through threatening or using physical force, and through exploiting the victim's inability to resist and more men than women reported sexual aggression perpetration towards a stranger through exploiting his/her inability to resist. Men's sexual aggression perpetration was more severe than women's, with more men than women perpetrating rape. 94 • In a Turkish study among university students, 28.9% of men and 14.2% of women reported at least one instance of sexual aggression perpetration. Similar to the previous study, men's sexual aggression perpetration was more severe than women's. 95 There are also a small number of studies of both men and women that use community samples. For example: • In a US longitudinal study with 11 waves of data among individuals who over the course of the study were aged 11-43, males comprised 90% of those who had perpetrated sexual assault.96 Close to one in ten males (8.8%) were identified as sexual assaulters, and 4.0% were identified as serious sexual assaulters (involving a completed forced intercourse or the use of physical force or injury). • In a study in the US, sexually violent acts were more commonly perpetrated by men than women. 49 • In a study in Iran, significantly more men (37%) than women (13.4%) reported at least one incident of sexual aggression perpetration. 97 Research among young people finds significant gender contrasts in perpetration of sexual violence: • A recent meta-analytic review, based on meta-analysis of 31 studies, found that boys had higher rates of sexual violence perpetration than girls in teen dating relationships (and lower rates of victimisation). 60 • A review of 113 studies on sexual aggression in countries in Europe found that rates of male perpetration were substantially higher than those for female perpetration, and rates of female victimisation mostly were higher than those of male victimisation (although rates for the latter were considerable in some contexts). 98 • Looking at older youth populations, a survey in ten European countries among young adults aged 18-27 found that a significantly greater proportion of men than women had perpetrated sexual aggression (16.3% versus 5%). 99 To note some other studies not included in the 2017 meta-analytic review above of 31 studies: • In a US study among high school students, 8% of the sample had perpetrated unwanted sexual activities in the last year, comprising 10.6% of males and 5.8% of females. 88 • In a US study among young people aged 14-21, 9% of adolescents had perpetrated sexual violence in their lifetime, and 4% of adolescents had perpetrated attempted or completed rape in their lifetime. There were no gender differences in overall rates of perpetration, although there were differences, for example, in ages and trajectories of perpetration. 100 • A US longitudinal study, involving four waves of data among 13 to 25-yearolds, identified three groups: (1) nonperpetrators, comprising 69% to 81% of the sample across waves, (2) sexual harassers, comprising 17% to 29% across waves, and (3) multiple perpetrators, who engaged in all types of sexual violence perpetration, comprising 1% to 3% of the sample. 101 Males were significantly more likely than females to be in the multiple perpetration and sexual harassment groups. 101 What are the most common strategies used in coercing or forcing others into sex? Studies among young people that disaggregate different forms of sexual coercion or sexual violence typically find that verbal coercion and alcohol or drug-facilitated or incapacitated sex are more common than assaults involving physical force. For example: • In a German study of first-year university students, the two coercive strategies of verbal pressure and the exploitation of the victim's incapacitated state were more common than a third strategy, the threat or use of physical violence. 102 • In a large-scale US study of high school students, the most commonly reported perpetration tactic was having had unwanted sexual activities with another high school student when she or he was under the influence of alcohol or drugs, and this was true for both males and females.88 This was more frequent than threatening to end the friendship or romantic relationship, using constant arguments, begging or threatening or using physical force.
• In a US study of high school students, alcohol or drug-facilitated or incapacitated sex was the most common form of sexual coercion (perpetrated by 8.5% of males and 4.1% of females), then sexual coercion (5.3% of males, 2% of females), and then physically forced sex (4.3% of males, 1.4% of females). 87 • In a systematic review of 77 studies among male university students in Canada and the US, sexual violence involving verbal tactics such as pressuring behaviour, expressions of anger or threats to the relationship was more common than sexual intercourse obtained via incapacitation, physical force or threats of physical force. 10 • In a Spanish study among male university students, among the 15.3% who reported having perpetrated sexual coercion, the most common strategies used were pressure and verbal manipulation, or the use of alcohol or drugs, whereas threats or the actual use of force were much less common. 103 In these community samples therefore, verbal coercion is the most common or one of the most common tactics among perpetrators, while physical force is the least commonly used tactic. 13 Not all studies find such patterns, however. For example, in a Turkish study among university students, the most common coercive strategy reported by 21.7% of men and 10% of women was the use or threat of physical force. 95 Regarding other strategies, 15% of men and 6.2% of women reported that they had exploited the other person's inability to resist and 8.6% of men and 3.7% of women reported the use of verbal pressure. It is possible that there are cultural differences in the typical strategies of sexual coercion, 95 or that verbal pressure is underreported because of lack of recognition of sexual coercion via this strategy.
Even among men who have been imprisoned for sexually violent offences, both nonphysical and physical tactics of perpetration are evident. For example, in a US study examining incarcerated men's perpetration of non-physical sexual coercion (e.g., verbal pressure or manipulation) and physical sexual aggression (e.g., incapacitation, physical force or threats), some offenders had used only the former while others also used the latter. 104 Studies that disaggregate different forms of sexual coercion may also find that coerced sexual touching is more common than coerced penetration. For example: • In a US study of university students where 2.1% reported any perpetration of sexual assault, sexualised touching was the most common form perpetrated and completed penetration was the least common form. 93 For those who reported perpetration, asked about the most significant incident, more than half reported that it involved sexualised touching only, whereas a smaller number involved attempted or completed perpetration. 93  According to this systematic review: • Close to one in three male university students (29.3%) in the USA and Canada have perpetrated sexual violence in their lifetimes. 10 • About one in five male university students (19%) have perpetrated sexual coercion (defined as any type of sexual intercourse obtained via verbal tactics such as verbal pressuring behaviour, expressions of anger, threats to the relationship and so on). • About one in 15 (6.5%) have perpetrated rape (defined as any type of sexual intercourse obtained via incapacitation, physical force or threats of physical force). 10 These findings are consistent with those of an earlier review of 120 studies conducted over the 1960s to 1990s in the USA among adolescent, university and adult populations. This found an overall prevalence rate for males' perpetration of sexual violence of about 25%, including a rate for rape of 4.7%. 106 Studies among university students in other countries also find substantial rates of sexual violence perpetration: • A study in Spain found that 15.3% of male students aged 18-23 had perpetrated some form of sexual coercion. 103 • A study of first-year university students in Germany found that 13.2% of men (and 7.6% of women) had perpetrated sexual aggression since the age of 14. 102 Focusing on heterosexual incidents, prevalence rates were substantially higher for men (11.6%) than for women (5.9%). • A study among Polish university students found that 6.8% of men (and 1.9% of women) reported at least one act of sexual aggression perpetration from the age of 15 to one year ago, and 8.7% of men (and 7.5% of women) reported at least one such act in the last year. 107 • A recent UK study found that one in nine male university students (11.4%) reported having perpetrated some form of sexual aggression in the last 24 months. 108 Some studies among male adolescents also document significant levels of sexual violence perpetration. For example:

Self-reported willingness to perpetrate sexual violence
A related stream of scholarship does not gather data on people's actual perpetration of violence, but on their self-reported likelihood of doing so if they knew they would suffer no consequences. There is a small body of scholarship for example on 'rape proclivity' among males. For example: • In a study among male university students in the US, 48% of college men acknowledged at least some likelihood of assaulting a woman, and 19% reported that it would be likely or very likely if they knew there would be no penalty or consequences for committing sexual assault. 113 • In a study of male university students in the US, 35.1% of the men indicated some likelihood of engaging in arguments or pressure to obtain sex play, 12.1% reported some likelihood to perpetrate sexual intercourse through arguments or pressure, and 9.2% reported some likelihood to do so through administration of drugs or alcohol. 114 On the other hand, very few men (less than 1%) reported any likelihood to use force to obtain sexual activities. Men who indicated some risk of perpetrating sexual aggression at baseline were more likely than other men to have actually perpetrated sexual aggression in the three months to follow-up. • In a study among male university students in England, 30% of the men endorsed behavioural propensity to initiate a victim into a multiple perpetrator rape scenario, and 23% endorsed behavioural propensity to intimidate a victim into a multiple perpetrator rape scenario. 115 Similar streams of research focus on other forms of violence such as sexual harassment, again assessing individuals' likelihood to harass. 116

Sexual orientation and sexual violence perpetration
Very few studies have examined associations between sexual orientation and sexual violence perpetration. In the systematic review described earlier of studies among North American male university students, none of the 78 samples reported perpetration rates specifically among sexual minority men. However, four other studies among males or mixed-gender samples find conflicting patterns regarding likelihoods of perpetration: • In a German study, university men who were bisexual (classified behaviourally rather than by self-identification) reported the highest rates of sexual perpetration, followed by behaviourally heterosexual men and behaviourally gay men. 102 • In a US study, there were no differences between sexual minority and heterosexual American college men in reported rates of sexual violence perpetration. 117 • In a large-scale US study of high school students, there were higher rates of perpetration of unwanted sexual activities among students who did not identify exclusively as heterosexual. 88 • In a recent US study, heterosexual men had the highest reported sexual violence perpetration rates among college men. 93 • Differences between these findings may reflect differences in the measures of violence and sexual orientation used. 118

Sexual harassment: self-reported perpetration
There has been very little research gathering self-report data on the extent to which people perpetrate sexual harassment. There are a small number of studies examining males' extent of sexual harassment perpetration or self-reported propensity to sexually harass, and they document substantial rates of sexual harassment perpetration. For example: • In a US study of male enlisted Navy personnel in their second year of service, 60% reported perpetrating at least one type of sexual harassment in the last two years. 119 Rates of perpetration for specific forms of sexual harassment were 57% for sexual gender harassment, 28% for sexist harassment, 21% for unwanted attention and 4% for sexual coercion.
Our review could identify only one study comparing men's and women's self-reports of sexual harassment perpetration: • In a US study among university students, men reported engaging in sexual harassment behaviour more than women, with significant differences for each of three types of harassment: sexist and derogatory jokes, unwanted sexual attention (sexual remarks or questions), and sexual coercion. 120

Image-based sexual abuse: selfreported perpetration
Image-based sexual abuse shows similar patterns, with males showing higher rates of perpetration than females in nearly all studies: • In a national Australian study, asked if since the age of 16 they had ever taken, distributed and/or threatened to distribute a nude or sexual image of another person without their consent, 11.1% of people had done so. A total of 12.0% of males and 6.2% of females reported having nonconsensually taken images and 9.1% of men and 4.4% of females reported having non-consensually distributed images. 121 • In a US study, asked if they had 'knowingly shared a sexually explicit image or video of someone else without his/her consent', 5.1% of people had done so, including 6.4% of heterosexual men and 2.7% of heterosexual women. 122 • In a survey of digital dating abuse among US university students, overall rates of perpetration were similar among males and females. 123 At the same time, men were more likely than women to report threatening to distribute embarrassing information about their dating partner(s) using the internet or a mobile phone, and pressuring their dating partner(s) to take a sexually suggestive/nude photo or video using a computer or mobile phone. • In a Canadian survey among young adults aged [18][19][20][21][22][23][24][25]19.7% of the sample had engaged in coerced sexting perpetration (pressuring someone to engage in sexting) in their lifetime and 14.9% had shared intimate images or videos of others as a source of entertainment, bonding, showing off or 'fitting in' amongst their peers. 124 Rates of perpetration of more severe forms of 'inappropriate intimate image-based behaviour' were low: with 5.1% having participated in the illegal distribution of intimate images; 3.2% having uploaded someone's nude images or videos online, to social media or to the victim's place of work; 1.6% having disseminated intimate images as a means of seeking revenge and/or humiliation of the victim; and 1.6% reporting sextortion (threatening to expose an intimate image as a means of coercing the victim into doing something). Coerced sexting was perpetrated more often by males than females, but there were few other gender differences. 124 • In a US study among middle and high school students, of those who reported at least one dating or sexual partner in the past 12 months, 8% acknowledged having pressured an intimate partner to sext -to send sexual messages or texts or nude or sexy photos. 125 In both sixth grade and ninth grade, boys were significantly more likely to have perpetrated coercive sexting than girls. 125 For example, among ninth graders, 13.4% of boys and 8.3% of girls had done so. Also, those who pressured a sexual or dating partner to send them sexts were significantly more likely to have perpetrated at least one form of sexual coercion (pressuring a dating partner to have sex without a condom, insisting on sex when partner did not want to and/or using threats to pressure a partner into having sex).
The perpetration of professional sexual misconduct, that is, in workplace settings, shows a particularly strong gender divide. For example: • An Australian study of notifications of sexual misconduct by health professionals documents that 88% of notifications were of male health professionals. These involved 0.6% of men and 0.03% of women in the health professions, suggesting that men were 20 times as likely as women to perpetrate sexual misconduct. Two-thirds (62%) of the female practitioner notifications involved inappropriate relationships, not harassment or assault. 126

The strategic use of particular tactics of control and abuse
People's use of particular tactics of violence, abuse or control is shaped by the resources available to them and the social locations they and their victims occupy. For example, perpetrators of intimate partner violence may make strategic use of their partner's social situations to intensify their control: • using the visa status of a partner who is a migrant and refugee or on a temporary visa, threatening them with deportation or criminal action to force compliance; • threatening to disclose a same-sex partner's sexual orientation to family or workplaces; 127 and • taking advantage of a partner's physical or intellectual disability to maintain control over them.

Drivers, predictors, causes, risk factors
What are the causes of violence perpetration? Why do some people and not others use violence? Why are rates of violence perpetration far higher in some settings, communities and countries than others?
The most common explanatory framework for answering such questions is the socialecological model: a framework for identifying and addressing the risk factors for violence that operate at different levels of the social order. It embodies the widespread recognition that perpetration is influenced by factors at the individual, relationship, community and societal levels. The ecological model highlights that risk factors for violence -that increase the likelihood of perpetration and victimisation -can be found at multiple levels of society. 128 The model also assumes that: • These levels are interconnected, such that violence is 'a multifaceted phenomenon grounded in the interplay among personal, situational and socio-cultural factors'. 129

The Ecological Model
Examples of structures, norms and practices found to increase the probability of violence against women, at different levels of the social ecology.

System and institutional
Organisational and community There is not scope here to review the range of risk factors identified for the forms of violence on which this report focuses. In addition, other recent publications provide comprehensive accounts of risk factors for the perpetration of forms of violence including:
This report noted earlier that many studies of violence perpetration and victimisation, particularly for domestic violence or intimate partner violence, only measure individuals' use of or subjection to a range of physically aggressive acts. Studies that then examine potential associations between violence perpetration or victimisation and various risk factors must rely, in part, on these same narrow acts-based approaches. In such cases therefore, they may not be examining the risk factors specifically for the perpetration of domestic violence involving asymmetrical and more severe violence and coercive control or patterns of domination and control. Instead, they may be examining the risk factors also or instead for the perpetration of relatively minor and reciprocal aggression unaccompanied by power and control.
Here, nevertheless, we highlight some sets of risk factors of interest at different levels of the social-ecological model.

Witnessing or experiencing childhood violence
One of the most consistent risk factors for violence perpetration documented in the scholarship is childhood exposure to domestic and family violence. That is, adults who as children witnessed or experienced these forms of violence show a greater likelihood of using violence than other adults who did not. This association is evident in a range of reviews and meta-analyses. For example: • In a systematic review of longitudinal studies that have prospectively investigated childhood and/or adolescent predictors of domestic violence perpetration and/ or victimisation among adult men and women in intimate relationships, child and adolescent abuse was identified as a significant predictor of domestic violence perpetration (and victimisation). 134 • A systematic review of dating violence risk factors among undergraduate college students notes associations between childhood exposure to violence and both male and female students' perpetration of domestic and dating violence. 16 • A systematic review of risk factors for violence against women in high-prevalence settings, based on 241 studies, pointed to the influence of exposure to other forms of violence (armed conflict, witnessing parental violence and child abuse). 137 • A review of literature on male perpetration of intimate partner violence, including 87 articles spanning 1980-2020, found that a common risk factor for perpetration in these studies was growing up in a violent home or witnessing violence at an early age. 133 • A meta-analysis on same-sex intimate partner violence finds that experiencing child abuse in one's family of origin is a risk factor for perpetration. 14 • There are also associations between intergenerational exposure to violence and the perpetration of sexual violence. 16 Social norms and practices that emphasise men's control and dominance over women Some reviews, however, find more mixed results: • In a systematic review of adolescent dating violence, of 20 studies using general samples, six found an association between adolescents' exposure to intimate partner violence and their own perpetration of adolescent dating violence, six found no association, and the remaining eight studies found mixed results (such as an association for boys but not girls, for girls but not boys, in heterosexual youth but not sexual minority youth, with some forms of violence but not others, and so on). 62 Six further studies with samples of high-risk youth again found mixed results.
The variability in these findings may reflect variations among these studies in age, developmental stage, exposure to other forms of victimisation and adversity, gender and environmental factors. 62 Nevertheless, it is clear that many of the individuals who use violence are also victims of violence, whether as children or in later life.

Violence-supportive and hostile masculine attitudes
Violence-supportive attitudes are a consistent predictor of the perpetration of domestic, family and sexual violence. A large volume of scholarship demonstrates that men are more likely to use violence against women and girls if they subscribe to attitudes that condone, minimise, excuse or justify that violence. This is evident in four recent meta-analyses and systematic analyses of well over 300 studies: • Hostile masculinity can be understood in terms of five components. There are significant relationships between each of the components and men's perpetration of sexual violence against women: • There is a consistent relationship between sexual aggression and sexual dominance -having sexual motives and feelings of gratification linked to having power over one's sexual partner. 149 • There is a consistent relationship between sexual aggression and hostility towards women: antagonistic or distrustful attitudes towards women. There is empirical support for hostility towards women as an associate of sexual aggression among a wide variety of male populations, including university students, student athletes, community members, military personnel and incarcerated men, and from studies using correlational, quasi-experimental, interventional and longitudinal designs. 149 Some studies find that hostility towards women interacts with other variables to predict sexual aggression. For example, one study found that college students who had high hostility towards women and lacked empathy were especially likely to perpetrate sexual aggression. 152 • Adversarial sexual beliefs -beliefs that male-female relationships are inherently exploitative and manipulative -have a significant association with sexual violence in most studies. 149 • Acceptance of rape myths -rapesupportive attitudes and false stereotypical beliefs about rape i.e. that most women lie about rape, or provoke rape through their dress or behaviour, and/or that a man cannot rape his wife or girlfriend -is associated with sexual aggression, including in studies among university students, junior high and senior high school students, student athletes, community, correctional inmates and military personnel. Acceptance of rape myths is associated with several forms of sexual aggression including various tactics (e.g., verbal coercion, technologybased coercion, physical force) and sexual acts. Longitudinal studies find that higher acceptance of rape myths is a significant predictor of future sexual aggression. 149 • Finally, acceptance of interpersonal violence -the belief that force is a legitimate way to gain compliance in sexual relationships -is consistently associated with sexual aggression perpetration. 149 Further, more recent studies continue to document the influence of these variables, such as rape myths, on male sexual violence perpetration or on self-reported proclivity to perpetrate. 153 Hostile masculinity has a more direct effect on men's use of sexual aggression than many other variables, and mediates the relationship between sexual aggression and various family based, relational and individual-level variables, such as childhood victimisation, witnessed interparental violence, attachment difficulties, psychopathic personality traits, general hostility, delinquency, peer attitudes and influence, general drinking behaviours and masculine gender role stress. 149 While some studies find that a combination of hostile masculinity and a preference for impersonal sex is particularly likely to produce sexual aggression, others find that the main effect of hostile masculinity is more significant than its interaction with an orientation towards impersonal sex, and that men with high levels of hostile masculinity perpetrate sexual aggression at similar levels to men with high levels of both hostile masculinity and impersonal sex. 149 Hostile masculinity thus seems a particularly significant driver of men's sexual violence against women.
Violence-supportive and gender-inequitable attitudes also shape men's perpetration of other forms of abuse such as workplace sexual harassment.
Men are more likely to perpetrate sexual harassment if they have 'high levels of hostile sexism, rape myth acceptance, authoritarianism, endorsement of traditional gender roles and masculine ideology, and low levels of agreeableness, openness to experience and empathy'; 154 endorse sexist norms, for example, of male dominance; 155 link sex and power; 116,154 have a 'social dominance orientation', a preference for hierarchical group relations; 154 or are predisposed to feel gender identity threats 156 or status threats, 157 and engage in sexual harassment as a way to bolster their gender identity or defend it when it is perceived as threatened. 154,155 Definitions of masculinity can influence perpetration among some men in a different way, in circumstances where they are unable to achieve conventional masculine status or feel that their masculine authority is under threat. 133 Some studies find that men's use of intimate partner violence is a reaction to 'masculine gender role stress' and an effort to reassert masculinity. In contexts where definitions of masculinity tie it to providing financial stability or having greater economic and occupational power than one's female partner, 'if men are unable to establish dominance over women through higher income, education, and occupational prestige, IPV [intimate partner violence] may serve as an alternate means to assert and express one's masculinity'. 133

Sexist and violence-supportive peers
Sexist, sexually hostile, and violence-supportive attitudes are not evenly spread among men and boys, nor are they random. Instead, adherence to such attitudes is influenced by social, cultural, interpersonal and individual factors. 149 For example, men are more likely to develop hostile masculinity if they are in contexts and cultures that value stereotypical male characteristics (such as power, toughness, dominance and status); they associate with anti-social and sexist peers; and/or they experienced or witnessed adversarial interpersonal relationships in childhood. 149 Peer support is an important influence on men's sexual violence perpetration. Men are more likely to be sexually aggressive if they have sexually aggressive peers, that is, male friends who themselves tolerate or perpetrate sexual aggression.
At least two processes shape this: peer reinforcement of sexually aggressive attitudes and behaviours and self-selection into violencesupportive peer groups and settings. 150 The norms, ideologies and patterns of interaction in peer groups can promote, justify and legitimise sexual violence. 158 Specific peer groups, such as fraternities and athletic teams in universities, may foster a hypermasculine culture in which men feel pressure or entitlement to use coercion and force for sex. In this sense, men may learn to use sexual violence in part by interacting with male peers who promote or justify sexual violence perpetration. Boys and men who perceive they will gain status or acceptance among male peers by having sex may use coercive or aggressive tactics to obtain sex to realise these gains. 109 Various studies point to the influence of male peer support: • In a meta-analysis of sexual violence perpetration in universities, peer approval of sexual violence was shown to increase the odds of sexual violence perpetration. 15 Individuals showed a higher risk of perpetrating sexual violence if their peer groups (in this case, fraternities) condoned this behaviour or if their peers approved of sexual violence. › Indeed, factors such as group membership and perceived peer approval of sexual violence may have greater influence on sexual violence perpetration than personal beliefs. In this meta-analysis hostile masculinity and rape myth acceptance did not strongly predict perpetration; fraternity membership and peer approval of sexual violence were more decisive predictors. 15 • In a second meta-analysis of studies, this time of studies over 2000-2021 of sexual assault perpetration by male university students in the US and based on 25 studies, peer approval of sexual assault was one of the strongest risk markers for perpetration. 142 • In an older review, a systematic qualitative review of risk factors for sexual violence perpetration based on 191 articles, sexual violence-supportive peer attitudes and behaviours, and hypermasculine or allmale peer groups emerged as significant influences. 18 Further recent studies continue to document such associations: • In a national longitudinal study among boys, those who perceived that their peers would respect them more for having sex were more likely to perpetrate sexual intimate partner violence in young adulthood compared to boys who did not endorse perceived peer approval for sex. 109 • In a study among male university students in the US, men reported higher likelihood of perpetrating rape if they perceived that their peers would not intervene as bystanders. 153 On the other hand, when men felt that among their peers there was a prevailing social norm of willingness to intervene, their own proclivity to perpetrate sexual violence was lower.

Situational variables
In recent years there has been increased attention to the situational factors that may shape violence perpetration and victimisation -to the elements of or processes in the immediate social and physical environment that might influence individual acts of violence. 150 This extends the ecological model, adding more dynamic factors to the largely static factors emphasised in some versions of the model in order to better understand and predict violence. An earlier expression of this attention is the 'background-situational' model of intimate partner violence, suggesting that while background factors (such as the experience of violence in childhood) establish the proclivity for violence, situational variables (such as substance use and relational conflict) increase the likelihood of aggression in a particular circumstance. 159 In other words, background factors predict who is likely to engage in domestic violence, while situational factors help to predict when violence is more likely to occur.
Attention to situational factors is informed by environmental theories in criminology, such as routine activity theory. This identifies three necessary conditions for criminal activity: a likely offender (one who is sufficiently motivated to offend), a suitable target, and the absence of a capable guardian. 160 With regard to domestic, family and sexual violence, perpetrators may deliberately engineer such conditions. A typical feature of domestic violence is perpetrators' isolation and control of their victims. Similarly, perpetrators of sexual violence may seek to isolate and control the individuals they intend to assault, using the environment to reduce interruptions or detection or relying on settings such as a fraternity house (an all-male university residence). 150 In terms of 'suitable targets', most sexual assaults are perpetrated by men who know their victim as an acquaintance at the least, therefore increasing trust and relationship proximity. Some men believe that particular women are more acceptable targets of sexual assault, such as women they consider to be promiscuous, manipulative or attention-seeking. 150 An example of a situational or 'dynamic' variable documented as a risk factor for the perpetration of intimate partner violence is separation. Men using violence against their female partners often will escalate the severity of their violence during and soon after separation. 161,162 Men with patriarchal or sexually proprietary attitudes are more likely than other men to begin or intensify physical or sexual violence against partners who are leaving or trying to leave their marital or cohabiting relationships. 163 Men using violence against intimate partners may also make use of post-separation contact as an opportunity for abuse, e.g., during visitation and the exchange of children, and may use children as tools for violence and control. 161 Another situational variable, in this case associated with the lethality of intimate partner violence, is firearm availability. A series of studies find that having a gun in the home greatly increases the risk of domestic homicide, that gun owners are far more likely than non-gun-owners to make gunrelated threats to intimate partners, and that there are higher rates of intimate partner homicide among women in states without laws restricting firearm access for individuals with a restraining order against them. 133

Settings and contexts
Risk factors for violence perpetration can also be found at the level of settings and institutional contexts. In relation to intimate partner violence and sexual violence, three collective contexts in which most empirical research has been done in this area are university fraternities (male residential colleges on campuses), sport, and the military. Attention has also been paid to the features of particular institutional contexts or settings that may facilitate violence perpetration.
Let us take university fraternities as a case study. University campuses and particularly university residences may involve increased risks of sexual and domestic violence because of alcohol consumption, 'hookup' culture, social or athletic groups fostering hypermasculine norms, fraternity and social events, and other factors. 164 There is a consistent association between men's membership of all-male university residences (fraternities) and their likelihood of sexual violence perpetration: • In a meta-analysis of university men's perpetration of sexual violence, fraternity membership was associated with an increased likelihood of sexual violence perpetration. 15 (Note that not all studies find such associations. For example, Gidycz et al. 165 did not find any relationship between fraternity membership or athletic participation and sexual assault perpetration, although they could not test for whether the social norms of these settings were violence-supportive or not.) 165 • This association is structured, in part, by attitudes. In a US study largely among white, heterosexual 19-year-old males, fraternity members were more accepting of sexual violence against women than non-fraternity members, in part because they more strongly endorsed traditional masculine norms, felt pressure from their friends to uphold masculine norms, and more readily viewed women as sexual objects. 166 • The association between fraternity membership and sexual violence perpetration may be shaped in part by self-selection, as two studies show: › In a US study among male incoming first-year university students, male students who intended to join a fraternity showed higher levels of self-reported proclivity to perpetrate rape than other male students. 153

›
In another US study, a longitudinal one, men who were interested in joining a fraternity and did join one showed higher proclivity to perpetrate sexual aggression and greater support for rape myths than non-interested non-members. 167 • At the same time, we should be careful of exaggerating the extent of peer support around male perpetrators. For example, a qualitative interview study of young male perpetrators in Sweden documented mixed responses to violence in their social networks. 168 The formal and informal characteristics of institutions, such as workplaces, also shape the perpetration of violence. Focusing on sexual harassment, for example, environmental or situational risk factors for higher levels of perpetration and victimisation in workplaces include organisational tolerance of harassment (e.g., through formal or informal policies and practices that allow sexual harassment to continue); 116,169-172 male-dominated environments and gender ratios; 170,172 significant power differentials within hierarchical organisations; 172 and the extent of uncivil behaviour in the organisation and the organisation's 'justice climate'. 173

Neighbourhoods, communities and cultures
Towards the more 'macro' end of the socialecological model, there are important influences on violence perpetration to do with the features of neighbourhoods and communities. For example: • A systematic review of community-level correlates of intimate partner violence (IPV) against women finds that higher levels of collective efficacy (e.g., neighbours' willingness to help each other) and social cohesion in communities are related to lower risks of women experiencing IPV, controlling for other community, family, relationship and individual factors. 174 • Rates of intimate partner violence perpetration and victimisation are higher in neighbourhoods with higher levels of economic disadvantage, more violencesupportive social norms, higher levels of social disorder and community violence, and greater numbers of alcohol outlets. 4 • A systematic review of neighbourhood factors and dating violence among youth finds that neighbourhood disadvantage is associated with dating violence, as are lower levels of social control and community connectedness (of communities' ability to advocate for themselves, uphold civic institutions such as schools, maintain strong social networks and high levels of trust and social support, and collectively monitor youth and appropriately sanction problem behaviour). 175 Violence perpetration is also shaped by entire cultures -by the social norms and social organisation of specific societies. There are associations between men's violence against women and various social and structural elements of gender inequality, including maledominated power relations in public life, as well as in families and relationships, and rigid and patriarchal gender roles. 176 In relation to sexual violence, for example, the feminist notion of 'rape culture' emphasises that sexual violence is underpinned by social and structural factors. Rape culture refers to 'the social, cultural and structural discourses and practices in which sexual violence is tolerated, accepted, eroticised, minimised and trivialised', 177 with such discourses and practices informing rape perpetration. 150 Violence perpetration also is influenced by macro-level social, economic and environmental factors and forces, such as colonialism, pandemics and climate change: • A cross-national study of factors shaping intimate partner violence found that structural factors, including colonialism, were associated with higher rates of this violence. The study identified three distinct risk contexts: 1) non-patriarchal egalitarian, low rates of homicide; 2) patriarchal post-colonial, high rates of homicide; and 3) patriarchal post-colonial conflict and disaster-affected. Intimate partner violence risk was highest in the two patriarchal post-colonial contexts. 178 • Pandemics such as the Covid-19 pandemic can increase rates of perpetration and victimisation through their influence on social conditions. In Australia, the pandemic was associated with an onset of physical or sexual violence or coercive control for some women, and an increase in the frequency or severity of ongoing violence or abuse for other women. 179 This may have been because Covid-19 brought greater contact between victims and offenders, increased social isolation and increased situational stressors. • Domestic, family, and sexual violence are also influenced by other large-scale processes, including climate change. 180,181 Similarly to the increases in reports of domestic, family, and sexual violence after natural disasters and in humanitarian settings, climate change also intersects with the drivers and exacerbates violence. Climate change increases the stressors on a household, such as financial stress and insecurity. The introduction and/or exacerbation of these risk factors increases the likelihood of violence occurring.

Diversities and overlaps in the use of violence
There is considerable diversity among perpetrators and in perpetration. Scholarship on domestic or intimate partner violence, for example, recognises that the people who assault their intimate partners are heterogeneous and emphasises that there are distinct types of perpetrators and perpetration.

Types of perpetrators and perpetration
In the intimate partner violence field, the most influential typology assesses types of domestically violent men (batterers) in terms of three dimensions: the severity and frequency of the batterer's marital violence, its generality (partner-only or also extrafamilial), and the batterer's psychopathology or personality disorder characteristics. 182 It identifies three types of domestic violence perpetrator: familyonly perpetrators, dysphoric or borderline perpetrators, and generally violent and antisocial perpetrators. While influential, such typologies also have been subject to debate and critique. [183][184][185][186] Nevertheless, a recent meta-analysis supports this work, emphasising a distinction among domestic violence perpetrators between partner-only and generally violent perpetrators. 187 This distinction also receives support in a recent examination of Australian data. In an examination of a cohort of family violence perpetrators from 2012 to 2016, 40% were classified as generalist perpetrators with records for committing both non-family and family violent offences, while 60% were specialists who were only recorded for family violence incidents and related offences. 188 Similar to the findings of previous studies, 189 male perpetrators were more likely than females to be generalist offenders.

Differing risk factors
The risk factors for different forms of perpetration of the one broad type of violence also differ. For instance, there is evidence that male perpetrators of sexual violence are not uniform, and that there are differences among them related to the severity, frequency and form of their use of sexual violence. For example: • A US study of men in prison found that there were shared risk characteristics associated with their perpetration of both non-physical sexual coercion and physical sexual aggression, but there were also distinct risk factors associated with whether they only used non-physical strategies of coercion or also used more violent sexually-aggressive strategies. 104 • Compared to university men who reported perpetrating only a single incident of sexual coercion or aggressive behaviour, repeat perpetrators had higher levels of traditional gender-role beliefs, callous attitudes towards women, and engagement in verbal pressure. 191 • In a US longitudinal study of male university students, men who reported perpetrating more than one incident of sexual coercion or assault during their four years at university also had higher levels of risky behaviours (such as high-risk drinking, drug use and number of sexual partners), sexually aggressive beliefs, and antisocial traits. 192 • Men at university who used alcohol as a tactic to obtain sex, compared to other perpetrators and to non-perpetrators, were more likely to have higher levels of alcohol consumption, to misinterpret women's actions, and to believe a woman's drinking was an indicator of their interest in sex. 191 Regarding sexual harassment, again, risk factors vary depending on the specific form of sexual harassment perpetrated.
This risk factor is linked especially to the form of sexual harassment termed gender harassment, aimed less at gaining sexual satisfaction and more at offending women, with the behaviour intended to enhance or protect the status of one's own gender group -of men. 154

Risk factors and categories of perpetrator
Risk factors for the one form of violence can differ for different categories of perpetrator, to do with gender, race and ethnicity, sexual orientation and other forms of social difference and inequality.
Focusing on sexual violence perpetration by men and women for example, there is 'evidence of differences in aetiology and motives for sexual violence and coercion by gender […] and possible differences in early adverse experiences'. 19 For example, a US study among heterosexual university students found differing predictors for men's and women's perpetration of sexual coercion. 193 The study examined four potential predictors of sexual coercion among heterosexual college men and women: prior sexual abuse, socio-sexuality (a desire for short-term, casual and impersonal sex), sexual compulsivity, and sexual dominance. Prior sexual abuse was a shared predictor, but men and women differed on the others. Key predictors of sexual coercion among men were sexual dominance and socio-sexuality, whereas the key predictor of sexual coercion among women was sexual compulsivity. Men's sexual coercion was driven primarily by dominance, power, and For example, one risk factor for men's perpetration of sexual harassment is 'gender identity protection', where sexual harassment is a means of bolstering men's gender identity or defending it when it is perceived as threatened. 54 control. Women's sexual coercion, on the other hand, 'appeared to be primarily driven by feelings of compulsivity, a distinct lack or loss of control over one's behaviours, and only secondarily supported by an attraction to sexual dominance'. 193 Drawing on this and other research, the authors suggest that some women are 'connection-driven coercers', driven by a desire for romance and intimacy but resorting to coercive tactics when their needs for sexual connection are thwarted or unmet.
For sexual abuse across racial and ethnic groups, there may be differences in risk factors, protective factors or the nature of effective treatments. 19 Among adolescents and adults there is some evidence of differences across race and ethnicity in typical clinical presentations, offence characteristics and the predictive accuracy of sexual risk tools. 19 Some risk factors for perpetration of intimate partner violence in same-sex relationships, such as internalised homophobia, are distinct from those in heterosexual relationships. A recent meta-analysis found significant associations between internalised homophobia and IPV perpetration. 194 In lesbian relationships, a distinctive risk factor for IPV is 'fusion', a high level of closeness or lack of boundaries between partners. This was found to be a risk factor in a systematic review of IPV among self-identified lesbians 195 and among women in a further meta-analysis reporting on IPV by, or against, same-sex partners. 14 Some studies of intimate partner violence (IPV) among lesbian, gay and bisexual respondents use samples based on identity or sexual behaviour. This is problematic as study participants may not be reporting on perpetration or victimisation in a same-sex relationship, for example with lesbian-identified women reporting on violence by a man. 14 Another meta-analysis focused on studies that explicitly indicate that participants were reporting on IPV in a current or previous same-sex relationship. It too found that internalised homophobia was a risk marker for the perpetration of physical IPV in same-sex relationships, among both men and women. 14 Other risk factors shared among men and women included being a victim of psychological abuse, alcohol abuse and experiencing child abuse in one's family of origin.

Co-perpetration
So far, we have emphasised diversity in perpetrators and perpetration. However, particularly from data among male populations, there is evidence that many of the individuals who perpetrate one form of violence also perpetrate others. Perpetrators of one type of violence are more likely to perpetrate other types of violence. 196 One of the most frequently documented risk markers for forms of violence, such as intimate partner violence or sexual violence, and for more specific behaviours such as intimate partner physical violence, is the perpetration of other forms of violence and abuse. For example: • In a meta-analysis of studies conducted from 2000 to 2021 of sexual assault perpetration by male university students in the US, the strongest risk markers for sexual assault perpetration were related to other forms of violence in intimate relationships (prior sexual assault perpetration, physical dating violence perpetration, psychological dating violence perpetration, and being a victim of physical dating violence). This suggests that 'these individuals might be perpetrating multiple forms of violence (sexual, physical and psychological) against their romantic partners'. 142 Similarly, men who perpetrate sexual harassment are also more likely than other men to perpetrate sexual assault.
For example, in a US study of male enlisted Navy personnel in their second year of service, 60% reported perpetrating sexual harassment in the last two years and 13% reported perpetrating sexual assault. The men who perpetrated the former were four times as likely as other men to perpetrate the latter. 119 Most men (86%) who reported perpetrating sexual assault had also perpetrated sexual harassment.
There is a significant pattern of co-occurrence of violence. For example, intimate partner violence and child maltreatment often co-occur within the same household. 197,198 Children exposed to intimate partner violence also are at greater risk of sexual abuse or other maltreatment, although figures on the likelihood of co-occurring intimate partner violence and child sexual abuse vary from 12% to 70%, depending on the character of the samples and data collection. 199 Regarding sexual harassment, at the level of individual characteristics (attitudes and beliefs and other psychological and emotional variables), perpetration has some of the same risk factors as perpetration of sexual aggression. 116 There has been little examination of how sexual harassment perpetrators compare with other sexual aggressors, and why some individuals go on to commit more serious forms of sexual violence while others do not. 116 However, a recent study among male US naval personnel finds a significant overlap between the risk factors for the perpetration of sexual harassment and those for sexual assault. 119 Both forms of perpetration were predicted by a modified Confluence Model, including factors such as a high numbers of sex partners, hostility towards women, delinquency and misconduct and heavy drinking. As the authors of this study conclude, 'sexually harassing behaviour is part of the spectrum of sexual aggression and may increase the risk of sexual assault perpetration'. 119

The first use of violence
There has been little research on initial violence perpetration (on when individuals first perpetrate violence) and the trajectories of this perpetration. Compared to the large volume of cross-sectional research, the body of longitudinal research (collecting data over time on the use of violence) here is small. Nevertheless, the existing research gives us some clues regarding initial perpetration.
Substantial proportions of adolescents perpetrate dating violence against their intimate partners and ex-partners, as documented 60 in reviews of this form of violence and as described in more detail earlier.

The perpetration of intimate partner
violence and abuse is visible, therefore, from people's first participation in sexual and intimate relationships in their early teens and adolescent years. [200][201][202] Among adults and others, a range of factors may contribute to the onset or escalation of abusive behaviours across personal, relationship or environmental and situational domains including, but not limited to, loss of employment, use of drugs and alcohol, pregnancy, separation, natural disasters or events such as football finals. 205 While young people may perpetrate dating or domestic violence, less is known about whether adult perpetrators of intimate partner violence typically begin using violence against intimate partners as adolescents. However, this seems likely given common patterns of adolescent dating violence. Two US studies find that among adolescents who are perpetrating dating violence, the most common pattern over time is for their use of violence to remain stable or to increase. 201,203 Sexual violence perpetration also starts young. Two major US studies find that the average age of first perpetration among males is 16: • In a US study among young people aged 10-21, the average age at first perpetration of sexual violence was between 15 and 16 years of age, depending on the type of sexual violence. 92,100 • In a US longitudinal study, with 11 waves of data, the most common age of first perpetration of sexual violence was 16. 96 By the age of 20, 88% of respondents who were going to become sexual assaulters had already done so. In other words, nearly nine in ten male sexual violence perpetrators reported their first assault by age 20.
The first study above also found a gender contrast in initial perpetration, with males starting younger. Males were almost all (98%) of the perpetrators who reported age at first perpetration to be 15 years or younger, and most (90%) of those who began at ages 16 or 17. 100 Putting this another way, large majorities of the youngest perpetrators of sexual violence are male.
Among adult male perpetrators of sexual violence, many committed their first acts of sexual violence as teenagers. For example: • In a US longitudinal study of male university students, 30% of the men had committed an act of sexual coercion or assault during their four years at university and, among them, large proportions of the single (33%) and repeat offenders (55%) had first committed an offence before university, i.e., before the age of 18. 192 • In a study among male school students aged 17-20 in Sweden and Norway, of those who reported having perpetrated sexual coercion, 23% of the Swedish sample and 20% of the Norwegian sample were under 15 the first time they perpetrated sexual coercion. 105 The same is not true for all forms of interpersonal violence. For example, the onset of child sexual abuse (CSA) perpetration shows a different pattern, a bimodal one, with onset peaking first in adolescence (at ages [13][14] and again in middle adulthood (at ages [32][33][34]. US data find that while most perpetrators of CSA are adult men, over onethird of perpetrators are male adolescents. 204

Trajectories of violence perpetration
What are the typical patterns of violence perpetration over time? The trajectories of individuals' use of abuse and violence are shaped by their life circumstances, experiences, and choices. The research on patterns of domestic, family and sexual violence reveals, however, that there are enough commonalities to map and group together perpetration trajectories in three main ways: non-abusive, low-level or high and persistent. 101,[200][201][202] There are variations in perpetration of specific types of abuse and violence and across trajectories, with some people demonstrating consistent and stable use of abusive behaviours, while others increase or decrease the frequency and/ or severity of violence. 101,200,201,205 A minority of people desist or increase their abusive behaviours enough to be considered to have changed the trajectory of their perpetration. 101,202 Among adolescents who are using violence in their dating relationships, common patterns are for this use of violence to increase over time into young adulthood, or to increase and then decrease. These studies suggest that older adolescents and young adults have a greater likelihood of intimate partner violence perpetration than younger adolescents.
• In a US longitudinal study among adolescents in 8th-12th grades (beginning when they were aged 13-14 and ending when they were aged 17-18), psychological abuse increased while physical and sexual abusive behaviour peaked and declined. 206 • In a second US longitudinal study among adolescents, perpetration of physical adolescent dating abuse increased from 8th to 10th grades (from ages 13-14 to 15-16), but then declined from 10th to 12th grades (from ages 15-16 to 17-18). 207 • A US longitudinal study among adolescents aged 12-18 at baseline found three trajectories over four years of data collection: (1) non-dating (37.3% of the sample); (2) increased dating, including both with and without perpetration of adolescent dating abuse (44.6%); and (3) high stable dating and adolescent dating abuse perpetration (18.1%). Among the second of these three groups, the likelihood of perpetrating adolescent dating abuse started at a low level but increased over the adolescent developmental period. Among the third group, the risk of perpetrating adolescent dating abuse was consistently high over multiple years. 201 There is similar diversity in the trajectories of perpetration for sexual violence among young adults, although persisting with perpetration is a common pattern. For example: • US studies among men at university find four patterns: consistently low or no sexual violence perpetration, decreasing perpetration, increasing perpetration, and consistently moderate or high levels of sexual violence perpetration across time. 150,208,209 • In a US longitudinal study involving four waves of data collection among 13-to 25-year-olds, the most common pattern was for young people's patterns of perpetration and non-perpetration of sexual violence to persist over time. Some young people deescalated their perpetration over time, for example, moving from perpetrating multiple forms of sexual violence to perpetrating only sexual harassment or not perpetrating at all, although this was less common. Escalation was the least common pattern. 101 Little is known about rates of violent reoffending, that is the extent to which individuals will repeat or continue their violent behaviour. There are data on recidivism, the likelihood of violent offenders reoffending after a criminal conviction or some other formal sanction, based on legal system and perpetrator treatment data. However, these data are likely to significantly underestimate actual rates of reoffending for domestic, family and sexual violence, given the very high levels of underreporting of these crimes, as well as the restricted definitions of recidivism used. 192 Some models of violence perpetration, for sexual aggression for example, have relied on notions of a static propensity towards violence among perpetrators. They assume that perpetrators are persistent in their perpetration across the life course, their propensity to commit crime is relatively stable across their lives and variations in their rates of offending are a function of fluctuating circumstances in opportunities and they start and repeat offending for the same reasons.210 Instead, there is evidence that rates of ongoing offending or recidivism are variable and uneven (although studies on recidivism rates largely come from police data and may underestimate actual rates), offending patterns are heterogenous even among persistent or repeat perpetrators, and some offend intermittently, with brief periods of offending followed by long periods of non-offending.210 However, the finding that, even among persistent offenders, sexual aggression remains relatively inconsistent and intermittent may reflect criminal legal data and its limitations, given that other studies find more stability in patterns of perpetration.
Some studies with community samples find substantial rates of sexual violence reoffending. For example: • In a US study of male university students surveyed at the end of each of their four years in university, 30% had committed an act of sexual coercion or assault over this period and, of these, two-thirds (68%) had engaged in repeated sexual coercion or assault. 192 • In a study among male school students aged 17-20 in Sweden and Norway, of those who reported having perpetrated sexual coercion, in the Swedish sample, 51% indicated that the behaviour occurred only once, whereas 12% reported sexually coercive behaviour on more than five occasions. Figures for the Norwegian sample were 53% and 14%, respectively. 105 Repeat offending by sexually violent perpetrators, and more severe offending, is more likely if their offending began earlier and was more severe, as studies among both convicted sex offenders and community samples find. 192 These findings suggest that: • Most individuals who commit sexual coercion or rape as young adults, for example at university, will continue to do so especially if, like the vast majority of perpetrators of sexual violence, they avoid criminal detection and sanction. • Many will have begun perpetrating sexual violence before adulthood, as teenagers. • The younger they were when they began perpetrating, and the more severe their violent behaviour, the more likely they are to continue offending and to perpetrate more severe forms of violence. 192

Trajectories of violence within a relationship
A different dimension of perpetration involves the pattern of violence perpetrated over time in an intimate relationship or family. There are diverse patterns and trajectories of perpetration in the use of domestic and family violence. Where individuals do escalate their violence against an intimate partner, its frequency and severity may differ as well. Some perpetrators escalate behaviours over time, especially where their use of violence is serious and persistent, while other perpetrators' patterns of escalation, and de-escalation, are episodic. 205 Looking at male perpetrators who have used lethal force against an intimate partner, an Australian study identified three main patterns of perpetration: fixated threat, persistent and disorderly, and deterioration/acute stressor. 211 The most common perpetration trajectory that led to intimate partner homicide (40%) was persistent and disorderly, characterised by complex histories of trauma, cooccurring personal factors, such as mental health challenges, and contact with the criminal legal system. Fixated threat was the second most common trajectory (33%), characterised by controlling and jealous behaviours, external appearance of being functional and respected, low contact with criminal justice, and relationship separation as a point of escalation. Deterioration or acute stressor was the least common (11%), characterised by having significant emotional, mental or physical health problems that were exacerbated by a life stressor that deteriorated their wellbeing and attitude towards their victim in an otherwise previously 'happy' relationship. 211

Reducing and stopping one's use of violence
Desistance is the gradual cessation of criminal behaviour. [212][213][214] Desistance describes the process of lessening criminal behaviour, whether its frequency or severity or both, whilst cessation is used to describe the end of criminal behaviour. Cessation is usually preceded by a period of de-escalation where criminal behaviour is less serious and/or less frequent. [212][213][214] It is well established that crime declines with age throughout the whole population, showing an 'age-crime curve', and so most offenders eventually desist from crime. [212][213][214] Serious offending, including violent crime, tends to peak in late adolescence/early adulthood then begins decreasing thereafter. 212,213 Although this is the general trend, there are differences in magnitude and duration of offending trajectories over the life course. The agecrime curve describes the modal pattern of offending, but some offenders deviate from this; some initiating crime younger, more frequently, and/or ending later in life. But, desistance from crime is the norm. 212 However, studies that show that most offenders 'age out' of criminal behaviour most commonly refer to generic criminal behaviour and/or violent offending outside of the home. The desistance literature only rarely refers specifically to intimate partner violence, family violence and sexual violence.
One study found that sexual offending also decreased with age, but the rate of decline was more gradual. 215 For those who had committed rape, the offenders were younger when their offending peaked/began to decline, whilst child sexual abusers did not begin to decline in their offending until age 50. The three factors were theorised as being key to rape and child sexual abuse: deviant sexual interests, opportunity and low self-control. 215 These three factors decline with rapists with age, whilst for child sexual abusers opportunity did not decline until much later in adulthood -explaining the difference in desistance. 213,215 Despite evidence showing that violent offenders usually end their offending, the causes which drive and maintain the process of desistance are unclear. 213 Violence desistance remains poorly understood due to definitional, operational and measurement inconsistencies. 212,213 Several of the conceptual frameworks used to explain desistance from criminal behaviour have not been specifically applied to violence and, therefore, may have limited relevance to determining factors that shape its cessation. 213 Although the process of desistance varies for different people, the literature shows that several factors may be instrumental in the cessation of violent behaviour.

External desistance
For violent offending in general, there is evidence that marriage, entering a stable relationship, parenthood, employment, education and/or military service are important factors for desistance. 212,213 This may be also due to these life events causing a separation with delinquent peers and/or other risk factors associated with offending. 212 These key turning points in life create social capital, social ties and informal social controls, which may make offending less acceptable. 212,214 Note, however, that in relation to domestic, family and sexual violence, some of these factors -such as entry into a relationship, marriage, parenthood and military service -may in fact be associated with violence commencing or escalating.

Internal desistance
Internal desistance is shaped by factors such as the cognitive transformation that can occur during therapeutic intervention. 214 This could also include identity transformation, rational decision-making and maturation -and these are all factors that are linked to the intentional and self-initiated changes brought about by individuals to desist their criminal behaviour. 212

Deterrence
Desistance may also be brought about by intervention by the criminal legal system. 214 For instance, Harris argues that the clearest distinction between sexual and non-sexual offenders is their treatment by the criminal legal system. 214 Desistance may therefore be motivation by the threat of another criminal legal sanction. 214

Gaps in the research
External desistance, internal desistance and deterrence may all intersect and factor into an individual's desistance, but are largely focused on criminal behaviours outside of the home -such as property crime, youth-related crime and violent crimes against non-family members. More research is needed on the desistance factors that contribute to cessation of intimate partner violence, family violence and other forms of violence against women. 213 External desistance factors like marriage, for example, seem unlikely to contribute to a cessation of intimate partner violencethe opposite may be true. More research is needed on how protective factors mitigate the risk of violence and what/how they contribute to maintaining cessation of violence. 213 Moreover, few studies examine desistance for female offenders. Research to date appears to show that women and men show a similar pattern of decline, but women's involvement in crime is far less frequent and tends to begin later in life, resulting in a sharper decline in the age-crime curve. 212 In theories of desistance, the role of the individual in the desistance is elevated whilst structural facilitators are given less weight -individuals' intentions and actions may be limited by their social circumstances, inhibiting their desistance. 212 How do those who occupy historically marginalised positions, such as women, First Nations people and LGBTQI+ people, desist from violent behaviour when they exist in compromised social positions with less access to protective factors associated with desistance? More research is needed to explore whether the benefits of factors associated with desistance extend to those who experience multiple and intersecting forms of oppression and disadvantage.

Being held to account
This report thus far has noted that substantial proportions of people engage in violent and abusive behaviour, including acts that meet legal definitions of crime. Yet very few people experience any kind of formal sanction for such behaviour.
Few perpetrators are held to account for their crimes.
Few individuals who perpetrate domestic or sexual violence are ever convicted of this crime. For example, in a major US longitudinal study, although 8.8% of males reported having perpetrated sexual assault, none had ever been arrested for such offences.96 This study's summary of the lack of consequences experienced by perpetrators can be applied more widely: 'Most individuals who actually commit an act that an official statute has labelled as an offense are never arrested. Of those who are arrested, a large percentage are not convicted despite having committed a statutory offense.' The quote continues: 'Further, many of those who are convicted are allowed to plead guilty to a lesser charge and are therefore never charged with or convicted of a specific sex offense. As a result, individuals who are finally convicted represent only a relatively small percentage of those who commit a sex act that is illegal where they live.' 96 The attrition of perpetrators through the criminal legal system begins before potential arrest or conviction, because of underreporting by victims. Many victimsurvivors do not report the assaults they suffer to authorities. For example, Australian crime victimisation data show that about 76% of women who experienced sexual assault in the last 12 months did not report the most recent incident to the police. 24 Such patterns have four consequences. First, those individuals who use violence are rarely held to account for their actions and many will continue to perpetrate violence, against both existing and new victims. Second, victim-survivors do not receive formal justice. Third, our knowledge of perpetrators and perpetration remains limited, particularly to the extent that this knowledge is based only on the tiny proportion of perpetrators represented by convicted offenders. 13 Fourth, this sparse evidence base obstructs our efforts to prevent and reduce perpetration.
Comparing the numbers of people who are sexually assaulted in the past year and the number of individuals who are found or plead guilty to sexual assault and related offences each year, by one estimate, only about 1.67% of perpetrators plead or are found guilty. 13

Conclusion
This report's mapping of existing research on the perpetration of domestic, family, and sexual violence highlights that, although there is emerging knowledge of the extent and drivers of violence perpetration, there is a great deal that we do not know. There are, however, emerging insights about the perpetration of domestic, family and sexual violence evident in existing scholarship. We know that there are feasible ways to gather data on violence perpetration. Existing studies, largely from outside Australia, demonstrate that the use of sexual and domestic violence is common, particularly by men and boys and, to a lesser extent, by others. Perpetration is driven by risk factors at the individual, relationship, community and societal levels. There is considerable diversity among perpetrators and in perpetration, but also patterns of co-occurrence and overlap. Perpetrators often start young, and trajectories of perpetration often remain stable. Finally, very few of the many individuals who commit acts of domestic, family or sexual violence ever receive formal punishment, face repercussions for their abusive behaviour, or are held to account by people who know them and organisations with which they interact.
What research is needed to understand violence perpetration? Here are nine high-level recommendations for future research. They are complemented by a further discussion, in the Appendix, of the methodological and ethical issues involved in research on perpetration.
1. Data collection on domestic, family, and sexual violence should include substantive attention to perpetration -to the prevalence and character of violence perpetration.
2. This should include, in particular, a regular, nationally representative, populationbased survey of the use of domestic, family, and sexual violence. 12 This would provide a consistent and comparable dataset on progress in reducing domestic, family and sexual violence. There should also be greater attention to potential data sources associated with the legal system and perpetrator programs, although other community-based methods are vital given just how few perpetrators enter such systems and programs.
3. Far more research is needed on the character, dynamics, pathways and drivers of perpetration. Focused studies of these aspects of perpetration are a necessary complement to a national perpetration survey, providing more intensive and complex forms of data collection than is possible in a national survey. 7. There should be greater examination of perpetration among different demographic groups and populations, including women, sexual and gender minorities, and others. 8. Research on violence perpetration should include both exploration of the predictors of desistance from and the cessation of violence (in the name of secondary and tertiary prevention) and the factors that protect against initial perpetration (in the name of primary prevention).
9. Research on perpetration should include greater attention to the more macro levels of the social ecology; exploring risk and protective factors for violence prevention at the level of organisations, neighbourhoods, communities and societies, and the interactions among them. 147 To prevent and reduce domestic, family, and sexual violence, we must have accurate and comprehensive knowledge of the prevalence, dynamics and drivers of perpetration. As two of this report's co-authors have noted elsewhere: 'If we do not know how many people are perpetrating domestic and sexual violence and why they are perpetrating it, how can we prevent it? If we do not know the conditions, contexts and drivers for the perpetration of violence, how can we prevent it? […] the lack of perpetration-specific data inhibits our ability to have a meaningful influence on domestic and sexual violence levels. We do not know enough to target effectively those people at risk of perpetrating such violence. Nor do we know enough about people who are beginning to use violence and who, without intervention, might continue to perpetrate and escalate violence until they come into contact with the justice system. We do not know enough about who to target and the protective factors on which we will need to build to divert people from perpetrating. ' 12 Preventing and reducing the perpetration of domestic, family and sexual violence must be at the heart of Australia's violence prevention efforts. After all, by definition, to prevent violence is to prevent violence before it is perpetrated. Rather than seeing domestic, family and sexual violence as problems only of victims and victimisation, we must see them also as problems of perpetrators and perpetration. In turn, we must collect robust Australian data on perpetration. As the Executive Summary to this report emphasises, doing so will have profound implications for the character and effectiveness of our violence prevention efforts.

'Violence is a problem for victims, but it is not a victims' problem.' -Lula Dembele, Survivor Advocate
It is time to make perpetrators and perpetration the focus of our national efforts to prevent and reduce domestic, family, and sexual violence.

Appendix: Measuring violence perpetration
What are the methodological issues in defining, measuring and analysing violence perpetration?
In this Appendix, the report identifies key methodological issues in producing data about the perpetration of violence. It begins with strategies based on the collection of original data on perpetration, such as quantitative surveys, before examining issues in drawing instead on administrative data.

Mapping perpetration
It is certainly possible to collect robust data, including at a nationally representative level, on the perpetration of violence. A strong recent example is the UN Multi-country Study on Men and Violence in Asia and the Pacific. Based on interviews with 10,000 men in Bangladesh, Cambodia, China, Indonesia, Sri Lanka and Papua New Guinea, this study collected both country-level and comparative data on physical and sexual violence. 64 While nationally representative studies on the perpetration of domestic or sexual violence are rare, the UN Multi-country Study and other studies demonstrate that such studies are both feasible and productive.
At the same time, producing data on the perpetration of violence requires careful attention to a series of methodological issues. The first is perhaps the most fundamental -what counts as violence.

Definitions and measures
Public and scholarly understanding of the nature and extent of violence perpetration is shaped by how researchers define, measure and analyse violence. 20 Inconsistent definitions and conceptualisations are used in the field, giving researchers an ongoing challenge of deciding what behaviour should be counted as violence. 216,217 For example: • Some studies define and measure any act of physical aggression in couple relationships as domestic violence, whereas others include an assessment of impact or consequences or the presence of coercive and controlling behaviours. • Acts of sexual coercion may involve verbal pressure, incapacitation, the use or threat of physical force or other means of coercion. Research that relies on the legal definition of rape may exclude an assessment of some of the broader forms of sexual coercion. 191 • Definitions of child sexual abuse may differ in childhood age criteria or type of behaviour being measured (i.e., forced intercourse or non-contact abuse). [218][219][220] As a result, there is a wide range of estimates of the prevalence of perpetration, muddying understanding of the nature and scope of the problem. 217 Measures of sexual violence perpetration may therefore lack validity when being applied to women's use of violence. 217 It is also clear that men and women experience violence differently, therefore measures that consider the unique experiences of men and women need to be developed. 226 To better understand and respond to violence, it is useful to define and conceptualise violence in a way that enables the measurement of multi-dimensional facets of violence; forms, context, frequency and severity of harm. [226][227][228] A broad definition involving 'any physical or sexual act that may cause harm' fails to inform the severity or context in which the act occurred, thus diminishing the overall impact of violence. 226

Measures of particular forms of violence
There is a lack of standardisation of measures of violence. Violence measurement scales are each designed to measure different forms of violence and are largely driven by different theories. 229 Different measures therefore produce very different estimates of prevalence and patterns of violence. For example, a study comparing sexual aggression perpetration scales found significant differences in prevalence rates among a sample of US college men. The authors suggest these differences may be a result of the differing social contexts being assessed and small differences in item wording. 230  41 • CTS studies' apparent findings of gender symmetry in victimisation or perpetration may reflect the influence of relatively low threshold items in the measure rather than potentially more serious or injurious behaviours. 234 For example, among the sexual coercion items, responses to the item on insisting on sex may reflect less harmful or even innocuous experiences among survey respondents. • The CTS distinguishes between minor and severe acts in the Physical Assault sub-scale, but this is based on general perceptions of their seriousness rather than evidence of which of these acts is most likely to cause injury or other harm. 40 • CTS studies often simply count the cumulative number of violent acts used, rather than also counting the frequency of violent acts and then weighting them by severity. Doing so is more likely to capture actual patterns of domestic violence. 70 • The CTS omits other violent behaviours known to be both serious and harmful, such as sexual violence, threats to harm or actual harm perpetrated against children or pets or other loved ones, stalking, and intimate homicide. 40 The CTS excludes many forms of financial abuse. • CTS studies exclude incidents of violence that occur after separation and divorce. They therefore omit various forms of postseparation abuse, although the period during and after separation often involves an escalation of abuse. 161 • Finally, CTS studies do not highlight situations in which there is no physical violence, but one partner is using coercive control against the other. We will return to this later.

Measures focused on violent acts
The original CTS was revised as the CTS-2 to improve concerns of validity, adding an assessment of sexual coercion and physical injury and revised wording to enhance clarity. 235,236 However, the new items are still limited. The sexual coercion scale only asks questions about penetrative intercourse and omits other sexually coercive acts, including unwanted groping or kissing and drug and alcohol-facilitated rape. In any case, many researchers do not use these additional items. 40 Because the CTS was designed to be used across a wide variety of relationships and families, and to focus on short reference periods such as the last year, it concentrates on more common but also less severe forms of violence in intimate relationships. This means that the measure excludes some of the most serious forms of violence, as noted above. 40 The CTS focuses on more common and more minor forms of aggression and, because these show less gender asymmetry than more severe forms of violence, CTS studies present a less accurate picture of gender differences in perpetration and victimisation. 40 Two further issues are, first, the reliability of participants' reports on perpetration and victimisation and, second, sampling. There is evidence, furthermore, that women are more likely than men to admit to their own perpetration of violence. 20 Male perpetrators are more likely than female perpetrators to underreport their use of intimate partner violence. 40 Studies find, for example, a larger gap between male partners' reports of perpetration and female partners' reports of victimisation than the reverse, female partners' reports of perpetration and male partners' reports of victimisation.
Regarding sampling, some researchers contend that individuals either perpetrating or suffering significant domestic violence are likely to drop out of survey samples, such that the only violence that will be detected is more minor, reciprocal forms of aggression in relationships, as we note below.
The validity of studies based on the Conflict Tactics Scale continues to be debated, due to the scale's lack of consideration of context and the contrast between the findings of CTS-based studies and those based on other measures. 226,231 Because the CTS treats violence in a highly decontextualised and abstracted way, the method is said to inevitably produce findings of apparent gender equality in domestic violence, while obscuring the actual patterns, meaning and impact of violence by men or women. 20 Jones and colleagues suggest that the validity of CTS-2 findings could be improved with additional measures that inform context. 225 Other researchers point to other measures of intimate partner violence that are more likely to produce valid data on patterns of perpetration and victimisation. 41 In any case, when they do gather data about the consequences of partner violence, even CTS-based studies typically find that men's violence against women causes more frequent and more severe injuries, greater fear, and worse psychological consequences than women's violence against men. 22

Sampling
Sample non-response may affect the accuracy of data collection. For example, community and population-based surveys of domestic violence are said by some researchers to be shaped by high rates of refusal, particularly among individuals either practising or suffering severe and controlling forms of intimate partner violence; what some call 'intimate terrorism' or coercive control. Individuals using violence against a partner, and those suffering violence at a partner's hands, are less likely than others to participate in such surveys, particularly where more severe violence is involved. 22 Such patterns mean that less severe forms of violence, such as situational couple violence, dominate in general surveys while more severe forms are visible instead in hospital, police and legal data.
This pattern has been said to help explain why community-based domestic violence surveys, many of which use the Conflict Tactics Scale (CTS), often find gender symmetry at least in men's and women's overall use of violence against intimate partners. Situational couple violence -the kind of violence that dominates in general surveys -is gender-symmetrical in terms of the extent of the use of aggression. In contrast, intimate terrorism -the kind of violence that is absent from general surveys, but more visible in hospital, police and legal data -is highly asymmetrical, with nearly all intimate terrorism perpetrated by men against women and only rarely perpetrated by women. 22 These patterns help to explain why some sources of data find significant gender asymmetries in domestic violence perpetration and victimisation while other sources find apparent gender symmetry. However, some researchers warn against overstating the influence of sampling on the divergent findings of different studies of domestic violence. Hamby 40 notes, first, that police and other criminological data include minor as well as more severe incidents and, second, that there are many community-based surveys that do not show gender symmetry in rates of intimate partner violence. Instead, she argues, the most significant explanation for why one class of measures based on the CTS produces gender symmetry when several other classes of measures do not, is that the CTS is poorly designed and overly broadly worded.

Self-reports of perpetration
Data on perpetration are often based on people's own reports of their use of violence. The data summarised in this report illustrate that certainly it is feasible to conduct research in which people acknowledge their use of domestic, family and sexual violence. Substantial proportions of survey respondents will report their own perpetration of violence, particularly if asked questions about specific behaviours. For example, in a systematic review across 77 studies among over 25,000 male university students, an average of 29.3% of the young men reported having perpetrated sexual violence in their lifetimes. 10 People are far more likely to acknowledge perpetration if asked questions about specific behaviours than if asked general questions about ever having perpetrated rape or domestic violence. For sexual violence, the most common measure of perpetration is the Sexual Experiences Survey -Perpetration, 237 used for example in 13 of 16 studies in a systematic review of men's perpetration of sexual violence in higher education institutions. 15  Very few men will report that they have 'raped' a woman, but sizable proportions will report using one or more of the coercive methods above to obtain a sexual act. In the systematic review among male university students, close to one-third (29.3%) reported perpetrating these forms of sexual coercion. In contrast, in the few studies in the review that also asked, 'Do you think you may have ever raped someone?', only about one in 100 men (from 0.4% to 1.5% with a mean rate of 0.9%) said yes. 10 Self-reported data on perpetration face the obvious challenge of underreporting. People may underreport their use of violence because violence is stigmatised, it is often criminal, or they may not recognise or name their behaviour as violent. Social desirability bias affects people's under-reporting of violence. For example, some early studies found that fewer men reported sexual assault perpetration when they were interviewed face to face. 234 Aware that they are acknowledging or endorsing a socially undesirable behaviour, some participants will report perpetration at a lower frequency or lesser severity than what occurred. 216 Reluctance to report acts of violence represents a conscious decision to preserve one's self-image or avoid perceived legal consequences. 216,238 On the other hand, a lack of memory or understanding as to the coercive or aggressive nature of one's behaviour contributes to unintentional nondisclosure. 238 In addition, some researchers suggest that the impact of socially desirable reporting on self-reports of violence has been overstated. 234 Nonetheless, several strategies can be implemented to enhance the accuracy of data collection in violence perpetration studies. For example, anonymity increases rates of self-disclosure in sensitive research topics. 216 A study found that using an anonymous assessment condition for selfreport of sensitive topics reduced the rates of participant termination and 'prefer not to respond' responses and increased the rates of frequency and incident data. 239 A more exotic strategy is the use of a 'bogus pipeline', in which participants are connected to devices that are described (falsely) as having the ability to identify deception. Studies find that men connected to the bogus pipeline are more likely than men in a control group to admit to sexual assault behaviour that meets the legal definition of rape. 234 The accuracy of self-reports of violence perpetration is also shaped by respondents' own attitudes towards violence, although this can lead to both more and less accurate reporting. On the one hand, people with more lenient attitudes towards violence may be more likely to report their own use of violence because they are less concerned about the social undesirability of this behaviour. On the other hand, people with more lenient attitudes may be less likely to report their use of violence because they are less conscientious in their reporting or because they are responding casually or carelessly to questions about behaviours they consider acceptable. 234 Concerns about the accuracy of self-reports of violence perpetration can equally be applied to the accuracy of self-reports of violence victimisation. Indeed, some studies find that people, both men and women, are more reliable reporters of their perpetration than their victimisation. 234

Item wording, framing and order
The words used in surveys to ask about violence perpetration make a difference to the accuracy of disclosure. 240 Using words such as 'rape' or 'sexual assault' may not align with the participants' perceived experience and contribute to non-disclosure. 241 Words that imply the least amount of culpability increase rates of reporting. 242 To minimise the risk of perceived stigmatisation and misinterpretation, terms such as rape have been omitted from the Sexual Experiences Survey self-report measure of sexual violence. 240 In child sexual abuse research, children respond better to behaviour-specific wording of questions about having been touched or kissed, rather than using words like abuse, violence or molestation. 243 Another issue is the subjectiveness of a person's experience and interpretation of survey questions which can generate varied estimates. However, clear use of language can improve the person's understanding of the concepts, thus enhancing the data's validity. 191 Question phrasing also affects recall and rates of self-reported perpetration. A study examining frame of reference in self-report surveys found higher rates of perpetration among men assigned a questionnaire asking about a list of tactics to obtain different types of forced sex, while men assigned a different version, asking first about the type of sex obtained and then the tactic, reported much lower rates, despite the overall survey information being identical. 241 It is possible that the first frame of reference used in the survey activates the person's memory or coding of an event. Tactics used to obtain sex may be a key component in sexual aggression scripts, therefore using these as a first frame of reference may activate relevant memories more than the type of sex. 241 Question or item order is a necessary consideration in self-report survey design. A study examining item order effect in different versions of the Conflict Tactics Scale found that randomisation of scale items (after the initial content on negotiation tactics) generated higher disclosure rates compared to other versions that ordered items in increasing severity. Ordering items by increasing severity may indicate to the participant a perceived notion of what is deemed less socially desirable, thus impacting under-reporting. 236 Also, long item lists can contribute to participant fatigue and reporting error. 244

Administration mode
Self-administered surveys are a favoured form of data collection in violence perpetration research as they provide participants with a high level of perceived confidentiality compared to face-to-face interviews, thus increasing honesty and response rates. [245][246][247] Self-administered surveys are also cost and time effective compared to face-to-face interviews. 242 However, if a researcher is able to form a connection with a participant in a face-to-face setting, there may be increased willingness to disclose sensitive information. 248

Recall and memory
Recall accuracy of past incidents of violence decreases as the length of time since the incident increases. 249 As a result, most measures assess more recent experiences (i.e., past year or past six months). The Conflict Tactics Scale, for example, was designed to assess past year experiences as a reference time period of over a year was seen as too long for accurate recall. 249 A study examining the accuracy of recalling sexual partners and behaviours over varying time periods found a lower error rate for the past month compared to longer time periods. Child sexual abuse studies involving adult participants are more prone to recollection bias compared to studies assessing the more recent experiences of children or adolescents. 219 Structured methods for eliciting personal information, such as the Life History Calendar method, can enhance the quality of data as they cue memory recall associated with the initial personal factors often associated with incidents of domestic violence. 249

Quantitative and qualitative methods
Quantitative research methods are commonly used to capture data on the prevalence of violence perpetration, however, qualitative research is useful to understand other aspects of domestic, family and sexual violence. For example, interview strategies can be utilised to assess and understand a perpetrator's perspective on their own use of intimate partner violence including their feelings, emotions, cognitions and conditions associated with their behaviour before and after an incident. 250,251 Understanding these factors is useful to inform treatment and prevention for perpetrators of violence. 251 In addition, qualitative methods can enhance the validity of assessments of the extent of perpetration, as two examples show. An interview study among female university students found that much of the violence they reported perpetrating was actually horseplay, mock violence or self-defence. A second qualitative study documented problems in how respondents interpreted items on a quantitative survey, allowing improvements to such quantitative methods. 70

Longitudinal studies
It is important to measure violence perpetration across the lifespan to improve the accuracy of prevalence rates and situational understanding. 252 Longitudinal studies also help to correct the gender biases of 'snapshot' surveys based on data collection at a single point in time or in a single relationship. Point-in-time studies may systematically underestimate men's perpetration and overestimate women's. For example, a longitudinal study among young adults found that: 'Men's violence in relationships leads to higher dissatisfaction in female partners than women's violence does for male partners. In turn, women who experienced violence were actually more likely to leave a relationship than men who experienced violence. One implication is that asking about violence at a particular point in time in a current relationship will be more likely to identify women's violence (since men are still in relationships with violent women) and less likely to identify men's violence (since women are more likely to have already left those relationships).' 70 Another study of intimate partner violence supports this in finding that men reported higher victimisation rates than women for victimisation in the current relationship, but women reported higher victimisation rates than men for victimisation in prior relationships. 70 Longitudinal studies of perpetration therefore are a useful complement to cross-sectional ones. Collecting survey data on violence perpetration at multiple timepoints is necessary to inform the complex dimensions of violence, including its behavioural patterns, frequency, escalation and desistance. 200,221 Administrative data An alternative source of information on violence is administrative data collected by police, hospitals and other services. It is important to note these data's limitations, however, as a source of information on the prevalence or patterns of violence perpetration: • Administrative data often focus on victimisation. • Administrative data present only a tiny proportion of the actual scale of violence, as all forms of domestic, family and sexual violence are underreported. Less than 40% of women who experience violence globally report it to police or healthcare services. 253,254 These reports also will not necessarily translate into criminal charges, meaning that these data fail to capture perpetration. • As administrative data are only gathered once people come into contact with the legal system or with medical and other services, these data are more likely only to capture the most severe cases. • Victims who engage with services and/or report to police are more likely than other victims to have adequate social supports or to have overcome various obstacles to seeking help. Such victims are often engaged in multiple systems, meaning that they are counted multiple times in administrative datasets, whilst other victims are not counted.
• Administrative data are limited by inconsistent legal definitions across jurisdictions regarding what counts as a criminal offence or as criminal violence, such as breaches of domestic violence orders. • Data collected by police and other legal agencies or systems are shaped by poor understandings of domestic violence, poor responses to victim-survivors and perpetrators, and by wider racist and unjust cultures and practices. The misidentification of victims as perpetrators is a powerful example of this, as discussed in more detail below. • Data collected by legal agencies or systems are likewise characterised by the over-policing of First Nation and ethnic minority communities, as described below.
The way that administrative data are collected -particularly by police -means that these data may be flawed. For example, the misidentification of victims as the primary aggressor often results in inappropriate legal sanctions against the primary victim. 255 This often occurs where there are conflicting claims of who perpetrated the violence, or apparent mutual violence, and often, a result of police intervention. 255 Data are generated through these interactions by police, courts and others, which can illustrate a picture of perpetration that is flawed, incomplete and/or does not capture drivers and motivations.
Misidentification often stems from a system's failure to contextualise domestic, family, and sexual violence within a pattern of power and control. A response to violence as single incidents, particularly when a primary victim may have used retaliatory or pre-emptive defensive violence, can make misidentification more likely.
The misidentification of female domestic violence victims as perpetrators is starkly evident in recent Australian experience: • Close to half (44%) of all cases of female domestic and family violence related deaths from 2015 to 2017 in Queensland were of women who had previously been identified as a respondent to a domestic violence order. 255 • The deceased person had been recorded as both a respondent and an aggrieved party in domestic violence orders in nearly all of the domestic and family violence related deaths of Aboriginal people. 255 Resulting administrative data can thus suggest that there are higher levels of violence perpetration among women or certain groups of women than is the case. Failure to appropriately identify the primary domestic abuser disproportionately impacts First Nations women, who are also more likely to encounter structural racism in their interactions with the criminal legal system. First Nations women are overrepresented as respondents/ defendants on domestic violence orders and in the breach of these orders. 255 Moreover, individuals collecting administrative data may have particular expectations of what particular forms of violence, such as domestic violence, 'look like'. This expectation may be embedded within Western ideas of nuclear family dynamics that can fail to account for different relationship structures within different cultural and ethnic groups. Administrative data place the responsibility on individuals to determine the primary aggressor and the forms of violence when they attend an incident. Therefore, in these cases, racialised sexism and other forms of discrimination shape the data.
As particular communities are over-policedsubject to disproportionate police attention and more likely to face legal sanctions -certain groups are more likely to come into contact with police and the legal system. This then overrepresents particular groups within the administrative data. For example, black men in lower-socioeconomic settings may be more likely to come into contact with police, whilst wealthier white men may have the means to shield themselves from contact with the criminal legal system. Male perpetrators who are white are less likely to be held accountable and less likely to be arrested and charged than male perpetrators who are black or men of colour. 19,256 In short, structural racism forms the context for the collection of administrative data, contributing to skewed and unreliable data.

What gets researched?
The final issue explored in this Appendix is the politics of data collection and, in particular, what gets researched.
The kinds of data and research available on violence and violence perpetration are shaped, in part, by political interests and social trends. What is researched is shaped by what will be funded, and funders' interests often align with what is topical or the focus of legislation.
As researchers shape their research to mirror the priorities of funders, this may skew the available data as certain types of data and particular forms of violence are focused upon, whilst others receive less attention. For example, recent research by the Sexual Violence Research Initiative 257 found that funding for research on violence against women is rare, and that current funding models rarely valued research as essential to addressing violence against women. Of the available violence against women research funding, most is dedicated to research on approaches to prevention and is centralised within high income countries. 258 The influence and competition for available funding then results in research, data and funding gaps.
The evidence base and available data are also shaped by advocacy. Advocacy draws necessary attention to particular cases and forms of violence. This can result in legislative change, but it also influences research priorities. For example, feminist advocacy in response to high-profile rape cases in the USA arguably led to volumes of research exploring sexual violence in US colleges (universities).
There is a large volume of research on the extent of sexual violence perpetration among US university students and not a single study in Australia. In Australia, high-profile domestic violence homicides, as well as public commentary by journalist Jess Hill and others, led to a national conversation about the criminalisation of coercive control. Possibly in response to this, Australia's National Research Agenda (2020), which determines Australia's National Research Organisation for Women's Safety (ANROWS) funding and research, includes multiple questions in relation to coercive control. In contrast, the Global Shared Research Agenda for research on violence against women and girls in lowand middle-income countries (2021) makes no mention of coercive control.
Whilst the influence on research of funders' priorities, advocacy and topicality is important and can make valid contributions to the evidence base, it can also result in the neglect and underfunding of important areas and issues. The literature presented in this report reflects the greater volume of research on victimisation and perpetration among some groups and settings more than others. The report reflects an evidence base where there are little available data on perpetration in general and on particular groups and relationship dynamics.