Developing and Testing New Domestic Abuse Questions and Approach for the Crime Survey for England and Wales

Abstract

Previous research highlighted that a fundamental rethink of the measurement of domestic abuse was needed in the Crime Survey for England and Wales (CSEW). The research reported here aimed to develop and test new questions on domestic abuse for the CSEW to improve the headline prevalence measure, including frequency of abuse, to develop a way of measuring controlling or coercive behavior within the overall prevalence measure, and to develop a measure of the impact of abuse. The research included focus groups and interviews with victims (n = 27) to assess a set of draft questions and cognitive testing of revised questions with victims and the general public (n = 42). A final set of 24 questions was developed for use with victims of both intimate partner and family abuse, with an additional question for family abuse. The new questions were found to echo victim experiences and were deemed acceptable and reliable measures by victims and the general public for domestic abuse, including controlling and/or coercive behavior and impact. An analytical approach was recommended to improve the headline prevalence measure of domestic abuse by establishing ‘high’ and ‘low’ abuse profiles using measures of both behavior and impact.

1. Introduction

The CSEW is a representative annual victimization survey administered by the UK Office for National Statistics (ONS), which asks around 34,500 people aged 16 and over resident in households in England and Wales about their experiences of a range of crimes in the 12 months prior to the interview, as well as their attitudes towards different crime-related issues and perceptions of crime. In contrast to police-recorded crime, the survey also captures incidents that are not reported to the police, and the estimates produced are unaffected by changes in police recording practice. Since 1996, it has provided measurement of domestic abuse via dedicated domestic abuse self-completion modules. There are two relevant modules: the ‘Domestic abuse, sexual assault, and stalking’ module (asked every year) and the ‘Nature of partner abuse’ module (asked every three years), with the most recent and continuous iteration introduced in 2004 (see Appendix A for details).

The measurement of DA in the context of national population surveys has been the subject of academic debate for many years (for a summary, see Myhill 2017). Briefly, critics have contended that questions relating to behavioral acts of physical and psychological aggression fail to capture what many victims experience as continuous abuse, and inadequate capture of context and impact leads to an obscuring of the gendered nature of abuse in headline measures of prevalence. This issue has been brought into sharper focus in England and Wales since a criminal offense of Controlling and Coercive Behavior in an intimate or family relationship was enacted in December 2015. In the same way criminal law was identified as not accommodating the lived experiences of many victims (Stark and Hester 2019), the measurement of abuse more generally and coercive control specifically in population surveys has been identified as lacking the crucial elements of intentionality on behalf of the perpetrator and negative perceptions of behavior by the victim (see Hamberger et al. 2017).

Developing new question content on such sensitive topics using survey instruments is clearly methodologically challenging, but fortunately, it is also an area in which significant progress has been made in recent decades as the cognitive aspects of survey methodology have become better understood. Methods including qualitative interviews, cognitive testing, analysis of response latency, and behavior coding have become increasingly well-established question testing tools in the repertoire of survey researchers, helping us to identify latent problems with survey questions and their delivery often missed by conventional survey pre-tests (for a review, see Presser et al. 2004).

The current questions on domestic abuse in the Crime Survey for England and Wales (CSEW), specifically questions in the self-completion module and the nature of abuse questions in the wider survey, were previously explored for gaps relating to the issues mentioned above. Research carried out between 2020 and 2021 (Hester et al. 2023), which included consultation with victim-survivors and stakeholders, found that the CSEW questions did not capture domestic abuse accurately nor reflect lived experience (especially amongst Black and minority communities and people with disabilities) and needed updating; that coercive control needed to be at the core of domestic abuse measurement; and while physical assault is an important part of measuring domestic abuse, establishing frequency through counting events is probably unfeasible. It was difficult to link harm to specific perpetrators, and the research highlighted that victims may not necessarily recognize coercive controlling behaviors while still in the relationship. Moreover, while recognizing the importance of time series data, most participants consulted (victims, user groups, and wider stakeholders) prioritized improvement in the headline measure of domestic abuse. It was concluded that a fundamental rethink of the CSEW self-completion module is required, with a wider set of questions about domestic abuse and perceived impacts than those used since 2004. To that end, further research was commissioned by the UK Office for National Statistics (ONS) to: improve the headline prevalence measure of domestic abuse, including frequency of abuse; develop and test new ways of measuring controlling or coercive behavior within the overall prevalence measure; develop and test new measures of the impact of abuse. This article provides an overview of this recent research to develop and test new survey questions on domestic abuse (DA) in the Crime Survey for England and Wales (CSEW).

It is now well established that asking sensitive questions in social surveys involves a complex response process with significant potential for measurement error as a result of problems with question cognition, recall, judgement, and response, especially for sensitive topics and question items (e.g., Sudman and Bradburn 1982; Tourangeau and Smith 1996). Although in-depth interviews with victims-survivors have informed various measurement frameworks for DA and coercive control, much of this work has been undertaken outside England and Wales, and past iterations of the CSEW self-completion module have drawn on existing international surveys and literature as opposed to primary research with victims–survivors. We therefore applied focus groups and interviews with victims–survivors as ‘experts by experience’ to assess questions, and cognitive interviews and question testing with victims–survivors to assess potential measurement error.

This paper adds to the existing literature by reporting research that seeks to address key historical debates on the measurement of DA in population surveys. Specifically, we report on the development of questions to measure coercive control and impact that address weaknesses identified in previous measures (Hamberger et al. 2017). The research also utilizes cognitive interviewing with both female and male victims–survivors from a range of communities in order to explore differences and similarities in perceptions of abusive behaviors and their impact in intimate partner and family abuse contexts and to see whether a single set of questions can adequately reflect their lived experiences and permit the identification of abuse profiles that go beyond simple prevalence.

DA and Coercive Control

The term ‘domestic abuse’ has various meanings, leading to difficulties regarding measurement. In the context of intimate relationships, DA may be characterized by a continuum from, at one end, “negative behavior” that involves one-off events and longer -lasting situational couple violence to particularly harmful, inequality-producing, and liberty-constraining, coercive-controlling violence abuse at the other (Stark 2007). Although any negative behavior in relationships should not be condoned, we nonetheless have to differentiate between these different forms of intimate partner behavior, as they require different types and levels of support and intervention. The research on family abuse is much more limited, although it suggests different victim experiences depending on the relationship with the perpetrator and context. Abuse of parents by their adult children, for instance, may take the form of teenage-to-parent abuse or abuse by adult children of elderly parents. Family abuse may also be exemplified by forced marriage or ‘honour’-related abuse and by groups of family members. While family abuse involves a wider range of possible perpetrators, many of the behavioral features and impacts of intimate partner abuse may nonetheless be replicated in family abuse. These are important issues for consideration in the development of survey questions on this issue.

In December 2015, legislation was enacted in England and Wales regarding a new criminal offense of controlling or coercive behavior in intimate or familial relationships as part of the Serious Crime Act, Section 76. The new offense defines coercive control as “a purposeful pattern of behavior which takes place over time in order for one individual to exert power, control, or coercion over another” (Serious Crime Act 2015, Section 76, p. 3). It includes any combination of physical, sexual, or emotional DA behaviors such as “assault, threats, humiliation, intimidation, or other abuse that is used to harm, punish, or frighten their victim” (Serious Crime Act 2015, Section 76, p. 3). The offense involves behavior on the part of the perpetrator that has had a “serious effect” on the victim, meaning that it has caused the victim to fear violence will be used against them on “at least two occasions” or that it has had a “substantial adverse effect on the victims’ day-to-day activities” (Serious Crime Act 2015, Section 76, p. 3). Coercive control thus defines what is deemed particularly severe, dangerous, and harmful DA. The new Domestic Abuse Act 2021 in England and Wales extends the coercive control offense to cover post-separation abuse, introduces a new, statutory definition of domestic abuse that covers both intimate partner and wider family abuse, includes economic abuse, and also introduces a specific offense of non-fatal strangulation. To accurately estimate the prevalence of these offending behaviors, they need to be ncorporated (more robustly) in the CSEW question set, as the existing CSEW has only a few questions relating to coercive controlling behavior or impact (Myhill 2015; Hester et al. 2023).

The UK Office for National Statistics recently tested a set of questions on coercive control involving intent of the perpetrator, pattern of behavior, and impacts of coercive control, but these were found not to be robust (ONS 2019). Moreover, comparison with the content of other domestic abuse surveys internationally also shows that some, such as the European Union Agency for Fundamental Rights (FRA) survey (FRA 2014), the Australian Longitudinal Survey of Women’s Health (ALSWH 2019), or the National Intimate Partner and Sexual Victimization Survey (US) (Smith et al. 2017), contain much more detailed items within their headline measure of domestic abuse. It is imperative that a wider range of topics be considered for inclusion to ensure an up-to-date and robust approach, such as technology-facilitated abuse, withholding essential care, and threats to harm oneself. Online and digital forms of coercive control include the monitoring of social media profiles or emails, abuse via Facebook or Twitter, and using GPS locators (Harris and Woodlock 2018). Other domestic abuse surveys include questions on these forms of abuse; the Australian Personal Safety Survey (PSS 2017) asks, for example, whether an ex/partner ‘Controlled or tried to control where they went or who they saw (e.g., constant phone calls, GPS tracking, monitoring through social media websites)’. Threatening to harm or kill themselves if the survivor leaves is addressed in a number of other domestic abuse surveys, such as the Scottish Crime and Justice Survey (SCJS 2019), which includes the item ‘Threatened to kill or attempted to kill themselves as a way of making you do something or stopping you from doing something’.

An ongoing concern in the literature has been the importance of accurately measuring effects related to gender as well as wider intersectional inequalities (Sanchez-Prada et al. 2023). While questionnaire-based surveys tend to assume that participants interpret questions in similar ways, men may underreport their perpetration of violence and may overreport victimization (Ackerman 2016). Gender asymmetries are more apparent if incidence and impacts are also included (Myhill 2015). The types of violence and abuse asked about also shape the forms of DA identified. A focus on physical violence and aggression provides more gender symmetry (Archer 2002), while including questions about sexual violence and coercively controlling behaviors provides greater gender asymmetry, as these are more likely to be experienced by women than men and in the context of coercive controlling violence (Johnson 2008; Graham-Kevan and Archer 2003; Hamby 2016). Studies involving DA in LGBT+ communities suggest that there may be differences in experiences for individuals in same-sex relationships, from family, and/or if identifying as bisexual or trans (Martinez and McDonald 2021; Donovan and Barnes 2020). Individuals of different ethnicities may also experience DA differently (Gill and Walker 2020), and parenting status is a likely issue in post-separation DA (Holt 2017). Research suggests that women with disabilities may suffer multiple forms of abuse, including disability-related abuse and neglect such as withholding medications, denying access to mobility devices, neglecting personal care, and preventing attendance at doctor’s appointments. It also shows that abuse can also be contextual (Plummer and Findley 2011; Thiara et al. 2011). The Australian Personal Safety Survey (PSS 2017) includes a question on whether withholding essential care: ‘Deprived them of basic needs such as food, shelter, sleep, or assistive aids’). We therefore need to take into consideration how measurement can capture different behaviors and consequences for people living in England and Wales, whether defined by sex (women and men), gender identity, sexual orientation, ethnicity, disability, age, or parental status.

As a means of differentiating between types of DA, researchers have used the notion of severity, although they are attempting to assess this in a variety of ways. Most surveys use some form of act-based questions and scales, but how these are rated, combined, and analyzed will affect measures of severity. Severity may be assessed via the type of violence used, usually based on physical violence being rated as more severe, and/or using incidence to differentiate levels of severity (e.g., Statistics Canada 2004). The Conflict Tactics Scale (CTS) provides a classic example of this, rating physical violence, physical injuries sustained, and frequency of tactics used as indicators of severity (see Hamby 2016). However, the CTS provides no consideration of the context; for instance, there is no differentiation between a push and severe physical violence leading to hospitalization. Using physical injury as a key measure of severity also has particular limitations, as some of the most injurious violence involves women using weapons against their male partners in protection and/or self-defense (Johnson et al. 2014) and precludes the impacts of a wider range of potentially abusive behaviors. Qualitative interviews with female survivors have consistently shown that the psychological, emotional, and coercive behaviors they experience have a greater impact and are more harmful long-term than physical violence, as also reflected in the English offense of coercive control. Moreover, frequency or incidence does not in and of itself indicate severity. Frequency may indicate what Stark called regular “fights” (Stark 2007) rather than ongoing coercive control.

A number of studies have used some form of cluster analysis to disaggregate higher and lower levels of DA severity. In Johnson et al.’s (2014) reanalysis of the U.S. National Violence Against Women Survey (NVAWS), physical violence was assessed by responses to a 12-item version of the CTS, with respondents categorized as “nonviolent” if they reported “no” on all of the items and “violent” if they reported “yes” to one or more items (p. 193). Severity was assessed via a severe violence scale based on the items that they deemed to have been conventionally identified as severe violence. The Coercive Control Scale was constructed from a subset of nine survey items that dealt with nonviolent control tactics used by the respondent’s partner. Using cluster analysis involving the Coercive Control Scale with a threshold of five revealed two clusters: high control, which the authors equated with coercive controlling violence, and low control, which they equated with situational couple violence. Although confirming patterns in previous findings (Johnson 2008; Graham-Kevan and Archer 2003), the results were deemed more accurate, showing that 22% of women experienced high-severity coercive controlling violence from ex-husbands, whereas only 5.4% of men experienced such coercive controlling violence from their ex-wives. Lower-severity situational couple violence was perpetrated more equally by men and women (7.4% of ex-husbands, 3.9% of ex-wives).

Myhill (2015) reanalyzed the existing CSEW data using the small number of questions in the interpersonal violence module that may reflect the impact of coercively controlling behavior to provide a measure of the severity and typology of coercive controlling violence for intimate partners. Respondents were characterized as having experienced coercive control if they said their partner had both “repeatedly belittled you to the extent that you felt worthless” and “frightened you by threatening to hurt you or someone close to you” (p. 362), deemed to reflect abuse that was ongoing, denigrating, perceived as threatening, and had caused a degree of fear. By contrast, all other respondents who reported physical violence or acts of emotional or psychological abuse were classified as having experienced only situational violence. Myhill’s analysis involved people who had experienced only one abusive relationship since the age of 16 (n = 3544). Of these, about one in 20 men (6%, n = 52) who reported some form of domestic violence and abuse (DVA) were found to experience what could be termed coercive controlling violence, and nearly a third (30%, n = 791) of the DVA reported by women could be classified similarly (as compared to general prevalence figures from the 2008/2009 CSEW, which estimated 24% of women and 12% of men had experienced nonsexual DVA). Experience of coercive controlling violence was found to involve more severe and more frequent physical violence and was more likely to persist over time than situational violence. Using past and current relationship data, Myhill (2015) thus ends up with similar, albeit slightly higher, results regarding exposure to coercive control violence to those of Johnson et al. (2014) using past relationship data.

Using clinical rather than national prevalence data, the authors (Hester et al. 2010; Hester et al. 2017) have also developed and applied a measure to ascertain profiles of abuse linked to severity in individuals in same-sex relationships and heterosexual men (the COHSAR measure). The approach was developed from detailed knowledge of women’s experiences of DA from qualitative research as well as drawing on existing surveys. This work indicated that an extensive set of questions on both behaviors and impacts makes it possible to explore their intersection and thus to statistically differentiate between experiences that constitute coercive controlling violence on the one hand and those apparent DVA behavior experiences without the similarly harmful impact that coercive controlling violence and abuse involve on the other. Given that coercive controlling violence is a pattern of behavior over time, the approach assumes that where individuals experience higher levels of DA behavior from a partner, this may be associated with a greater impact on respondents. By combining statistically both DA behaviors and impact in the initial COHSAR study with men and women in same-sex relationships, this association was indeed found to be the case, reflected in the relationship between the frequency of incidents of DVA and their impact on respondents’ lives (Hester et al. 2010). Because the frequency of incidents and impacts of DVA can be assumed to be theoretically interdependent (coercive controlling violence involves behavior over time and harmful impacts), establishing the optimal threshold of the impacts and DA items was achieved by maximizing the statistical “fit” between these scales using one-way analysis of variance (ANOVA). Through ANOVA, the optimal fit between the two variables was determined, resulting in four distinct groups or profiles with high or low experience of DA behaviors reported and high or low impact. When the COHSAR approach was applied to 707 heterosexual men in a clinical sample, about half of the men, 371 (52.5%), reported experiencing at least one form of DA behavior during their lifetime. Of these, 31 men can be classified as experiencing high abuse and high impact. Given the general assumption that experiencing high abuse and high impact are commensurate with coercive controlling violence, the 31 men can thus be classified as experiencing such DVA, i.e., 4.4% of the entire sample of 707 men.

The current research, carried out between November 2021 and April 2022, aimed to develop a new question set and analytical approach for implementation in the CSEW that overcomes the limitations and problems identified in the existing set of domestic abuse and nature of violence questions identified above.

2. Materials and Methods

The research involved:

• development of a draft set of domestic abuse (DA) Questions (Draft 1), which were assessed via focus groups and interviews with 27 victims and survivors, leading to Draft 2.
• cognitive testing of the new Draft 2 DA Questions, with samples of the general public (n = 19) and victims and survivors (n = 23) to assess the suitability of the questions in relation to intimate partner and/or family abuse, and consultation with survey users.
• Development of a final Draft 3 DA Question set relating to intimate partner abuse and a Draft 4 relating to family abuse (for later testing by the ONS with large samples).
• Consultation with user groups and other stakeholders (arranged by ONS and not reported in detail here).

The Draft 1 questions with suggested changes to wording are listed in the body of the findings below. The subsequent draft question sets are detailed in Appendix B and Appendix C.

A range of ethical considerations were taken into account during the research, including safety and support from specific organizations for victims–survivors participating in the research and ensuring the security of online interviews and data. The research obtained ethical clearance from the National Statistician’s Data Ethics Advisory Committee and from the University of Bristol Research Ethics Committee. Informed consent was obtained from all subjects involved in this study. A Participant Information Sheet and a Consent Form meeting the GDPR requirements for transparency were provided to all potential participants ahead of the focus groups, interviews, and stakeholder consultations.

2.1. Developing Draft 1.0 Questions

To keep the length of any future iteration similar to the current CSEW (a funder requirement), we anticipated that a set of 20–30 questions could be utilized to cover prevalence, coercive control, and impact. We included a series of questions on physical, non-physical, and sexual abuse so as to cover physically abusive and coercively controlling behaviors and included possible measures of frequency (e.g., never, sometimes, often) to explore with participants.

To develop the Draft 1 questions, we drew on findings and recommendations from our previous research for the ONS, which mapped the content and structure of the existing DA modules, made comparisons with survey user needs, included extensive stakeholder input, and looked at issues from the wider literature and issues related to the Domestic Abuse Bill, especially definitional issues (Hester et al. 2023; ONS 2023). A number of considerations arising from previous work were taken into account as a result:

• How do we distinguish DA involving power imbalance as opposed to couple conflict?
• How do we design questions that are appropriate for and accurately reflect the experiences of both female and male victims–survivors?
• How can we differentiate victims of ‘one-off’ abuse that has little or no reported impact from those suffering ongoing, coercive abuse?
• How do we establish the three key facets of coercive control, i.e., intentionality of the abuser (versus motivation), negative perception of the controlling behavior by the victim (i.e., impact), and the ability of the abuser to obtain control through the deployment of a credible threat (Hamberger et al. 2017)?
• How do we identify victims for whom the abuse is mainly repeated physical assaults, victims who suffer primarily non-physical coercion, those who experience both physical and non-physical abuse, and those who also experience sexual victimization and post-separation stalking?
• How do we mitigate against the possibility that experiences which are not abusive are reported to victimization surveys? Or, indeed, that respondents may under-report experiences?
• How can we link a specific perpetrator to the abuse reported?
• Would ‘current partner’ provide an adequate proxy measure for abuse experienced in the ‘last 12 months’?

The resulting set of 25 Draft 1 questions for assessment with the 27 victim–survivors covered:

• Relationship status (do they have current partner? If not, route to questions about the ex-partner);
• Context of the relationship (who holds the power in the relationship);
• Coercive controlling behaviors (emotional abuse, isolation, jealousy, minimizing, denying and blaming, economic abuse, coercion, threats and intimidation, using privilege);
• Physical assault and use of weapons;
• Sexual coercion;
• Impact of non-physical and physical abuse.

We explored whether the Draft 1 questions would be suitable for capturing experiences of DA from partners/ex-partners and also from family members. We also took into consideration how analytical approaches might capture different victim ‘profiles’ more robustly than mere descriptive measures of prevalence. This included consideration of how questions might be asked to allow for scaling and to combine measures of behavior and impact to better reflect the diversity of experiences.

2.2. Focus Groups and Interviews to Assess Draft 1 with Victims–Survivors

To refine the question areas in Draft 1 (and help the later development of the Draft 2 DA question set to be tested via cognitive interviews), qualitative exploratory work was carried out with 27 victim-survivors (as experts by experience) via three online focus groups with 19 participants and eight online individual interviews. Focus groups were preferred as a method given the short time frame for the research, but participants were given a choice of focus group or individual interview. Recruitment of participants was facilitated with the help of project partners, who had links to a wide range of victim support services. We purposely sampled the participants for this stage to ensure coverage of individuals from different age groups, Black and Minoritized groups, different sex and gender groups, as well as those with and without children (see

Table 1. Focus groups & interview participants (n = 27).

		n	%
Sex/gender	Female	21	78
	Male	5	18
	Non-binary	1	4
Age *	16–29 yrs	7	26
	30–44 yrs	10	37
	45–59 yrs	7	26
	≥60 yrs	2	7
Sexual orientation	Heterosexual/straight	24	89
	Bisexual	2	7
	Lesbian	1	4
Ethnicity	White: English/Welsh/Scottish/Northern Irish/British	12	44
	White: other	1	4
	Mixed/multiple: White & Black Caribbean	1	4
	Mixed/multiple: Other	1	4
	Asian/Asian British: Indian	2	7
	Asian/Asian British: Pakistani	4	15
	Asian/Asian British: Bangladeshi	2	7
	Asian/Asian British: Chinese	1	4
	Asian/Asian British: Other	3	11
Children *	Have children	10	39
	Do not have children	16	61
Disability	Have physical/learning disability or chronic health issue	6	23 *

* excludes missing data n = 1.

), which was a requirement from the funders. The participants had experience with abuse from intimate partners, and some also had experience with family abuse. The focus groups and interviews were organized via the support services, who were also available to provide further support to participants if required.

The University Zoom account was used for the online focus groups and interviews, as this provides a secure facility with recording that is directly stored on the University’s encrypted server. To aid online participation, one member of the team shared slides of the questions we were interested in exploring with the participants and asked questions. Another team member made field notes and asked additional questions where appropriate.

We wanted to find out: whether the proposed draft questions on abusive behaviors and their impacts would adequately reflect lived experience of domestic abuse; whether it would be appropriate to ask the same questions about an ex-partner and/or family member as for a current partner; and whether key aspects of behavior and/or impact might still be missing.

The focus group and interview data were analyzed thematically, mapping feedback for individual questions to allow further development of the questions and question areas in Draft 2.

2.3. Developing Draft 2 Domestic Abuse Questions

Based on the findings from Stage 1, the research team, in consultation with the ONS, considered what were key question areas for inclusion in Draft 2, what questions may be omitted, and what changes needed to be made to clarify the meaning of the questions. A further important consideration was how to ascertain ongoing DA prevalence, whether to use abuse suffered in the last 12 months and/or by a current partner, asking about previous partner(s), and asking about family members.

This resulted in a Draft 2 question set of 19 questions. We decided to include experiences relating to a current partner and those over the last 12 months in the testing, along with asking about a previous partner.

From the Stage 1 findings, it became clear that nearly all the questions would apply to both intimate partner and family DA (wider family and related to ‘Honor-Based’ Violence). We therefore decided to use one version of Draft 2 in the cognitive testing, but we made sure that we asked participants about its applicability in both intimate partner and family DA contexts (Appendix C).

2.4. Cognitive Testing of Draft 2.0 with General Public and Victims-Survivors

In stage 2, we subjected the revised draft domestic abuse module emerging from the focus groups (Draft 2) to formal cognitive testing with the general public and with victims–survivors of domestic abuse. Walklate (2014) problematizes the victimization survey method itself, noting problems of respondent bias arising from recall and cognition issues in crime surveys. Differences in operational definitions and question wording between different types of studies certainly reflect their different aims and framing and can also strongly influence respondent behavior (see also Singer and Couper 2011). We decided to use cognitive interviewing, which is designed to identify cognition, recall, judgement, and response problems associated with survey questions as well as any sensitivity issues arising for respondents (see, e.g., Collins 2003; Schwarz 2007; Willis 2005). However, testing for measurement error in survey questions intended for online self-completion based on respondent verbal reports in face-to-face cognitive tests is problematic, especially in the context of sensitive question content on experiences of domestic abuse. To ensure coverage of non-DA and DA experiences, we therefore used both general public and victims–survivor samples.

A general public sample (GPS) with no experience of DA consisted of 16 previous CSEW participants and three from an LGBT+ network. A general public sample was initially recruited from CSEW survey respondents who had previously agreed to be recontacted by ONS and were selected on an opt-in basis. The sample was selected from previous CSEW respondents who did not report domestic abuse, although participants may have experienced sexual abuse (during the interview, we found that one had experienced family abuse). We were aiming for a sample with a range of demographic features based on age, gender identity, sexuality, ethnicity, household type (1 + children/none), disability, and age at completed full-time education (related to funder requirements). All of the CSEW respondents who finally took part in the research (n = 16) were heterosexual. In order to include the views of people from LGBT+ groups, we recruited three further participants via an LGBT+ network, specifying that we were interested in hearing from people who had not experienced DA.

A further 11 participants were ‘experts by experience’ i.e., people with lived experience of mainly intimate partner domestic abuse, recruited via a wide range of DA support services. To ensure adequate coverage of all forms of DA, another 12 participants with specific experience of family abuse were recruited via DA support services. See

Table 2. Cognitive interviews sample intimate partner and family abuse (n = 42).

		n	%
Sex/gender	Female	29	69
	Male	12	28
	Non-binary	1	2
Age *	16–29 yrs	5	12
	30–44 yrs	13	30
	45–59 yrs	12	29
	≥60 yrs	12	29
Sexual orientation	Heterosexual	37	88
	Bisexual	4	10
	Gay woman/lesbian	1	2
Ethnicity	White: English/Welsh/Scottish/Northern Irish/British	30	71
	White: European	1	2
	Mixed/multiple: Other	1	2
	Asian/Asian British: Indian	4	10
	Asian/Asian British: Pakistani	3	7
	Asian/Asian British: Other	2	5
	Black African	1	2
Children	Children in household	20	48
	No children	2	5
Disability *	Physical/learning disability/chronic health issue	12	29
Complete education **	≤16 yrs	6	21
	17–18 yrs	8	28
	19–22 yrs	10	34
	≥22 yrs	5	17

* excludes missing data n = 1, ** excludes missing data n = 13.

for participant details.

3. Results

3.1. Findings from Focus Groups and Interviews with Victims–Survivors Regarding Draft 1 Questions

In this section, the questions from Draft 1 used with the 27 participants in focus groups and interviews are detailed in italics alongside the findings related to the specific questions.

3.1.1. Framing the Questions

The pre-amble in the CSEW self-completion module provides an opportunity to set the context for the questions. We presented participants with the following preamble:

‘The next set of questions may seem very personal, but it is very important that we ask them to help the Home Office and other agencies understand more about domestic violence and abuse and how best to deal with it.’

We asked participants: “If somebody was answering this whilst experiencing abuse in a relationship, do you think it would make a difference if it was framed as a crime? Or would it put you off if it said domestic violence and abuse? Do you think it should mention the Home Office? If so why?”

The majority of participants felt it was better to frame the CSEW questions in the context of domestic violence and abuse (DA) rather than in a crime context. Their reasons related to their individual experiences of abuse. For example, some survivors who had experienced serious threats and severe physical and mental abuse and had become homeless following abuse preferred the word ‘crime’ because it reflected the seriousness of the behaviors they experienced. Others expressed more general concern about distinguishing between crime and domestic abuse. The mention of the Home Office (HO) in the pre-able to the questions was particularly problematic for survivors who had concerns about immigration status or knowledge of those who had such concerns.

3.1.2. Questions to Establish the Context/Rules of the Relationship

Based on a question used in the COHSAR survey (Hester et al. 2010), which aimed to establish the power dynamics within a relationship, we presented participants with the following example question:

• Q. Thinking about your relationship with your current partner/ex-partner/family member:

• Who makes all the big decisions? [usually you/usually partner/both equally]
• Who defines your role or duties within the relationship? [usually you/usually partner/both equally]

In general, participants were in favor of including a question to establish the power dynamics within a relationship, and many provided suggestions for improved wording to avoid potential confusion. For example, participants questioned what was meant by ‘big decisions’ as that could relate to financial decisions or decisions involving children, which may be decided consensually in non-abusive relationships. There was an overarching perception that ‘who has the final say on most decisions’ captured the issue of power imbalance to the greatest extent. The respondents indicated the need to differentiate between taking responsibility, decision making, and who has the final say.

3.1.3. Questions to Capture Coercive Controlling Behaviors

Overall, participants agreed that the coercive controlling behavior questions (below, 18 measures across 5 shorter questions to ensure only 2–5 measures under each) would generally capture their lived experiences of abuse, but they also provided a number of suggestions for what could or should be changed/added in terms of the specific wording (suggested wording underlined below for ease of reference).

• Q. How often has your current partner/ex-partner/family member: [Response options: Never/once/at least once a month/at least once a week/most days/everyday]

• deliberately deprived you of basic needs [‘necessities’ rather than needs] (e.g., food, sleep, medication or assistive aids [‘assistive technologies’ or ‘mobility aids’] (e.g., such as refusing to get your wheelchair fixed or taking away your hearing aids]
• deliberately prevented you from getting medical assistance (e.g., for injuries or illness) [or from accessing support services]
• tried to turn your family, friends or children against you or tried to convince them you were crazy [or stupid]
• blamed you for their violent behaviour/alcohol/drug misuse/self-harm
• told you that you’re imagining things, dismissed or minimised what you’re saying or twisted things
• controlled, or tried to control, your spending or access to [benefits/the household income
• refused to pay their share of rent/mortgage or household bills, child support payments, etc.
• stolen/taken money meant for the children (e.g., child benefit or tax credits, child’s savings or birthday money)
• stolen/taken your money, wages[/benefits], property or deliberately got you into debt
• prevented, or tried to prevent, [made it difficult for you] you from working or studying
• insulted, belittled, or humiliated you, either in public or in private
• commented on your physical appearance (looks, weight, clothes, hair) in a derogatory way or [dictated how you should look or dress]
• told you that you are a failure or not good/‘feminine’/‘manly’ enough
• told you no one else would want you
• monitored who you talk to or where you went, for example, by constantly phoning you when you are out of the house, or using GPS tracking, monitoring your texts, emails, letters, or Facebook/Twitter/other social media sites
• stopped you [or discouraged you], from seeing or talking to family, friends, colleagues, or children either by telling you not to, or making it difficult for you to do so
• acted possessive/jealous or unfairly accused you of being unfaithful
• dictated how you should look or dress [move to number 12 above]

Overall, participants thought that the coercive control questions were comprehensive enough to capture most of their experiences while also making suggestions for further improvement. For example, with regards to the question about deliberately depriving the respondent of basic needs, we needed to be clearer about what we mean by ‘assistive aids’ and give an example or two to capture the seriousness of how abuse can manifest amongst people with disabilities. A key message overall was that we must capture within these questions the notion that the abusers’ behaviors are used primarily to undermine the victim–survivor and their confidence.

In relation to capturing ‘gaslighting’ behaviors (where an abuser makes someone question their sanity, reality, or experiences), we might need to include ‘saying you did something when you did not. Another example of gaslighting behavior is when the abuser convinces the respondent/others that they are stupid.

We then asked participants the following question:

• Q. As a way of controlling (what) you (do), how often has your current partner/ex- partner/family member: [Never/once/at least once a month/at least once a week/most days/everyday]

• threatened to use something personal against you (e.g., your sexuality, physical or mental disability, illness/health condition, age, ethnicity, religion or immigration status)
• threatened to report you to the authorities [or ‘threatened to use the authorities’] (e.g., social services, family courts, immigration office, health/mental health services/GP, police, professional bodies) as a way of making you do something or stopping you from doing something?
• threatened to harm or kill themselves.

Participants suggested this question should use a yes/no response format, i.e., ‘has your partner…’ rather than ‘how often’, and also queried whether respondents would necessarily recognize that these actions are a way of controlling them.

• Q. How often has your current partner/ex-partner/family member: [Never/once/at least once a month/at least once a week/most days/everyday]

• threatened to kill you and you believed they were capable of doing so
• threatened to kill your children, pets or family members and you believed they were capable of doing so
• deliberately destroyed property or thrown/smashed things to intimidate or frighten you

Regarding threatening to kill you or others, participants suggested that it might also be better to answer yes/no to these questions rather than how many times (“is it necessary to know how many times? surely once is enough”). Participants also queried whether there was a need to include the ‘and you believed they were capable of doing so’.

Participants raised a number of further points about the questions on coercive controlling behaviors:

• The questions should reflect the escalation of abuse which could be achieved by re-ordering the questions/measures included above but also by asking an additional question to ascertain whether the behaviors have increased in frequency and/or severity over time.
• The questions should somehow capture the temporal aspect of abuse, i.e., the fact that abuse fluctuates.
• We need to reflect on the different ways that children are weaponized by the abuser to hurt the victim-survivor.
• We should add ‘deliberately prevented you from accessing services’ (e.g., including support services as well as medical/health services).
• We should extend the threats to family members and friends.
• It was not clear to some what we meant by the word ‘harm’.
• The questions are designed only for monogamous relationships and families.
• We need to consider third-party involvement in a relationship, for example, the incidence of control and abuse by other family members linked to the couple’s relationship (e.g., abuse and control by in-laws).
• The difficulty is that behaviors could be deemed positive rather than abusive, e.g., constant messaging/phoning, surveillance, or acting jealous, which can be perceived by a victim-survivor as an act of ‘care’ or ‘love’ rather than control.
• Consider adding a question on reproductive control.

Next, we asked participants whether we should also include the following question in addition to the question on threats to harm. We were interested in whether they thought respondents would actually admit to children or others being harmed by their abuser(s).

• Q. As a way of controlling (what) you (do), how often has your current/ex-partner/family member: [Never/once/at least once a month/at least once a week/most days/everyday]

• harmed, or tried to harm, your children, family members or pets
• harmed, or tried to harm, themselves

The general consensus was that if keeping this question in mind, it would be better phrased as a yes/no question (i.e., asking ‘has your partner…’ rather than ‘how often…’).

3.1.4. Questions on Sexual Coercion

Participants were asked about the following example questions on sexual coercion:

• Q. How often has a current/ex-partner/family member made you take part in any form of sexual activity when you did not want to, or you were unable to refuse?
• Q. How often have you agreed to sexual activity because you were afraid of what your current/ex-partner/family member might do if you refused?
• Response options:

• Never/once/2–5 times/6–10 times/more than 10 times; OR
• Never/once/at least once a month/at least once a week/most days/everyday

Should we frame the second question in terms of ‘fear’ of what their partner might do? Should we use words such as ‘afraid’ or ‘terrified’? Or are victims more likely to agree to sexual activity because it is less hassle or to appease them, i.e., as opposed to acting specifically out of fear? The consensus was that both ‘appease’ and ‘through fear’ should be included.

There was a consensus amongst participants that these questions needed to be included but should instead be asked, “Has your current/ex-partner/family member made you take part….” rather than “How often has”? which would elicit a yes/no response instead. It was suggested that asking ‘how often’ immediately assumes that this has happened, i.e., rather than giving the respondent permission to answer whether it has or has not happened to them. In general, participants favored the option ‘never/once/at least once a month/at least once a week/most days/every day’.

The consensus was that sexual coercion is about power. Both female and male participants said they had experienced pressure to have sex when they did not want to and did it to appease or because they were worried about negative consequences (e.g., being subjected to emotional or verbal abuse if they refused).

3.1.5. Frequency—Counting or Words, and What?

All participants agreed that measuring the frequency of abuse precisely is not feasible or appropriate, and it is better to use options such as ‘never, once, frequently or weekly, monthly’, etc. Most victim–survivors cannot remember or count how many times specific abusive acts happened to them.

To capture escalation within the frequency measures, some participants suggested adding ‘increased over time’. The ‘everyday’ category is especially important to include to capture the constant battle some victim–survivors of DVA are experiencing. Some also noted that there is a big gap between ‘once’ and then ‘once a month’.

3.1.6. Questions on Physical Assaults

We asked participants about the following questions on physical assault:

• Q. How often has a current/ex-partner/family member:

• attacked you in a way which caused a physical injury (e.g., cut, scratch, black eye, bruising, burn, fracture, broken bones, nose or teeth, internal injury or miscarriage)
• tried to strangle or suffocate you
• attacked you in a way which DID NOT cause a physical injury
• spat at you, pulled your hair, held you down or restrained you, or locked you in (or out) of a house/room
• used a knife, gun or other weapon against you

• Response options:

• Never/once/2–5 times/6–10 times/more than 10 times; or
• Never/once/at least once a month/at least once a week/most days/everyday

Participants suggested that the order of the questions could be improved to reflect the escalation of abuse. Some also queried the meaning of ‘attacked you in a way that did not cause a physical injury’, suggesting that this is not clear and needs clarification.

Some participants said they did not realize what the abuser was doing was an ‘attack’: “They keep repeat doing it..…pushing you and they say it was a mistake and not deliberate, then it escalates to pushing you downstairs and then you get injured”. Others noted that ‘fracture’ and ‘broken bone’ are the same thing and that ‘fracture’ be removed.

Participants suggested splitting the question ‘spat at you, pulled your hair, held you down or restrained you, or locked you in (or out) of a house/room’ into two questions. It was also pointed out that if you have a disability, an abuser may not have to physically restrain you, i.e., with their own hands, as they could restrain you in other ways, such as removing your disability aid or deliberately withholding essential medication, e.g., as a form of ‘chemical control’ (Walker et al. 2023).

Participants suggested that wording and language are important because the language that the perpetrator uses to explain or excuse their behavior becomes the narrative of the victim–survivor.

3.1.7. Questions on Impact of Coercive Controlling Behavior

We asked the participants whether the following questions adequately cover the different impacts of abuse or whether we might need to provide more examples of impacts that are not covered. Also, we highlighted that for questions 10 and 11 (below), we suspect that perpetrators might answer yes to these, e.g., if they were refused contact with children by the courts or were removed from the home as a result of a Domestic Violence Protection Order.

All the participants agreed with the range of impact examples listed below and found them to be very relevant. However, they all stressed that they only recognized these impacts once they had left the relationship. This was the result of their abusers’ coercive controlling behaviors, mentioned to a greater extent by the female participants. Some wording changes were suggested (underlined in what follows):

• Q. Thinking about your current partner, ex-partner, or family’s behaviours, did they affect you in any of the following ways? [tick all that apply]

• felt like you had to watch what you say or do because you were frightened of what they would do if you upset them. [e.g., watch out for mood swings, being in constant alert]

Some mentioned that this constant state of alert and feeling in constant battle 24/7 needs to be captured.

2.

feared for your life, or the life of your children, family members or pets.

Participants suggested these might have quite different impacts, with one participant highlighting that her ex-partner threatened her once, but it did not impact her as much as him threatening her children.

3.

felt like you had to work harder or significantly change your routine, behaviour, or appearance to stop them leaving or hurting you.

4.

anxiety or feeling unable to cope, sleep or concentrate. [e.g., nightmares, flashbacks, and avoidance]

Many wanted Post Traumatic Stress Disorder (PTSD) included as a term. However, as it is a specific mental health diagnosis, PTSD-like symptoms such as nightmares, flashbacks, and avoidance may be more suitable.

5.

loss of self-confidence/self-esteem.

6.

depression, suicidal thoughts, or self-harm

A female survivor suggested adding self-neglect by not seeking help when injured because of anxiety and depression. Another suggestion is that depression and suicidal thoughts should be separate.

7.

loss of trust in partner and/or others.

8.

given up work, education, hobbies or involvement in other activities (e.g., volunteering or religious activities).

Participants suggested that ‘made it difficult to’ should be added alongside ‘given up’ as a respondent might not recognize the intention behind the behavior even though the abuser is attempting to restrict their autonomy/freedom.

9.

Lost your job or feared losing your job.

10.

Lost your home or feared losing your home or becoming homeless.

11.

Lost your children or believed you would lose your children.

The impact of coercive controlling behaviors on victims–survivors post-separation needs to asked about explicitly.

3.1.8. Questions on Impact of Physical Assault

We asked the participants how well the question below captures the impact of physical domestic abuse and whether it makes sense to ask about the impacts of non-physical abuse separate from physical abuse—or if all impacts should be asked about at the end. Suggested changes to wording are underlined.

• Q. Thinking about your current partner, ex-partner, or family member’s behaviors, did they affect you in any of the following ways? [tick all that apply]

• Physical impacts (e.g., high blood pressure, pain, unwanted pregnancy, infection, or disease)
• started using drugs. Exacerbation of existing conditions
• Injuries that did not require medical attention (e.g., bruising/scratches/black eye)
• Injuries that did require medical attention (e.g., broken bones/teeth, internal or other injuries, miscarriage, burns)

Overall, participants felt the injuries could easily be recognized as impacts of physical attacks. However, they said they might not recognize that some of the health issues they have were caused by physical violence. There should therefore be both short-term and long-term impacts, and these should include non-physical effects.

3.1.9. Asking about Last 12 Months versus Current Partner

CSEW has previously only asked about the ‘last 12 months’ but not about a ‘current partner’. Our previous research indicated that individuals found it difficult to identify abuse in a current relationship and recommended that both ‘current partner’ and in the last 12 months’ required testing (Hester et al. 2023). This was indicated again in the current research in relation to coercive behaviors (see Section 3.1.7 above). Other prevalence surveys (e.g., FRA 2014; WHO 2005) also identify current partners. To assess these issues regarding intimate partner abuse, we asked participants if their experience with their current partner would provide a similar measure as asking for experience in the last 12 months. There was a range of views and an overall lack of clarity, indicating a need for further testing. Responses included “It is better to use current and ex-partner”, “12 months is arbitrary”, and “…people don’t say anything about current partner”.

3.2. Findings from Cognitive Testing of Draft 2.0 DA Questions

As outlined in the Methods, Draft 2 DA questions were developed following qualitative testing of the Draft 1 question set. Further testing using cognitive interviewing was conducted on the Draft 2 DA questions on coercive behaviors, impacts, and revised prevalence measures in relation to both intimate partner and family abuse. The same set of questions regarding behavior and possible impacts were tested for both intimate partner and family abuse, with an additional question regarding forced marriage for family abuse. Respondents were asked the questions in relation to partners (current or ex-) during the past 12 months or since age 16, and in relation to a range of possible family members (father, mother, brother, sister, son over 16, daughter over 16, male cousin, female cousin, uncle, aunt) in the last 12 months and since age 16. Draft 2 is in Appendix B.

3.2.1. Cognition

Overall, most of the Draft 2 questions raised very few issues in terms of cognition, with the majority of respondents suggesting the questions were easy to answer, straightforward and clear, and were able to give relevant examples where applicable and when prompted or during ‘thinking aloud’ tasks. However, the testing of the questions did reveal that Q14, which set out to measure change in frequency and severity of abuse, did raise some recall validity issues, signifying that this question should probably be removed from the module/question set. The feedback indicated that for a few of the questions, we should possibly revise the wording of particular items and possibly amend the examples given in order to make the question less ambiguous.

3.2.2. Sensitivity of Questions

Some of the behaviors asked about in the questions, e.g., comments on a partner’s clothes/appearance could be deemed to be or may seem innocuous to many, particularly those who have not experienced domestic abuse. However, those very same behaviors may also be experienced as part of a pattern of particularly abusive and controlling behavior by others. Thus, it is important that we are as specific and clear in our questions as possible.

The overall response from the sample was that these questions were not overly intrusive or uncomfortable. Victims–survivors in particular suggested they would be happy to answer these questions because they felt it was necessary to ask them as the questions described what they had been through. Therefore, an emerging message from respondents is that we must not worry too much about the sensitivity of the questions, as most respondents agreed that these questions need to be asked as they reflect the lived reality of victims–survivors.

3.2.3. Establishing the Context/Dynamics within the Relationship (Q4)

In relation to Draft 1, we considered the topic of context and dynamics within the relationship and looked at how to improve the delivery of such a question. However, the cognitive testing subsequently showed that the question does not work well.

3.2.4. Measuring Frequency

As in the focus groups and interviews regarding Draft 1 questions, in the cognitive interviews relating to Draft 2 victim–survivors, it was clear that it was not possible to ‘count’ individual acts of abuse. The cognitive testing indicated respondents preferred, for questions exploring non-physical coercion and control, a yes/no response as opposed to frequency scales. Many such behaviors are (implicitly) ongoing, and it is therefore preferable to use qualifying words such as ‘repeatedly’ where appropriate.

3.2.5. Measuring Escalation in Frequency and Severity of Abuse (Q14)

The question regarding escalation of abuse was found not to work as tested and would need more development and further testing or deletion.

3.2.6. Testing Recall: ‘Last 12 Months’ versus ‘Current Partner’

In relation to Draft 2, we tested recall, asking participants what time period they were thinking of when they were recalling the information to answer the questions (especially in relation to intimate partner abuse), i.e., were they thinking only about the last 12 months, recently, or about the whole of the time they were together/the whole relationship? Findings suggested that people who are still in an abusive relationship, female victims more so than male survivors, might not necessarily be able to answer the questions.

We asked about our current partner and ex-partner simultaneously. Some respondents were answering the questions about abuse experienced at the hands of their ex-partners. If they had more than one ex-partner, they would answer the questions with reference to whichever relationship was relevant. For example, one female victim-survivor said she was thinking about two abusive ex-partners when answering the questions, reporting the non-physical coercive and controlling behavior thinking about one ex-partner, and then answering the physical abuse questions thinking about the other ex-partner.

3.3. Establishing Abuse Profiles

We were able to carry out initial work to establish profiles using the general COHSAR approach (Hester et al. 2010) using the Draft 1 DA questionnaire responses regarding behavior and impact items. Four participants (one female, two male, and one non-binary) had provided details of their experiences (that also fitted the Draft 2 DA questions), and we mapped their responses to assess if they could provide scaled behavior and impact profiles to exemplify severity (i.e., high levels of behavior/high levels of impact and low behavior/low impact). This provided an initial indication that the DA questions on behavior and impact and the number of questions being asked would enable profiles to be measured using statistical methods involving general linear models (ANOVA, logistic regression).

3.4. Developing Drafts 3 and 4 Domestic Abuse

Based on the work from Drafts 1 and 2, including the cognitive testing, we were able to draft version 3 of the DA questionnaire relating to intimate partner abuse and Draft 4 relating to family abuse for further testing by ONS (Appendix C). The cognitive testing of Draft 2 indicated that the questions generally worked well for both intimate partner and family abuse and did not identify question response problems. The non-physical items in Draft 2 were found to be clear and well received, covering an accurate range of coercively controlling behaviors alongside the physical and sexual DA behaviors. The feedback was positive from users/stakeholders (in meetings arranged by ONS), as well as the general public and victim–survivor samples.

Only two questions (Q4 about relationship decisions and Q14 about escalation of abuse) were highlighted as either requiring further work or deletion. Neither of these questions feed into the DA profiles and were thus omitted from Drafts 3 and 4.

The impact items in Draft 2 were found to be clear and well received, covering an accurate range of physical and non-physical harms. The feedback was positive from users/stakeholders as well as the general public and victim–survivor samples, and the impact questions were especially commended. The impact questions were found to measure both short-term and longer-term impacts of DA.

The Draft 3 DA set of questions regarding current or previous partner behavior resulted in 24 questions, and Draft 4 regarding family behavior included the same 24 questions plus a further question asking about forced marriage. The core 24 questions cover relationship status, non-physical coercive control, sexual abuse, physical violence, and the impacts of such behaviors. They include one question about the frequency of physical violence experienced.

Following completion of our research, the ONS tasked Kantar Public (the CSEW survey contractors) to carry out usability testing of the new questions in Drafts 3 and 4 in February 2023. This involved face-to-face interviews with 10 individuals, 5 of whom had lived experiences with DA, and 5 members of the general public. Based on the testing, recommendations to improve the usability of the new questions included:

• Changes to the layout of the questions and providing more clarity about what the survey includes;
• Instructions about how to respond to questions and the reference period for questions;
• Some minor wording and format changes to the questions.

The ONS thus adopted both sets of questions with minor revisions, and at the time of writing, they are testing these in a split sample (ONS 2023).

4. Discussion and Conclusions

The research set out to develop and test a new set of questions on DA for the CSEW. It aimed to: improve the headline prevalence measure of domestic abuse, including frequency of abuse; develop a way of measuring controlling or coercive behavior within the overall prevalence measure; and develop a new measure of the impact of abuse. As detailed above, we have fulfilled these aims by developing and testing new questions that provide measures of controlling or coercive behavior and impacts. Moreover, we have recommended an analytical approach that will significantly improve the headline prevalence measure of DA and includes both coercive controlling behaviors and impacts. We have incorporated and tested a measure of the frequency of abuse.

The testing of Drafts 1 and 2 showed that the DA question areas we have developed provide a valid and comprehensive set of questions to measure the types of abuse research users want to know about, with an appropriate range of behavior and impact questions. Our findings also shed new light on the potential hazards of asking sensitive questions in DA surveys, such as respondent concerns about data disclosure risks and intrusiveness. For both victims–survivor and general public participants in this study, question content itself was not felt to be overly intrusive, but mode of delivery is vital, with self-completion being overwhelmingly preferred by participants, a finding consistent with wider scholarship on these issues (e.g., Tourangeau and Yan 2007).

We were able to incorporate into the development of Draft 1 and following Drafts the issues arising from our previous research, as listed in Section 2.1 above, and consider these further below. Survivors across the whole range of participants, including men and women from minoritized communities, etc., were especially complementary regarding the extent to which the questions echoed their experiences. The DA questions can be used to measure intimate partner DA and were also found to be appropriate for family DA (wider family and ‘honor’-based abuse). The questions on coercive control include the key facets of the intentionality of the perpetrator via deliberate acts, impacts, and threats (Hamberger et al. 2017). We have also tested whether questions may elicit non-abusive experiences or under-report experiences (for instance, regarding current partners).

Capturing the frequency of physical violence is an important aspect of DA and also features in the offense of coercive control. Due to the strong view from victims–survivors in the current research that it is not possible to accurately count discrete acts of violence, we recommend a banded approach that captures whether respondents have experienced a single or repeated act of physical violence. This will also enable ‘one-off’ and infrequent acts of physical abuse that may constitute an important abuse profile associated with situational violence to be identified (see Johnson 2008). Other questions ask about non-physical coercion, sexual victimization, and post-separation stalking, thus allowing identification of victims for whom one or particular forms of abuse are relevant.

As identified in previous studies (Hester et al. 2023; Nybergh et al. 2013; Johnson et al. 2014), asking about domestic abuse during ‘the last 12 months’ presents a challenge in that people who are still living in an abusive relationship will often not recognize or articulate what they are experiencing as abuse (and thus not answer positively questions in a survey that specifies it is about domestic abuse). In the current research, a strong message from female victim–survivors in particular was that it is often not possible to recognize and name their experiences while in the maelstrom of an abusive relationship. They also felt uncomfortable and constrained when individual questions were framed over an arbitrary time period of 12 months. We have identified a method for the CSEW that enables victims to better report a wider range of abuse occurring in both current and previous intimate relationships while retaining the capacity to estimate a 12-month prevalence rate and thus retain the existing time series (Draft 3 Appendix D), and thus does not require ‘current partner’ to be a proxy measure for ‘the last 12 months’. We note, however, that where difficult choices need to be made, we believe it is fundamentally right to prioritize accurate representation of the lived experiences of victims–survivors over essentially arbitrary reporting timeframes.

As is the case for most DA surveys, headline prevalence in the CSEW has been presented as a descriptive statistic using the frequency of physical abuse and the frequencies of other behaviors linked to demographic information. However, as outlined earlier, this may result in misidentification of DA with respect to severity and a lack of attention to differences in experience, for instance, between men and women (Sanchez-Prada et al. 2023; Johnson 2008). To obtain a more nuanced and realistic picture of DA that can be more useful to policy makers and support agencies, we propose that the ‘COHSAR’ analytical approach be applied to the DA question data to identify abuse profiles and provide headline prevalence for coercively controlling violence and abuse. Our approach emphasizes the importance of recognizing the wide-ranging impacts of abusive behavior and drawing on this information in estimating the prevalence of DA to identify those more likely to require agency intervention. As outlined earlier, it is possible to construct reliable indices of DA behavior and impact by aggregating question items, and our existing work has shown high levels of scale and test-retest reliability for indices of abusive behavior and impacts (Hester et al. 2010, 2017). Given this, we can then model the empirical relationship between abuse and impacts to identify an optimal threshold for distinguishing between ‘high’ (high abuse/high impact) and ‘low’ (low abuse/low impact) DA groups. General linear models (ANOVA, logistic regression) offer a rigorous framework for identifying such a threshold by maximizing between-group and minimizing within-group differences. Our initial profiling work in the current research suggests such an approach would be viable using the new CSEW questions. We suggest the high/high profile should be used to provide a headline prevalence measure. It is then possible using standard inferential methods (e.g., chi square, logistic, and multinomial regression) to examine the distribution of these DA profiles across the CSEW sample on the basis of socio-demographic characteristics and different abusive behavior profiles, thereby attaining a more nuanced picture of the particular groups to be targeted for services. We suggest the two profiles of high/high and low/low be published to provide comparisons between those in need of services and support and those who are much less likely to need such support. It should be noted that the ONS has agreed to test this approach with a larger and split sample (ONS 2023). This approach would take the estimation of headline prevalence significantly beyond the use of descriptive statistics and frequencies used in other victimization surveys.

Linking the experience of DA to a specific perpetrator helps us to understand ‘who does what to whom’ and thus feeds into the new UK government policy on DA, which includes work to tackle perpetration. Such an approach is generally lacking in victimization surveys, including the CSEW, and our previous research found it was a difficult issue to resolve (Hester et al. 2023). In the current work, we have included questions about perpetrators in Drafts 3 and 4 that will allow some general conclusions to be made regarding ‘who does what to whom’. Questions Q1 to Q5 have been added to Draft 3 to identify the sex of current and previous partners/perpetrators (see Appendix C). We have also added questions along similar lines with regard to family DA, using categories such as father-in-law, mother-in-law, uncle, aunt, male cousin, and female cousin to denote perpetrator and perpetrator sex (see Draft 4 Appendix C). It should be noted that the ONS has adopted this approach for testing (ONS 2023). However, we recognize that it will remain difficult to disaggregate and identify specific perpetrators where individuals had more than one partner or more than one perpetrator. This is a limitation that may need revisiting depending on the size of this group, as identified in further ONS testing.

Finally, the dynamics and context of intimate relationships have been another key area that others have argued is important to understand in DA. Sanchez-Prada and colleagues, for instance, urge that surveys “provide an adequate contextualization of the behaviors under study, such as questions that can evaluate the positions of power within the relationship” (Sanchez-Prada et al. 2023, p. 1044). We asked participants if we should include a question that will allow the survey to capture where the power lies in terms of decision making, negotiation between partners, whether somebody is allowed to voice their opinion, has freedom of personal choice, or possibly pick up on the presence of traditional gender roles or entitlement. This was an aspect that did not work, and the limitation remains.