Evaluating a new measure of domestic abuse

The way domestic abuse manifests is constantly changing. Considering this and legislative changes, we have undertaken a user engagement, research, and testing programme to improve the collection of Crime Survey for England and Wales (CSEW) data on domestic abuse.

This article concludes our evaluation, providing the evidence behind our decision to change to using new survey questions in the CSEW from April 2025.

The questions on domestic abuse asked as part of the self-completion section of the CSEW have largely remained the same since they were introduced in April 2004, allowing for a long, comparable time series. However, there were issues with the data, mainly: 

• they did not align with the definition of domestic abuse introduced in the Domestic Abuse Act (DAA) 2021

• they excluded the offence of controlling or coercive behaviour, which was introduced in the Serious Crime Act 2015

• they did not measure the number of incidents or frequency of abuse 

• there was a greater user need for data to understand the nature of the abuse 

The new questions provide opportunities to explore and present the lived experiences of victims of domestic abuse to aid government, law enforcement, victim services, academia and the general public in better supporting victims, and ultimately to help prevent domestic abuse.

This article provides the findings for each evaluation criterion in Section 8 of our Developing a new measure of domestic abuse methodology article, and for the associated research questions outlined in Section 4 of our Redevelopment of domestic abuse statistics: research update, published in November 2023.

We aim to publish headline estimates from the new questions alongside our Crime in England and Wales: year ending March 2025 bulletin in July, and more detailed estimates in our annual domestic abuse compendium, which will be published later in 2025. We will work with users to provide the support needed to understand the new estimates.

In addition to this article, we have published a summary in our Redevelopment of domestic abuse statistics: research update May 2025 article

2.1 Crime Survey for England and Wales

The Crime Survey for England and Wales (CSEW), previously known as the British Crime Survey (BCS), is a face-to-face victimisation survey for people aged 16 years and over. People resident in households in England and Wales are asked about their experiences of a range of crimes. The main aim of the CSEW is to provide robust trends for the crime types and population it covers; the survey does not aim to provide an absolute count of crime and has notable exclusions.

The CSEW has a complex structure consisting of a core set of modules asked of the whole sample, a set of modules asked only of different sub-samples, and self-completion modules.

Self-completion modules are used in the CSEW to collect information on topic areas, such as domestic abuse, that respondents could feel uncomfortable talking about to an interviewer. The use of self-completion on tablets allows respondents to feel more at ease when answering questions on sensitive issues, because of increased confidence in the privacy and confidentiality of the survey.

The self-completion modules include questions on:

• drug use and drinking

• gangs and personal security

• domestic abuse

• sexual assault

• stalking

Historically, we have also rotated between modules covering the nature of partner abuse, the nature of serious sexual assault and the experience of abuse during childhood.

More information about the CSEW can be found in Section 2 of our User guide to crime statistics for England and Wales: March 2024.

2.2 Domestic abuse in the Crime Survey for England and Wales

A domestic abuse module was first introduced in the CSEW in 1996. It was permanently included from April 2004. The survey questions have remained largely unchanged since 2004, enabling a long, comparable time series. The two exceptions to this are:

• a small redesign in 2013 (for more information see our Intimate personal violence methodological note (PDF, 497KB))

• the suspension of the face-to-face CSEW in the year ending March 2021 because of the coronavirus (COVID-19) pandemic

The age range of those asked the domestic abuse questions has changed over the years. Originally, those aged 16 to 59 years were asked the questions. The upper age limit was increased to 74 years in April 2017 and then removed in October 2021, as explained in Section 2 of our User guide to crime statistics. The comparable time series dating back to 2004 is for the 16 to 59 years age group and does not include older victims of domestic abuse.

When the offence of controlling or coercive behaviour in an intimate or family relationship was introduced as part of the Serious Crime Act 2015, new questions were developed and evaluated through the Office for National Statistics (ONS) Domestic Abuse Statistics Steering Group (DASSG). Between April 2017 and March 2019, the newly designed questions were added to the existing domestic abuse questions on the CSEW on a split-sample basis.

Analysis of the new questions, as outlined in our Developing a measure of controlling or coercive behaviour, April 2019 article, concluded that there was uncertainty in whether the measure adequately captured victims of the offence. Working closely with topic experts, it was decided that the questions required further development and testing before we could be confident that they were fit for purpose. As such, the questions tested in the split-sample experiment were removed from the survey in April 2019. This work became the foundation for this redevelopment project.

2.3 Scope of the domestic abuse question redevelopment

Work undertaken to redevelop the domestic abuse survey questions has focused solely on the main domestic abuse module on the self-completion part of the CSEW. The data collected through this module are used to derive the headline prevalence measures. Redeveloping the module on the nature of partner abuse was not part of this work, however, this is something we aim to do in the future.

Changes to the wider CSEW were also not within the scope of this redevelopment project. For example, although some of our early work, outlined in Section 3 of our Redevelopment of domestic abuse statistics: research update November 2021 article, explored mode of response, this project's aim was not to change the response mode of the self-completion modules. In addition, the CSEW is a victimisation survey of people aged 16 years and over. It also does not ask people whether they have committed a crime or intended to cause harm. This work did not consider changing this approach and this has resulted in some limitations when working towards existing domestic abuse definitions.

It was also important to consider the length of the survey and the respondent burden associated with increases in survey length. For the year ending March 2023 CSEW, our CSEW technical report (PDF, 1.7MB) stated that the average interview length for victims of crime was 73 minutes compared with 47 minutes for non-victims. This showed that changes to the domestic abuse questions would have consequences across the entire survey and it was therefore important to be mindful of the impact of including additional questions.

The last time we added new domestic abuse questions to the CSEW, we attempted to do so in a way that wouldn't affect the current measures. Given this was unsuccessful, this project did not rule out making more significant changes which could affect the comparability of the estimates and therefore the continuation of the existing time series. Following the first stage of research, as highlighted in our Redevelopment of domestic abuse statistics: research update November 2021 article, most users felt that improving the headline measure was more important than continuing the current time series, given the substantial issues with the data currently collected.

2.4 New domestic abuse questions

Following extensive testing, as outlined in our Developing a new measure of domestic abuse: April 2023 article, we produced a set of new domestic abuse questions (PDF, 471KB). These questions were added to the CSEW on a split-sample trial from April 2023 until March 2025, with half of respondents answering the new questions and the other half answering the existing questions.

The new domestic abuse questions are split into two parts. The first part asks questions on intimate partner abuse, while the second part asks questions on family abuse.

The questions first collect information on experiences since the age of 16 years. These questions aim to capture data on experiences over the course of the respondent's life, from the age of consent. If the respondent states they have experienced any of the abusive behaviour questions since the age of 16 years, they are asked if they experienced any of those behaviours in the last year (12 months). Respondents are then asked questions on the impacts of those behaviours since the age of 16 years. We do not ask questions on impacts experienced in the last year because of limited survey space and the cognitive burden associated with asking whether the impacts they felt or experienced happened both in the last year and since the age of 16 years.

Throughout this release, we refer to questions about abusive behaviours and impacts since the age of 16 years, and in the last year (12 months) as “since the age of 16 years questions” and “in the last year questions”, respectively. We also refer to the corresponding time periods as the “since the age of 16 years time period” and the “in the last year time period”.

Within the intimate partner questions, respondents are asked about abusive behaviours experienced by their current partner, ex-partner(s), or both. At the beginning of the module, we ask if respondents have a current partner or ex-partner(s); "are you currently in a relationship with an intimate partner, for example a boyfriend, girlfriend, husband, wife or civil partner?" and "since you were 16, have you had a previous intimate partner or partners?". Depending on the answers, respondents are presented with abusive behaviours and asked if a current or ex-partner was the perpetrator.

Within the family abuse questions, respondents are asked about abusive behaviours they have experienced from one or more family members. This includes father, mother, son, daughter, brother, sister, or any other family member who was aged 16 years and over (for example grandparent, grandchild, cousin, uncle, or aunt). It also includes family members who are step, adopted, foster, or in-laws.

In April 2023, our Developing a new measure of domestic abuse methodology article outlined our strategy for evaluating the success of the new questions and how we would make a decision on the future of domestic abuse measurement in the Crime Survey for England and Wales (CSEW). Setting this strategy before collecting any data from the new questions was essential to ensure the transparency of our approach.

We developed a set of evaluation criteria in line with the quality requirements of the UK Code of Practice for Statistics and Eurostat's European Statistics Code of Practice, which sets out five dimensions for measuring the quality of statistical outputs. The dimensions are:

• relevance

• accuracy and reliability

• timeliness and punctuality

• comparability and coherence

• accessibility and clarity

We published a set of 12 evaluation criteria, as detailed in Section 8 of our Developing a new measure of domestic abuse: April 2023 article. To expand further on the evaluation criteria, we published details of 29 research questions we would use to assess the criteria in Section 4 of our Redevelopment of domestic abuse statistics: research update November 2023 article. For example, within the relevance quality dimension there are four evaluation criteria and 10 associated research questions.

The assessment of the evaluation criteria required a range of sources, including, but not limited to, data collected from the split-sample trial, our research work and user engagement. The sources were used to answer the research questions linked to each criterion. This allowed us to evaluate the success of the new domestic abuse survey questions.

The remaining sections of this report cover each of the five quality dimensions and the subsections show the associated evaluation criteria. Information is presented against each to provide evidence as to whether the criterion has been met. We used this information to make the decision to stop the split-sample trial and include the new domestic abuse questions on the CSEW permanently from April 2025, replacing the existing questions. We could not continue with the split-sample trial because of the impact on the quality of the domestic abuse estimates, as well as the sexual assault and stalking estimates, because all would continue to be affected by reduced sample sizes.

Relevance is the degree to which statistics meet the current and potential needs of users. For more information, see the Government Analysis Function's Quality statistics in government article. This section evaluates the four criteria we set in the context of the relevance quality dimension.

4.1 The new questions and proposed outputs meet the priority requirements of users

Ensuring that our statistics continue to provide the most accurate information and meet the needs of users is central to the redevelopment of the domestic abuse survey questions. User engagement has been a core part of the redevelopment work, and this evaluation criterion captures the importance of meeting our users' requirements.

To assess this criterion, we considered how the requirements of users were gathered, what those requirements were, and whether they have been met. We recognise that meeting all user requirements is likely to be an unachievable target, therefore the focus is on several priority requirements.

4.1.1 User engagement

Since the beginning of the redevelopment work in 2020, we have sought continuous involvement and feedback from a wide range of users to ensure their needs and requirements were considered. As well as collecting requirements, we have also gathered feedback on the new survey questions, and our outputs. At each stage of the redevelopment process, user engagement has taken place through a variety of methods, including user surveys, one-to-one meetings, and open events. User engagement has also included representatives from a variety of organisations across support services and charities, the academic sector and other government departments.

A summary of the main user engagement activities we undertook between 2020 and 2022 is given in Section 5 of our Developing a new measure of domestic abuse: April 2023 article. In this section, we have outlined our user engagement for 2023 onwards.

Office for Statistics Regulation review, April 2023

Alongside the publication of the new survey questions and our strategy for evaluating their success, the Office for Statistics Regulation (OSR) conducted an independent assessment of our user engagement, as well as our process for developing new domestic abuse questions. The result of the assessment was a letter from OSR endorsing the approach we took towards developing new questions. They also stated that "the team carried out extensive engagement with a variety of different stakeholders at multiple points throughout the redevelopment process".

User feedback, April 2024

In meetings with users, we explained the structure of the new questions and discussed methodologies for defining a victim. We asked for feedback and received suggestions for alternative methods and ideas for further analysis. Overall, the feedback was positive, with users content with the direction of the research towards including the new questions.

User event, May 2024

We hosted our third user event on the domestic abuse redevelopment in 2024. The event was open to anyone with an interest in statistics on domestic abuse. It was publicised using Eventbrite and by email to stakeholders who had previously engaged with our work or expressed interest in the domestic abuse data. We provided an update on the redevelopment work, outlined the structure of the new survey questions, presented preliminary analysis, explained how we would evaluate the new questions and shared our future plans. Users were given the opportunity to ask questions, however, provided little feedback regarding the evaluation strategy, trial outputs and future timelines.

User survey, winter 2024 to 2025 

Alongside our Redevelopment of domestic abuse statistics: research update December 2024 article, we published a user survey. The survey was publicised through the article and by email to stakeholders. We received 14 responses. We asked for feedback on:

• the content of the new questions

• the new estimates

• the methodology used to produce the new estimates

• supporting information around the estimates and the trial outputs

In addition to these engagement activities, we have also met with many users on a one-to-one basis to discuss their requirements in more detail and answer any questions.

4.1.2 Priority requirements

In 2020, we ran a user survey to understand how Crime Survey for England and Wales (CSEW) domestic abuse data were currently used, and to understand future requirements for domestic abuse statistics. The University of Bristol's report summarised that the main findings from this survey were:

• the majority of respondents stated that improving the measure of domestic abuse was more important than continuing a comparable time series

• respondents expressed substantial support for continuing to ask about experiences of domestic abuse from the age of 16 years

• respondents favoured producing separate estimates of abuse by a family member and by a partner or ex-partner

• respondents ranked prevalence the highest in terms of topic importance, followed by impact and then nature; however, several respondents noted that ranking these against each other was difficult, as they are equally important in understanding domestic abuse

• some respondents felt the connected nature of types of domestic abuse meant the linking of abuse to specific perpetrators was important

These findings were central to the development of the new survey questions and were collated alongside additional requirements received throughout the project. A total of 50 unique requirements were collected through the various forms of communication and user engagement outlined previously.

Discussions with a core group of users in February 2021 focused on narrowing down the requirements to identify themes. Users were sampled purposively to provide a mix of policy makers, third sector organisations and academics. Further information on the group's members and discussions that took place are outlined in the University of Bristol's report.

This information, alongside that collected in the user engagement events and follow-up conversations with individual users, was used in part of the research led by the University of Bristol. The outcome of this research was a set of six priority requirements and recommendations for the new domestic abuse statistics. In priority order, these were:

• that the headline prevalence measures of abuse should be revised and improved

• that the survey should measure coercive control, following robust question development and testing

• that impact should be measured

• that the measurement of physical assault should be disaggregated

• that further consideration should be given to measuring the frequency of physical (and sexual) victimisation

• that consideration should be given to matching abuse to a specific perpetrator

Further information on each of these is available in Section 3 of our Redevelopment of domestic abuse statistics: research update November 2021 article.

4.1.3 Meeting requirements

User feedback has been gathered throughout the various stages of the project to assess whether the new survey questions and proposed outputs met their needs. For example, at our user event in 2022, we asked users several questions including:

"How confident are you that the work we've done to redevelop the questions will address the issues with the existing questions?"

"Are the data collected from the new set of questions likely to meet your needs?"

The feedback received was positive, with 84.6% of the 26 users who answered the poll saying they were "somewhat confident" or "very confident" that the work we had done would address the issues with the existing questions. Over three-quarters (78.6%) said the data that would be collected from the new questions would meet or mostly meet their needs.

Our user survey conducted in winter 2024 to 2025, following the release of our data from the split-sample trial, provided additional evidence showing users supported the move to the new questions. Around three-quarters (76.9%) of respondents agreed that the new survey questions should be added to the CSEW permanently.

Of the 50 unique user requirements gathered throughout the redevelopment project, our assessment showed that 70.0% of these would be fully or partially met through the new survey questions. We recognise that meeting all user requirements is an unachievable target. This is largely because of the limitations of a victimisation survey, including limited survey space and the appropriateness of questions or topics for the survey mode. Focusing specifically on the smaller number of priority requirements recommended by the University of Bristol research, our assessment showed the following results.

Improving the headline prevalence measures

The new survey questions ensure that the requirement for prevalence measures for both since the age of 16 years and in the last 12 months, by a partner and family member, can be produced. Following thorough question testing, the new measures will more accurately measure victims' lived experiences than the existing questions (see Subsection 5.2 in Section 5: Accuracy and reliability), providing greater detail through the more comprehensive list of abuse types collected. In addition, the headline prevalence measures will align with the definition of domestic abuse outlined in the Domestic Abuse Act (DAA) 2021 see Subsection 7.2 in Section 7: Comparability and coherence.

Measuring coercive control 

As controlling or coercive behaviour is the only domestic abuse-specific criminal offence, it was highlighted by users to be a priority when developing new survey questions. It is a complex offence, recognised to be at the core of domestic abuse, and there is currently no agreed measurement instrument internationally.

The questions we have developed will measure many elements of the offence within the abusive behaviours asked. However, our research found that we cannot collect all the information needed to determine whether the offence has occurred through a quantitative self-completion survey with limited question space. Therefore, the headline domestic abuse prevalence measures we will produce will not fully align with the controlling or coercive behaviour offence definition. Our work to develop abuse profiles will provide a way of differentiating "types" of victims, without producing a singular measure of controlling or coercive behaviour. For more information on the offence and how the new questions align to the offence of controlling or coercive behaviour, see Subsection 7.3 in Section 7: Comparability and coherence.

Measuring impact 

The new survey questions capture abusive behaviours experienced. They also capture a wide range of potential impacts experienced as a result of those behaviours since the age of 16 years. Users felt that having data on impact was important as a means of determining severity and harm. The data on impact will be used in the identification of different abuse profiles (see Subsection 4.4 in Section 4: Relevance for more information).

Producing a disaggregated measure of physical assault 

Non-fatal strangulation, which is now a criminal offence, was not specifically asked about in the existing questions. The new questions ask the respondent whether someone "used force on you or attacked you". Follow-up questions then ask respondents about specific types of assault. As shown in the data from the split-sample trial, published in December 2024, data can be disaggregated to allow the publication of estimates on physical abuse overall, as well as more specifically on responses to the following behaviours:

• "physically held you down or restrained you"

• "strangled, choked or suffocated you"

• "used a knife, gun, or any other object as weapon against you"

Measuring frequency of physical and sexual victimisation 

As described in the University of Bristol's report, findings from cognitive testing with victims showed that they felt it was not possible for them to count individual acts of abuse. This was because of issues with recall and the appropriateness of such questions. For questions exploring non-physical coercion and control, a "yes" or "no" response was preferred, as opposed to frequency scales. These behaviours can be ongoing, and it was suggested that qualifying words such as "repeatedly" should instead be used, where appropriate. For acts of physical violence only, victims felt that measuring frequency may be possible. However, a banded approach to measuring frequency (for example, "once" or "two to five times") was preferred. As such, this requirement has been partially met, with the new questions capturing some information on the frequency of physical victimisation.

Matching abuse to a specific perpetrator 

Although not possible with the existing questions, being able to link victims, and the type of abuse they experienced, to their relationship with the perpetrator or perpetrators was something users felt was an important requirement in aiding interpretation of the data. Respondent safety and well-being is our top priority when asking questions on such sensitive topics, therefore we decided not to include questions on numbers of perpetrators (see our Redevelopment of domestic abuse statistics: research update November 2021 article for more information). However, the new questions will allow for a more detailed exploration of whether specific abusive behaviours were perpetrated by a current partner, an ex-partner or a family member, than the existing questions. The sex of the perpetrator will also be known, and for family members it will be possible to determine the relationship of the perpetrator to the victim (for example mother, uncle, cousin).

In summary, of the priority requirements recommended by the University of Bristol research, four have been fully met, with the other two partially met. Of the 50 unique requirements, 30.0% have not been met. Topic area requirements not met include understanding first-time victims of abuse, information on children witnessing abuse, and keeping the current time series.

The question redevelopment has aimed to balance meeting as many, often competing, requirements as possible. However, it has not been possible to meet every requirement. Constraints such as survey length, cognitive burden on respondents, interpretation of questions and the quality of the data have all contributed to our decisions about which questions to include on the survey. Our user engagement throughout the question redevelopment has ensured that requirements were documented, assessed and prioritised, and feedback continuously sought to ensure that these were being met. Our evidence shows that user feedback about the move to the new questions is positive, and that the proposed outputs will meet the majority of user requirements.

4.2 Data collected from the new questions can produce prevalence estimates and breakdowns at the same level of detail as is currently produced

One of the main aims of the redevelopment of domestic abuse survey questions was to produce a set of questions that could be used to produce prevalence estimates. Under this evaluation criterion we outline:

• the existing and new question designs

• our methodology for producing prevalence estimates from the new questions

• the decision-making process we have gone through to arrive at this methodology

• the resulting estimates and breakdowns available

4.2.1 Existing question design

Before outlining the method for producing estimates from the new domestic abuse questions, it is important to explain the method applied to the existing questions.

Using the existing CSEW domestic abuse questions, we produce prevalence measures for the since the age of 16 years and in the last year time periods. This provides a lifetime measure and a more recent measure.

The more recent measure tends to vary more year on year and over time, whereas the lifetime measure remains relatively stable, and year-on-year changes are largely a result of sampling variation. More information can be found in our Domestic abuse prevalence and characteristics data tables.

The existing domestic abuse estimates are derived from 16 questions. Eight questions ask respondents if they have experienced abuse from a partner. The remaining eight questions ask if they have experienced abuse from a family member. For each relationship type, there are four questions that cover the time period since the respondent was aged 16 years and four questions that cover the respondent's experiences over the last year. A respondent is only asked questions on experiences in the last year if they have experienced any of the abuse specified in the questions that ask about experiences since the age of 16 years. There are a further two questions asking the relationship between victim and perpetrator in the family abuse questions. 

A further 48 questions ask respondents if they have experienced each of the abusive behaviours since the age of 16 years and in the last year:

• by a partner

• by a family member

• by both a partner and a family member

Of these questions, 24 are in the sexual assault module and 24 are in the stalking module.

Respondents are asked in the last year questions if they responded "yes" to the since the age of 16 years questions.

Using the existing questions, we produce estimates for the since the age of 16 years and in the last year time periods, for the following abusive behaviour types, with relationship in brackets:

• overall domestic abuse

• partner abuse

• family abuse

• non-physical abuse (partner or family)

• threats (partner or family)

• physical abuse (partner or family)

• sexual assault by rape or penetration (partner or family)

• indecent exposure (partner or family)

• unwanted sexual touching (partner or family)

• domestic stalking (partner or family)

The abuse types outlined can be broken down by demographic characteristics subject to sample sizes and statistical disclosure control constraints.

4.2.2 New question design

The new domestic abuse questions were trialled in the same location on the CSEW self-completion module as the existing domestic abuse questions. The questions were tested in a split-sample trial, meaning that half of the respondents received the existing questions, and the other half received the new questions. Limited space on the survey meant that respondents to the new questions did not receive the existing sexual assault and stalking questions that traditionally sit on the self-completion module.

The new questions have been designed to ensure estimates can be produced for the since the age of 16 years and in the last year time periods. These measures were published as part of the trial outputs accompanying our Redevelopment of domestic abuse statistics: research update December 2024 article.

The questions are split into a partner section and then a family section. Within each section, respondents are asked about abusive behaviours and then the impact(s) of those behaviours. Unlike the existing questions, respondents need to provide a response to each behaviour and impact. In the partner section, respondents can state whether the behaviours they experienced were perpetrated by a current partner, ex-partner(s), or both.

Respondents are presented with questions asking whether they have experienced abusive behaviours since the age of 16 years. If they say that they have, they are then presented with questions asking whether they have experienced those behaviours in the last year. Then, if a respondent has experienced any abusive behaviours, they are presented with a set of questions asking about the impacts of the abuse they experienced since the age of 16 years. We do not ask about the impacts of abuse experienced in the last year.

The way the new questions are asked and presented is very different to the existing questions. The new questions are presented in a grid-style format allowing respondents to select multiple abusive behaviours and impacts where applicable (see Subsection 7.1 in Section 7: Comparability and coherence for an example question). Respondents are also asked whether a current partner, ex-partner(s), or both, were the perpetrator of those behaviours within the grid for the partner section. The grid format continues in the impact sections of the question set, where a respondent can select multiple impacts.

The abusive behaviour types and relationships (shown in brackets) captured by the new questions are:

• overall domestic abuse

• partner abuse (partner, current partner or ex-partner)

• family abuse

• emotional abuse (partner, current partner, ex-partner or family)

• economic abuse (partner, current partner, ex-partner or family)

• health abuse (partner, current partner, ex-partner or family)

• marital status-related abuse (family)

• threats (partner, current partner, ex-partner or family)

• domestic stalking (partner, current partner, ex-partner or family)

• domestic sexual assault (partner, current partner, ex-partner or family)

• physical abuse (partner, current partner, ex-partner or family)

Like the existing questions, the abuse types outlined can be broken down by demographic characteristics, subject to sample sizes and disclosure control constraints.

4.2.3 Defining a victim

Defining a victim of domestic abuse using data from a victimisation survey is not without its limitations given the complexity of many victim's experiences and the space available on a survey to ask about these. Given this, the definition presented is not a criminal definition in the same way that the Domestic Abuse Act (DAA) 2021 does not make domestic abuse a criminal offence. This section outlines the research process we used to decide how to define a victim of domestic abuse, using the new questions on the CSEW.

We considered using the impact questions alongside the behaviour questions to define a victim. However, that would mean we could only produce a prevalence measure for the time period of since the age of 16 years. Use of lifetime measures would not provide an accurate picture of progress when assessing policy change, as estimates accounting for abusive behaviours and their impacts since the age of 16 years would not show if the abuse occurred in the last year or, for example, 50 years ago. Therefore, given the requirement for an in the last year prevalence measure, we excluded using the impact questions to define a victim. Subsection 4.4 in Section 4: Relevance covers how we plan to use the impact questions.

Defining a victim based only on the behaviour questions meant deciding how many behaviours someone had to have experienced to be defined as a victim.

Using the existing questions, someone is defined as a victim of "any domestic abuse", if they answer "yes" to any of the domestic abuse-related questions asked on the self-completion module. They may have experienced one of the behaviours asked about or all of them. In some cases, we do not know exactly which behaviour they experienced, as multiple behaviours are included within the same question with no option to state which behaviour was experienced.

To determine how to define a victim using the new questions, we explored a number of options.

One option was to create a definition that aligned as closely as possible with the definition introduced in the DAA (see Subsection 7.2 in Section 7: Comparability and coherence for further information on the DAA). The DAA states, "it does not matter whether the behaviour consists of a single incident or a course of conduct". Therefore, using the new questions, someone could be defined as a victim of domestic abuse if they have selected one or more of the abusive behaviours.

One of the limitations of this definition is that it does not convey the spectrum of victims. In addition, the estimates may include people who have, for example, experienced one of the abusive behaviours but may not necessarily be a victim of domestic abuse, according to the legal definition.

With the data collected, we looked at the possibility of applying different definitions of a victim to understand if the definition that aligned most closely to the DAA was the most appropriate. For measures based purely on the behaviour questions, we produced trial prevalence estimates using different definitions. Those definitions were:

• selected 1 or more abusive behaviours

• selected 2 or more abusive behaviours

• selected 3 or more abusive behaviours

• selected 4 or more abusive behaviours

The results of applying those definitions showed that the greater the number of abusive behaviours selected, the lower the prevalence rates for the since the age of 16 years and in the last year time periods (Figure 1). When going from one or more behaviours to two or more behaviours, the since the age of 16 years prevalence rate fell from 26.1% to 19.6%. For the in the last year time period, the rates went from 8.0% to 5.0%.

When split by sex, for the in the last year time period, the gap between prevalence rates increased as the number of behaviours selected increased (Figure 2). This suggests that female victims were more likely to say they had experienced a greater number of abusive behaviours than male victims. The same relationship was seen for the since the age of 16 years time period.

Exploring the different victim definitions demonstrates that if we were to select a threshold above a single abusive behaviour, we risk omitting someone who has experienced one behaviour but is a victim. For example, between April 2023 and September 2024, of all respondents that said they had experienced a single abusive behaviour since the age of 16 years, nearly 1 in 5 (18.6%) selected a behaviour within the abuse type of domestic sexual assault or physical abuse.

To avoid omitting victims, we could pick certain behaviours, but we would risk discounting certain victims' experiences and not aligning to the DAA, which does not differentiate between abusive behaviours.

To define a victim of domestic abuse using the new survey questions, we have undertaken thorough research, acknowledging the effect the definition will have on the estimates produced. We have concluded that for our measures (since the age of 16 years and in the last year) we will define someone as a victim of domestic abuse if they have selected one or more of the abusive behaviours. For more information on the definition of a victim for different abuse types, see Section 11: Concepts and definitions. 

This definition remains consistent with one of our priority requirements to improve the headline prevalence measures, ensuring data align with the definition of domestic abuse introduced in the DAA. As outlined in our Developing a measure of controlling or coercive behaviour, April 2019 article, we undertook extensive cognitive testing of the new survey questions with both victims and non-victims of domestic abuse. This meant that the new abusive behaviours included within the questions aligned as closely as possible with the types of behaviours outlined in the DAA, and reflected lived experiences. 

The definition does not exactly match the offence of controlling or coercive behaviour, which was introduced in the Serious Crime Act (SCA) 2015. Further information on the SCA, including the criteria for defining a victim, is provided in Subsection 7.3 in Section 7: Comparability and coherence. The space available on the CSEW for questions on domestic abuse does not allow for the collection of the specific information required to meet the SCA definition, therefore it is not possible to produce estimates that accurately measure the offence.

Each of the abuse types for which we collect information includes elements of controlling or coercive behaviours, as outlined in the Home Office's Controlling or coercive behaviour: statutory guidance framework. This makes it extremely difficult to separate what is and is not controlling or coercive behaviour, based on the behaviour questions.

We acknowledge that the definition has limitations. It is not a criminal definition in the same way that the DAA does not make domestic abuse a criminal offence. Therefore, it is likely that the resulting estimates may include some people who would not be classed as victims of domestic abuse, especially in a criminal setting. However, we believe that the definition chosen balances this with the risk of omitting victims that have experienced one abusive behaviour.

It is therefore important that we produce our additional measure of domestic abuse, which accounts for abusive behaviours and their impacts. The measure provides context, and profiles the types of victims that are included in the prevalence measures, based only on abusive behaviours (see Subsection 4.4 in Section 4: Relevance for more information).

4.2.4 Resulting estimates of domestic abuse

The resulting prevalence estimates using our definition of domestic abuse, based on the DAA, were published in our Redevelopment of domestic abuse statistics: research update December 2024 article. The estimates were published as official statistics in development. Following our decision to include the questions permanently on the CSEW, the numbers in these tables should be treated as official statistics.

Within the tables published in December 2024, we included prevalence estimates for more abuse types than are available from the existing questions. The new questions also provide the opportunity to publish more detailed breakdowns, for example, whether the abuse type was perpetrated by a current partner or an ex-partner. The trial outputs presented an example of what can be produced, and we plan to continue developing these outputs based on user need.

We estimated that, for the year ending March 2024, 3.9 million (8.0%) people aged 16 years and over had experienced domestic abuse in the last year. Of these, 2.3 million (9.5%) were women and 1.5 million (6.5%) were men.

The estimates showed for the year ending March 2024, 12.6 million (26.1%) people aged 16 years and over had experienced domestic abuse since the age of 16 years. Split by sex, 7.4 million (30.3%) were women and 5.1 million (21.7%) were men.

More details can be found in our Redevelopment of domestic abuse statistics, data from split sample trial, England and Wales dataset.

4.3 The majority of users are satisfied with the outputs produced from the new questions

Alongside our Redevelopment of domestic abuse statistics: research update December 2024 article, we published a set of trial outputs, using data from the new questions.

We published the trial outputs in December 2024, in part, to gather user feedback on their format and structure. The trial outputs were not designed to be a final set of data tables, but instead to provide an example of the type of data we could release from the new questions. Although the table formats are similar to the CSEW tables currently published as part of our annual Domestic abuse prevalence and victim characteristics dataset, the new questions allow for the publication of estimates on a greater number of abuse types. We also included tables on the types of physical abuse victims have experienced since the age of 16 years, and the frequency of physical abuse experienced, to give an idea of the types of new tables we could produce.

The findings from our winter 2024 to 2025 user survey showed that, regarding the trial outputs:

• 45.5% of respondents were "very satisfied" or "quite satisfied"

• 18.2% were "neither satisfied nor dissatisfied"

• 36.4% were "dissatisfied" or "very dissatisfied"

We received a range of feedback suggesting additional tables. However, there was no clear consensus from users on what we should publish in the future beyond the data presented in the trial outputs. Where possible, we will add users' suggestions to our publication plan for future outputs.

The feedback showed users liked the additional breakdowns presented in Tables 1a and 1b in the trial outputs. In these tables, we published breakdowns for health abuse and marital status-related abuse for the first time and plan to continue to publish these in our domestic abuse compendium.

Users liked Table 5 in the trial outputs, which presented, for the first time, the prevalence of physical abuse, since the age of 16 years, by type of physical abuse and sex. Based on this feedback, we plan to include this table in our annual domestic abuse compendium in the future. One user requested we produce this table for the in the last year estimates. As the new questions only ask about serious physical abuse since the age of 16 years, not in the last year, it is not possible to publish this information.

Users requested that we publish a time series table, similar to Tables 3, 4 and 5 in our existing Domestic abuse prevalence and victim characteristics tables. We did not produce a time series for the trial estimates, as the changes to the questions mean the estimates produced from the new questions are not comparable with those produced from the existing questions. However, when two years of data are available from the new questions, we plan to produce a time series of prevalence estimates in both our Crime in England and Wales: Annual supplementary tables and our Domestic abuse prevalence and victim characteristics dataset. For more information on the comparability of the estimates, see Section 7: Comparability and coherence.

Users found Table 6 in the trial outputs, which presented frequency of physical abuse experienced since the age of 16 years, helpful. The current table is broken down into "once", "2 to 5 times", "6 to 10 times" and "More than 10 times". One user asked for frequency to be published ungrouped. However, we are unable to produce this, as the new questions do not ask respondents to specify the number of times they have experienced physical abuse. The findings from cognitive testing, published in our Redevelopment of domestic abuse statistics: research update November 2022 article, concluded that it was not possible to accurately count separate acts of abuse within the survey, and instead the questions focus on capturing frequency of physical violence using a banded approach.

One user requested that we produce a table presenting perpetrator relationships for family abuse. The new questions ask victims about the type of familial relationship they have with the perpetrator, and list 14 relationships. This is something we will look to publish in the future in our domestic abuse compendium.

Some users wanted the new outputs to include information on the impacts of domestic abuse. As outlined in Subsection 4.4 in Section 4: Relevance, we are currently working on producing abuse profiles that will take into account the abusive behaviours victims have experienced and the impacts of these. At this stage, we do not have data to share, but aim to include these in our domestic abuse release, which will be published later in 2025.

One user asked for the three health behaviours to be published individually, split by sex. This may be something we can look at publishing in the future, however, it will depend on whether enough people report each behaviour to ensure the estimates are robust and not disclosive.

Some of the information users requested we include in the new domestic abuse outputs would not be possible, as collecting the information is beyond the scope of the redevelopment work, or the information is not captured by the new questions.

The CSEW data collected will, in time, be available through the UK Data Service, meaning users can conduct their own analysis of the data and produce their own breakdowns. We are always open to ideas on additional tables that can be produced from the new questions and users can contact us with suggestions by emailing crimestatistics@ons.gov.uk.

4.4 Methods for producing abuse profiles have been tested and shared with users

When the new domestic abuse survey questions were developed, they were designed with the intention that the abusive behaviour and impact questions could be used together as outlined in our Redevelopment of domestic abuse statistics: research update November 2022 article.

The aim of the abuse profiles is to group together different types of victims that experience domestic abuse based on the abusive behaviours and impacts they have selected. The profiles will show the spectrum of victims based on their experiences since the age of 16 years. However, it will not be possible to differentiate between when the abusive behaviours occurred and when the impacts of those behaviours were felt. They could be recent, but they could also date back many years.

When the new questions were trialled on the CSEW on a split-sample from April 2023, we prioritised the development of the methodology for defining a victim outlined in Subsection 4.2 in Section 4: Relevance, and the overall evaluation of the questions, for example, question non-response analysis. This has been conducted while continuing the publication of data from the existing questions, as shown in our Domestic abuse in England and Wales overview: November 2024 bulletin.

The development and challenges of defining a victim of domestic abuse since the age of 16 years and in the last year reinforced the importance of producing abuse profiles. The profiles would provide greater context for the broader prevalence measures.

Deriving the profiles is more complex than deriving the two prevalence measures presented in our Redevelopment of domestic abuse statistics: research update December 2024 article. We are currently exploring a cluster analysis approach to derive abuse profiles for the since the age of 16 years time period. The aim is to apply clustering techniques to identify distinct subgroups, or clusters, of victims who experience similar levels of abusive behaviours and impacts. We are exploring the feasibility of applying hierarchical methods to find the optimum number of clusters.

Although we are unable to share any results at this stage while we continue to explore methods, we aim to share more information in our annual domestic abuse compendium later in 2025. We remain flexible in our choice of methodology and will communicate further as we progress.

We recognise that we have not made as much progress as we intended when this criterion was set. We have plans over the coming months to continue testing methods and sharing the outcomes. We are confident that we will be able to meet this criterion in the future and have therefore concluded that this does not affect our decision to permanently include the new domestic abuse questions on the CSEW.

Accuracy is defined as the closeness between an estimated result and the (unknown) true value. Reliability is defined as the closeness of early estimates to subsequent estimated values. For more information, see the Government Analysis Function's quality article. This section evaluates the three criteria we set in the context of the accuracy and reliability quality dimension.

5.1 The new questions do not lead to a significant increase in non-response rates or non-response bias

To assess the accuracy and reliability of the new questions, we set out two criteria to explore the effects that they had on both non-response rates and non-response bias. It is important to understand whether the new questions have resulted in an increase in overall non-response compared with the existing questions. In addition, analysing the characteristics of those who have not responded to the new questions compared with the existing questions will show if there is an increased bias.

5.1.1 Defining non-response

Non-response in this section refers to cases where respondents answered questions on the face-to-face Crime Survey for England and Wales (CSEW) and the self-completion modules before the domestic abuse questions, but refused all, or part, of the domestic abuse module. Demographic information is collected earlier on in the face-to-face part of the survey, which enables us to conduct non-response analysis by characteristics.

The split-sample experiment allows us to compare non-response rates for the new and existing questions.

For this analysis, we have split non-response into the following categories:

• non-response to the partner abuse questions

• non-response to the family abuse questions

• non-response to both the partner abuse and family abuse questions

For the new questions, a respondent would be included in the non-response for the partner abuse questions if they refused to answer the partner screener question (which identifies whether they have ever had a partner) or if they answered the screener question but then refused to answer all of the since the age of 16 years behaviour questions. For the existing questions, the screener questions are less detailed, but the same logic applies as for the new questions.

For both the new and existing questions, everyone who agreed to complete the self-completion part of the CSEW receives the family abuse questions. Therefore, someone is classed as a non-respondent if they refused to answer all the family abuse since the age of 16 years behaviour questions.

5.1.2 Overall non-response

Numbers presented in this section cover the period of April 2023 to September 2024. Non-response rates to the existing questions presented in this section are consistent with rates seen since 2020.

Between April 2023 and September 2024, the non-response rate for the new questions was 1.4 percentage points higher than for the existing questions (3.2% compared with 1.8%, Figure 3). This equates to 630 and 334 respondents, respectively. Although this is an increase, the numbers are still very low in the context of just under 20,000 respondents receiving each set of questions in this time period.

The higher non-response rate for the new questions was mainly the result of the family abuse question non-response, which increased to 4.3% for the new questions compared with 2.2% for the existing questions. In contrast, the partner abuse question non-response rates were similar for the new and existing questions (5.0% and 4.7%, respectively). More detailed statistics on non-response are available within our accompanying supplementary tables.

Although the non-response rates for the new questions are higher, the increase is somewhat expected. The new domestic abuse questions are substantially longer and more detailed compared with the existing questions. It is possible that this may have had an impact on non-response, particularly for the family abuse questions, as most respondents would have already answered a large number of questions on partner abuse before they are presented with them.

Although the non-response rates have almost doubled in some cases, the overall non-response rate for the new questions remains relatively small. The level of non-response has therefore not affected our decision to include the new questions. We will continue to monitor the non-response rates from the new questions over time.

5.1.3 Characteristics of non-respondents

Assessing non-response rates by respondent characteristics is important in identifying any potential impact on estimates derived from the questions. Our accompanying supplementary tables provide more detail on non-response rates by different characteristics.

Non-response bias is a feature of estimates derived from surveys and can be seen by analysing non-response rates by characteristics. The existing questions have a level of non-response bias, which has continued into the new questions, though at differing levels.

Non-response rates for the new partner and family abuse questions for women and men were higher compared with the existing questions (Figure 4). When comparing by sex for the new questions, the rates were slightly higher for men compared with women. This is a change from the existing questions where non-response rates were slightly higher for women than men; a trend we have generally seen since 2020.

Similarly, non-response rates in each age group were higher for the new questions than for the existing questions, but were still low. The exception to this was the 16- to 19-year-old age group, where the non-response rate was higher for the existing questions (2.4%) than for the new questions (1.0%, Figure 5). However, it should be noted that the numbers of people not responding to the new and existing questions aged 16 to 19 years are extremely small.

The difference in non-response rates between the existing and new questions was largest for the over 60 years age groups however, the rates for the over 60 years age groups remained low at below 3.0%. We will continue to monitor non-response rates as we receive more data.

We conducted further non-response analysis on other characteristics including:

• ethnicity

• religion

• disability

• sexual orientation

• whether the respondent was UK born

The findings generally showed the non-response rates followed the trends seen in the overall rates where the new questions had a slightly higher rate than the existing questions. However, the rates remained low and there were no specific characteristics that did not follow this trend. More information on the non-response rates can be found in our accompanying supplementary tables.

5.2 The new questions accurately measure lived experiences of victims of domestic abuse

To assess whether the new questions accurately measure lived experiences of victims of domestic abuse, we considered several elements. We have undertaken extensive user engagement and testing with respondents and victims, analysed the confidence intervals of the estimates produced, and examined in detail the responses to a number of questions. This section outlines these elements.

5.2.1 User feedback

In Subsection 4.1 in Section 4: Relevance, we outlined the user engagement we have undertaken throughout the question redevelopment. One of the main aims of the engagement was to obtain feedback.

When the University of Bristol developed the new questions, they did so as part of a consortium including the College of Policing, Women's Aid Federation of England, Men's Advice Line, Imkaan and Welsh Women's Aid. In addition, draft questions were shared with users including the Domestic Abuse Commissioner's Office, Home Office, Verian, ManKind Initiative and academics who had expressed interest in the redevelopment. We asked users their views on the draft questions and the feedback fell into five themes.

• Users were generally satisfied with what the questions were aiming to measure, but most provided some comments on their wording.

• The majority of users were satisfied with the list of domestic abuse behaviours and their ordering.

• Users felt asking participants about abuse in the last 12 months, as opposed to asking about a "current partner", was more quantifiable and would provide a clearer prevalence measure.

• Asking about "current partner abuse" as a proxy measure would not capture post-separation abuse.

• The majority of users agreed that a lifetime (since the age of 16 years) prevalence measure was not necessary every year.

The feedback was shared with the University of Bristol to aid them in updating the questions. Following receipt of the draft domestic abuse questions, we held a stakeholder event in October 2022 where we presented details of the new questions to approximately 60 users. Feedback gathered was generally positive with 84.6% of attendees either "very confident" or "somewhat confident" that the new questions would address the issues with the existing questions. Users raised some areas of concern, for example, the importance of ensuring that there was no sex bias in the questions. We addressed these concerns by making amendments to the questions.

From April 2023, we included the new questions on the CSEW via a split-sample trial. Following this, throughout 2023 to early 2025, we continued to meet with users, this time on an individual basis to receive further feedback. The main themes of this feedback were:

• future outputs from the questions

• how to derive prevalence

• the impact of losing the existing survey questions

• the time series of data

For all these, strong cases have been made from a range of users with different requirements and backgrounds. We have acknowledged every user's feedback and implemented changes where possible, while recognising that it was not possible to meet everyone's requirements.

Some of these themes also came through in the results of the user survey we conducted between 3 December 2024 and 5 January 2025. In the survey, we gave users another opportunity to feed back their views. We asked to what extent they agreed with the statement:

"The new domestic abuse questions accurately capture the lived experiences of domestic abuse victims."

Of the 13 responses to this question, 53.8% of people either "agreed" or "strongly agreed", 7.7% "neither agreed or disagreed" and 38.5% "disagreed" or "strongly disagreed".

When asked whether they thought some abuse types were missing from the questions, half felt there were, with the other half feeling the questions captured all types. Generally, users felt that there were some aspects missing from some of the abuse types covered by the survey. In addition, specific requests were made for:

• broader questions on technology-facilitated abuse

• questions specifically identifying faith-based abuse

• questions on child victims witnessing domestic abuse

The survey also asked if there were any groups of victims that would not be identified through the new questions. Over half of respondents (53.8%) felt that this was the case, specifically noting:

• child victims of domestic abuse

• those in communal establishments

• male victims

• victims with learning difficulties

• transgender and gender nonconforming victims

As participants in the CSEW are aged 16 years and over and resident in households in England and Wales, its scope does not extend to children or those living in communal establishments.

We have many users of our domestic abuse statistics that come from a wide range of professions all with valid and competing requirements. Throughout the redevelopment process, we have developed questions with user requirements in mind, and we have sought and acted on user feedback. This has been done alongside balancing user requirements for other crime statistics on a survey that has limited space.

Throughout the process, feedback on the new questions has been largely positive. However, we acknowledge that the results of the most recent user survey showed that over a third of users (38.5%) felt the questions did not align with lived experiences. Although this demonstrates that some users may not be completely satisfied with the contents of the new questions, the majority (76.9%) felt that the new questions should be included permanently on the CSEW.

There are always areas for continuous improvement that can be made with survey questions. One of the benefits of the new questions, compared with the existing questions, is that we can more easily adapt the question wording, without the need for another large-scale redevelopment project. Therefore, we can change some of the questions in the future if there is a clear user need to do so.

5.2.2 Testing questions with victims

Between December 2020 and February 2023, we conducted thorough testing of the new domestic abuse questions. This involved a group of 90 participants, consisting of both those who identified themselves as victims of abuse and those who did not. More information on the characteristics of participants can be seen in our Developing a new measure of domestic abuse: April 2023 article.

Results from the cognitive interviews showed that the new questions raised very few issues in terms of cognition, with the majority of people indicating the questions were easy to answer, straightforward and clear. Some people, particularly those who had not experienced domestic abuse, felt that the behaviours asked about in the questions were not harmful and did not mind being asked about them. In contrast, others understood the same behaviours as abusive and controlling. People also found the impact questions clear, with victims stating they covered an accurate range of physical and non-physical harms and measured both the short-term and longer-term impacts of abuse.

As shown in the University of Bristol's research report, asking respondents to think about any domestic abuse they have experienced by a partner presented a challenge for some victims, as those who were still living in an abusive relationship often did not recognise what they were experiencing as abuse. The new intimate partner question set has therefore been designed so that respondents are encouraged to think separately about abusive behaviours experienced by their current partner, ex-partner(s), or both.

Overall results from the testing showed that people did not find the new questions overly intrusive or uncomfortable. Victims stated they would be happy to answer the questions because they felt it was necessary to ask them as they described what they had been through.

Based on the feedback from victims, we are confident that the questions developed will measure lived experiences of domestic abuse as accurately as is possible on a survey limited by space. More information on the outcomes of the testing can be found in our Redevelopment of domestic abuse statistics: research update November 2022 article and our Developing a new measure of domestic abuse: April 2023 article.

5.2.3 Comparing the new and existing questions

In our Redevelopment of domestic abuse statistics: research update December 2024 article, we explained that the estimates produced from the new CSEW questions would not be comparable with the estimates from the existing questions (see Subsection 7.1 in Section 7: Comparability and coherence for more information). However, when we outlined the research questions we would use to assess the evaluation criteria in our Redevelopment of domestic abuse statistics: research update November 2023 article, we stated that we would compare the widths of the confidence intervals around the estimates from the existing and new questions. Confidence intervals give an indication of the degree of uncertainty and provide information on the precision of the sample estimate. They specify a range of values likely to contain the unknown population value. These values are defined by lower and upper limits.

If the confidence interval is much wider for one estimate compared with another, it may suggest there is an issue with the accuracy or reliability of that estimate. The purpose of the comparisons is not to understand how much we may be overestimating or underestimating abuse with either set of questions, nor should attempts be made to combine the estimates or analyse differences between the estimates. This is because the existing and new questions are not comparable for the reasons stated in Subsection 7.1 in Section 7: Comparability and coherence.

We cannot compare estimates to infer statistically significant differences because they are drawn from different question sets. We have only made comparisons of confidence intervals for estimates that are measuring the same broad abuse type. There are some abuse types, for example, marital status-related abuse, which are not covered in the existing questions. This section presents the confidence intervals for the headline prevalence measures. However, further breakdowns, including by sex, are available in our accompanying supplementary tables.

For estimates since the age of 16 years, the confidence intervals were marginally wider for the new estimates compared with the existing estimates (Figure 6). The width of the domestic abuse since the age of 16 years estimate, derived from the existing questions, was 1.9 percentage points, compared with 2.1 percentage points for the estimate derived from the new questions. For estimates of family abuse, the confidence interval widths were almost the same. Split by sex, the differences in confidence interval widths followed the same trends.

Figure 6: The confidence intervals for the new since the age of 16 years estimates were slightly wider than for the existing estimates

Confidence intervals and estimates from the existing and new questions for the since the age of 16 years time period, England and Wales, year ending March 2024

Notes:

1. Data presented from the new domestic abuse questions are not comparable with other domestic abuse estimates produced from the CSEW. Comparisons should only be made regarding the widths of the confidence intervals.
2. Data presented are based on a split-sample. Caution should be taken because of the impact of the reduced sample size on the quality of the data.

Download the data

For estimates in the last year, the 95% confidence intervals followed a similar trend where the intervals were wider for the new estimates compared with the existing estimates (Figure 7). The width of the domestic abuse in the last year estimate, derived from the existing questions, was 1.0 percentage points, compared with 1.3 percentage points for the domestic abuse estimate derived from the new questions. Split by sex, the widths followed a similar trend.

Figure 7: The confidence intervals for the new estimates were slightly wider than the existing estimates for the in the last year time period

Confidence intervals and estimates from the existing and new questions for the in the last year time period, England and Wales, year ending March 2024

Notes:

1. Data presented from the new domestic abuse questions are not comparable with other domestic abuse estimates produced from the CSEW. Comparisons should only be made regarding the widths of the confidence intervals.
2. Data presented are based on a split-sample. Caution should be taken because of the impact of the reduced sample size on the quality of the data.

Download the data

The analysis presented should be treated with caution because comparisons between abuse categories, such as our overall domestic abuse estimates, have been made despite there being large differences between the questions used to derive the confidence intervals.

Although not completely comparable, the analysis is necessary for us to understand the possible uncertainty around the estimates. Our analysis of the 95% confidence intervals suggests the widths of the confidence intervals are similar between the existing estimates and new estimates, although slightly wider for the new estimates.

5.2.4 Response to impact questions

Adding the impact questions substantially increases the number of survey questions for a respondent to answer. Before the split-sample trial, we set a research question to explore if respondents were answering the impact questions or if there was a drop off in response for those who received them.

The new questions include two sets of impact questions, one asking about the impacts of partner abuse experienced and the other asking about the impacts of family abuse experienced.

Between April 2023 and September 2024, we found that a very small number of victims did not answer the impact questions (1.1% of victims of partner abuse and 1.9% of family abuse victims). We also found that 0.5% of partner abuse victims did not answer the partner impact questions and then went on to refuse the family abuse behaviour questions. Again, this suggests victims are not deterred from continuing the survey after answering a larger number of questions. This is in line with the findings from the cognitive testing with victims of domestic abuse.

Split by sex, the non-response rates were very similar, with men having a slightly higher non-response rate to the impact questions (Figure 8). Analysis by other characteristics was limited, given the very low numbers of refusals to the impact questions.

We can be confident that only a small number of victims who identified they had experienced an abusive behaviour did not then answer the corresponding impact questions. Following the inclusion of the new domestic abuse questions on the full CSEW sample, we will monitor the effects on response to the sexual assault and stalking modules that follow these questions.

5.2.5 Response to physical abuse questions

We advise against comparison between the new and existing questions for all abuse types. However, for physical abuse, as the new question is very similar to the existing question, we have made some comparisons.

The wording for the new physical abuse question is:

"Since you were 16, has a current partner or an ex-partner used force on you or attacked you, for example, pushed, shoved, hit or kicked you".

The existing question wording is:

"Since you were 16 has a partner or ex-partner ever used force on you? For example, they may have pushed you, slapped you, hit, punched or kicked you, choked you or used a weapon against you".

The wording for the equivalent family member questions follows the same format.

The physical abuse question is positioned differently within the question sets, and the new question is presented in a grid format, unlike the existing question. These factors may affect response.

The prevalence estimates for the since the age of 16 years time period were higher for the existing questions than for the new questions (Figure 9). Split by sex, the prevalence rate was higher for women and for men for the existing questions compared with the new questions.

Physical abuse prevalence rates for the in the last year time period were higher for the new questions compared with the existing questions (Figure 10).

We identified that data collected from the question on physical abuse would be important to analyse given the similarities between the new and existing question wording. The abuse since the age of 16 years estimates, derived from the existing questions, were higher for all people. This was driven by the prevalence rate for women. A reason for this could be the position of the question within the survey, or differences in how the question is presented. However, the in the last year estimates are higher when derived from the new questions. It is unclear why we are seeing these differences, and we will continue to monitor as we collect more data. Although we are including the new questions permanently on the CSEW, we will remain flexible with wording changes or question positioning if it is necessary to do so to improve the accuracy of the estimates.

5.2.6 Response to frequency questions

During questionnaire development, we asked victims of domestic abuse about the frequency of abuse experienced, specifically physical abuse, as outlined in Section 2 of our Redevelopment of domestic abuse statistics: research update November 2022 article. Consistently they expressed reservations, both in terms of recall and appropriateness. When developing the new questions, victims stated that measuring frequency of both physical and non-physical abuse with numbers was not feasible, as most could not remember or count how many times abusive incidents had happened to them. They suggested using a banded approach, but only to the physical abuse question, as something they could respond to.

We took this advice, and the new questions ask victims of physical abuse if they have experienced force or been attacked "once", "two to five times", "six to ten times", or "more than ten times" since the age of 16 years. We ask this question once in the partner abuse set of questions and then again in the family abuse set of questions. The partner physical abuse frequency question is presented in a grid format, where the victim can state how many times a current partner, ex-partner(s), or both, have physically abused them. A respondent will only see the physical abuse frequency question if they state a partner or family member "used force on you or attacked you, for example, pushed, shoved, hit or kicked you".

We published estimates based on this question in our Redevelopment of domestic abuse statistics, data from split sample trial, England and Wales dataset. In addition, we explored how many victims refused to answer the frequency questions. We found the percentage of ex-partner and family member physical abuse victims that refused the respective frequency question was relatively low (2.3% and 2.8%, respectively). For current partner physical abuse victims, the refusal rate was higher at 7.1 % (Figure 11). Given the non-response rates and the feedback from testing, we will continue to monitor these questions following the permanent inclusion of the new questions on the CSEW.

Timeliness is the time gap between the publication date and the reference period for the statistics. Punctuality is the time lag between the actual and planned dates of publication for the statistics. For more information, see the "Timeliness and punctuality" subsection of the Government Analysis Function's Quality statistics in government article. This section evaluates the criterion we set in the context of the timeliness and punctuality quality dimension.

6.1 The new questions do not affect the timeliness and punctuality of our domestic abuse publications

An important factor in the redevelopment of the domestic abuse questions was ensuring that the estimates and outputs from the new questions can be produced to the same timeline as those from the existing questions.

Alongside our Redevelopment of domestic abuse statistics: research update December 2024 article, we produced a set of trial outputs. The format of the data tables was similar to the existing outputs. However, they contained additional information and breakdowns including more abuse types, a breakdown of the physical abuse behaviours, and the frequency of physical abuse. User feedback showed that the majority felt the tables were "very easy" or "easy" to access (81.8%) and interpret (72.7%). We plan to continue to produce outputs from the new questions in a similar format in the future. For more information on the trial outputs and user feedback, see Subsection 4.3 in Section 4: Relevance.

When producing the trial outputs, we found the time taken for production was similar to the time taken to produce the existing outputs. Because of the greater number of estimates being produced, marginally more resource will be required. However, we do not anticipate that this will affect our timeline of table production and publication.

We plan to continue publishing domestic abuse prevalence measures in our quarterly Crime in England and Wales: Appendix tables and in our Domestic abuse in England and Wales overview bulletin, published as part of our annual domestic abuse compendium. In the future, we may decide to publish additional breakdowns from the new questions. We will ensure that any additional output production does not affect the planned timeline of the headline estimates.

Part of this evaluation criterion was to assess whether we could produce more complex statistics, such as the modelling of abuse profiles, within this same timeline. As detailed in Subsection 4.4 in Section 4: Relevance, we are working on developing a method to produce abuse profiles. These will take account of the abusive behaviours experienced by victims, as well as reflecting the impacts of those behaviours. We will continue to communicate further as we progress with the development and aim to share findings in our Domestic abuse in England and Wales overview bulletin, which will be published as part of our annual domestic abuse compendium later in 2025.

In future years, we anticipate that we will publish the abuse profiles as part of the domestic abuse compendium, and that this will not affect the publication of the other prevalence estimates. However, we will only know this after the development has finished. Once we have developed an approach and shared it with users, we will seek their views on the frequency of producing abuse profiles.

Comparability is the degree to which statistics can be compared over time, by region or by another domain. Coherence is the degree to which the statistical processes that generate two or more outputs use the same concepts and harmonised methods. For more information, see the "Comparability" subsection in the Government Analysis Function's Quality statistics in government article. This section evaluates the two criteria we set in the context of the comparability and coherence quality dimension.

7.1 The comparability between estimates produced using the new questions and the existing time series is clear to users

In our Redevelopment of domestic abuse statistics: research update December 2024 article, we explained to users that the estimates produced using the new domestic abuse questions are not comparable with the existing estimates, meaning the existing time series cannot be maintained. As the existing questions fail to measure the information required by users, it is therefore to be expected that the newly developed questions will not be comparable. In this section, we provide more detail on why the questions are not comparable.

7.1.1 Question structure

The existing questions in the domestic abuse module were first designed in 1996, and permanently added to the Crime Survey for England and Wales (CSEW) from April 2004. The questions have remained largely unchanged since, enabling a long comparable time series for 16- to 59-year-olds. For more information on the existing time series, see Subsection 2.2 in Section 2: Domestic abuse and the Crime Survey for England and Wales.

We conducted extensive testing to design a new set of questions to address these issues and added them to the CSEW on a split-sample basis in April 2023. While the structure of the new questions is different from the existing questions, there are still some broad similarities. Both question sets include questions on partner abuse, followed by questions on family abuse. The respondent is initially asked partner screener questions and if the respondent has not had an intimate partner, they will skip to the questions on family abuse. In both versions, questions ask respondents about their experiences in the last year (12 months) and since the age of 16 years. However, while the overall layout of the survey module remains similar, there are substantial differences in how the questions are structured.

Image 1: Example question from the existing domestic abuse question set

Download this image Image 1: Example question from the existing domestic abuse question set

.png (107.5 kB)

Image 1 shows one of the existing domestic abuse questions. The question asks the respondent:

"Since you were 16 has a partner or ex-partner ever frightened or threatened you in any way".

The question then provides examples of abusive behaviours:

"For example, they may have threatened to hurt you, to kill you, to use a weapon on you, or to hurt someone close to you (such as your children)"

The respondent is then presented with four response options in a list from which they can select only one:

1. Yes

2. No

3. Don't know or can't remember

4. Don't wish to answer

Most questions in the existing domestic abuse question set follow a similar format.

Image 2: Example question from the new domestic abuse question set

Source: Crime Survey for England and Wales (CSEW) from the Office for National Statistics

Download this image Image 2: Example question from the new domestic abuse question set

.png (32.7 kB)

Image 2 shows one of the new domestic abuse questions. The question asks the respondent:

"Since you were 16, has a current or an ex-partner"

The question then lists abusive behaviours that they can respond to, using a grid format. Wording for current partner, ex-partner, or both, is included in the question dependent on the respondent's answers to the current and ex-partner screener questions. There are six abusive behaviours listed in the grid rows:

1. Deliberately destroyed or damaged property or your belongings, for example, punched walls, threw or smashed things

2. Threatened to use sensitive personal information about you to discredit you, for example, your disability or mental health condition, sexuality or immigration status

3. Threatened to hurt or kill themselves if you did not do what they wanted

4. Used the "authorities" as a threat, for example, social services, mental health services, police, immigration agencies or your employer

5. Threatened to hurt or kill you

6. Threatened to hurt or kill someone or something you care about, for example, your children, family, friends, or pets

The respondent is presented with four or five response options (dependent on whether they have or had a current or ex-partner) for each behaviour in the grid columns, from which they can select only one per row:

1. Current partner

2. Ex-partner(s)

3. No

4. Don't know or can't remember

5. Don't wish to answer

This means that respondents who state in the screener questions that they have a current partner and have also had an ex-partner, can report whether they experienced each abusive behaviour from their current or ex-partner(s). Most questions in the new domestic abuse question set follow a similar format.

This grid format was implemented after research, as described in the University of Bristol's report, found that asking respondents to think about domestic abuse experienced by a partner was challenging for some victims. This is because those who were still living in an abusive relationship often did not recognise what they are experiencing as abuse. Prompting respondents to think about a current partner increases the likelihood that this type of abuse is reported by respondents, meaning a greater number are likely to be identified as victims of domestic abuse. The remaining response options are unchanged.

The grid format of the new questions means that respondents can respond individually to each of the abusive behaviours. This means we can identify the specific abusive behaviours victims have experienced from those listed in the question. This offers more detail than the existing questions.

7.1.2 Question content

The new questions collect information on more abuse types than the existing questions. In addition, for abuse types that were covered by the existing questions, the new questions include more behaviours. This is in line with our aim to align the estimates with the Domestic Abuse Act (DAA) 2021 and to capture controlling or coercive behaviour (CCB).

The physical abuse question is similar in both question sets. The new and existing question sets both ask victims if they have experienced "force" in the last 12 months or since the age of 16 years, and if they have been "hit" or "kicked". Unlike most of the other new questions, the physical abuse question only asks about one behaviour, similar to the structure of the existing questions. However, although the estimates are produced in a similar way, they are not entirely comparable because of the different examples and question format. More information about comparisons between the new and existing physical abuse questions can be found in Subsection 5.2 in Section 5: Accuracy and reliability.

The new questions ask about the impact of domestic abuse. While the existing questions ask respondents if they had experienced being "repeatedly belittled to the extent that you felt worthless", this was not separated out from the other abusive behaviours. The existing questions ask respondents who had experienced physical abuse if they had been "injured (even if only slightly) as a result of the force used upon you", with examples of "bruises, black eyes, cuts, scratches, or broken bones". However, they did not ask about the impact of any of the other abuse types.

During the research led by the University of Bristol in 2020 to 2021, users and victims highlighted that it was important for the survey questions to account for the impact of abusive behaviours in deriving domestic abuse measures. In the new questions, if a respondent reported that they had experienced one of the abusive behaviours, they would be asked further questions on the impacts of this. There are six impact questions, made up of 23 potential impacts that respondents can select.

By asking more questions about a wider range of abusive behaviours and abuse types, as well as their impacts, there are more opportunities for victims to align their lived experiences to the questions. This, alongside the structure of the new questions, means that the estimates produced from them are not comparable with the estimates produced from the existing questions.

7.1.3 User feedback

We provided information on the comparability of the estimates in Section 2 of our Redevelopment of domestic abuse statistics: research update December 2024 article. Alongside this, we published a user survey, where we asked to what extent users agreed with the statement:

 "It is clear why the estimates produced using the existing questions and estimates produced using the new questions are not comparable."

Of the respondents:

• 76.9% "strongly agreed" or "agreed" with the statement

• 7.7% "neither agreed nor disagreed"

• 15.4% "disagreed" or "strongly disagreed"

The percentages are based on a relatively small number of people.

We also asked to what extent users agreed with the statement:

"The benefits of the new estimates outweigh the benefits of maintaining a comparable time series with the existing questions."

Over half (61.5%) of users "strongly agreed" or "agreed", 30.8% "neither agreed nor disagreed" and 7.7% "disagreed" or "strongly disagreed".

While the permanent inclusion of the new questions on the CSEW means we are unable to maintain the existing time series, we acknowledge that it is important to ensure the comparability of the new domestic abuse estimates in the future, to enable a new time series to be produced. The new questions have been designed to minimise the need for large updates in the future, ensuring that, as far as possible, the behaviours captured will remain largely relevant. However, to ensure the questions continue to accurately capture victims' lived experiences, we may, in time, need to update the wording of the questions or the behaviours and impacts within them.

If updates to the wording of the abusive behaviours or impacts are made, the prevalence estimates should remain comparable year on year. The abuse types being measured will remain the same, as will the structure of the questions themselves. For example, the questions will continue to measure emotional abuse, however it is perceived or defined at the time, using the same grid structure. We will use the same methodology to produce the estimates, as outlined in Subsection 4.2 in Section 4: Relevance.

There are advantages to maintaining the existing time series, mainly for the analysis of trends over time and combining multiple years of data. However, the new questions will allow us to better capture the lived experiences of domestic abuse victims and produce better quality data. It would not be beneficial to continue the split-sample trial to maintain the existing time series because of the impact on the quality of the domestic abuse estimates, as well as the sexual assault and stalking estimates.

The majority of users felt it was clear why the estimates were not comparable and that the benefits of the new questions outweighed the benefits of maintaining the existing time series. We have implemented steps to ensure the new estimates can remain comparable in future years, to ensure a new time series can continue to be produced in the future. Once the data are available through the UK Data Service, we are happy to speak to users about any analysis they are doing on trends or combined samples to support their analysis.

7.2 Estimates produced using the new questions closely align with government definitions of domestic abuse in the Domestic Abuse Act 2021

One of the main reasons for redeveloping the domestic abuse questions on the CSEW was that they did not align with the government definition of domestic abuse as set out in the Domestic Abuse Act (DAA) 2021. We set a research criterion assessing the alignment of the estimates from the new questions with the definition, and whether any differences were clear to users.

This section outlines the definition set out in the Act, explaining where the new questions do and do not align.

7.2.1 The Domestic Abuse Act 2021

The DAA 2021 was introduced to raise awareness and understanding about the devastating impact of domestic abuse on victims and their families. The DAA does not make domestic abuse a criminal offence. Instead, it defines domestic abuse in a more detailed and informative way than was previously available.

7.2.2 The victim and perpetrator must be aged 16 years and over

The DAA states both the victim and perpetrator must be aged 16 years and over. The CSEW asks people aged 16 years and over about their experiences of crime and abusive behaviours, and the new domestic abuse survey questions ask about experiences since the age of 16 years and in the last 12 months. We have routed questions asking about experiences in the last 12 months so that only respondents aged 17 years and over are asked them. This ensures that only experiences where the respondent was 16 years and over are captured.

The introductory text for the family abuse questions reminds respondents to think only about experiences with family members who were aged 16 years and over at the time of the abuse, and each question prompts respondents to think about this family member when responding. The intimate partner introductory text does not specify this, however, the partner screener questions specify the definition of a partner, including age.

7.2.3 The victim and perpetrator must be personally connected

The DAA provides examples of "personally connected" relationships. The new questions ask respondents if they are currently or have been in an intimate partner relationship since the age of 16 years. The examples provided include girlfriend, boyfriend, husband, wife and civil partner. If respondents have not had an intimate partner, they will be directed straight to the family abuse questions.

The DAA states that the victim and the perpetrator may be relatives. The introductory text provides an extensive list of family relationship examples, and prompts respondents to include step, adopted, foster and in-law family members. Each question asks about "a member of your family (other than a partner)", reminding the respondent to answer with the family members in mind.

The DAA states that the victim and perpetrator may "have, or there has been a time when they had, a parental relationship in relation to the same child". The new questions ask about ex-partners, meaning they should capture people who have shared responsibility for a child but are no longer partners. There may be rare circumstances where two individuals share parental responsibility for a child but have not been in an intimate relationship or are not family members, and this may not be captured by the new questions. However, it would also not have been captured by the existing questions.

7.2.4 Behaviour may be "towards" the victim despite the fact it consists of conduct against another person

The DAA outlines that someone may be a victim of domestic abuse through experiencing the impact of abusive behaviour aimed towards another person, for example, the victim's child.

The new questions ask respondents if their current partner, ex-partner or family member has "threatened to hurt or kill someone or something you care about, for example, your children, family, friends or pets".

The new impact questions ask if respondents believed the perpetrator would hurt their children (emotionally or physically) or someone they care about.

7.2.5 Children as victims of domestic abuse

The DAA states that children who "see, hear or experience the effects of the abuse" and are related to the victim or the perpetrator are victims of domestic abuse.

A child is defined as a person under the age of 18 years. As the CSEW is an adult victimisation survey, we do not specifically collect information on child victims. Although the survey does cover 16- to 18-year-olds, the new questions do not explicitly ask about this. They do ask about threats to someone or something you care about, meaning that some child victims may be captured. However, it would not be possible to identify whether the child was present when the threats were made. To meet this data need, we have been assessing the feasibility of a survey to measure the prevalence of child abuse. For more information see our Exploring the feasibility of a survey measuring child abuse in the UK article.

7.2.6 Threats to disclose private sexual photographs and films with intent to cause distress

The DAA states that threats to disclose private sexual photographs and films is an abusive behaviour when the perpetrator intended to cause distress.

The new questions ask respondents if a current partner, ex-partner or family member has "shared, or tried to share, intimate photos or videos of you without consent". They do not explicitly ask about threats to disclose, however they do include attempts. They ask about threats to use sensitive personal information to discredit, meaning some victim's experiences may be captured by these questions.

The CSEW is a victim-based survey, so we cannot assess the perpetrators' intentions.

7.2.7 Strangulation and suffocation

The DAA states that strangulation and suffocation must be intentional to be considered abusive behaviour. If a victim has consented, this is considered to be a defence unless:

• serious harm is caused

• the perpetrator intended to cause harm

• the perpetrator was reckless as to whether they would cause harm

The new questions ask respondents if a current partner, ex-partner or family member has "strangled, choked or suffocated you" since the age of 16 years. As the CSEW is a victimisation survey, we cannot assess the intent of the perpetrator. The question also does not ask if there was consent. However, it is asked within the context of domestic abuse and respondents will only see it if they said that someone has "used force on you or attacked you".

7.2.8 Abusive behaviour

The DAA states that behaviour is abusive if it consists of any of the following:

• physical or sexual abuse

• violent or threatening behaviour

• controlling or coercive behaviour

• economic abuse

• psychological, emotional or other abuse

The Domestic Abuse Statutory Guidance (PDF 1,343 KB), provided alongside the DAA, lists 82 example abusive behaviours, of which the new questions align or partially align with 48 (58.5%). Some questions capture multiple behaviours.

The abusive behaviours that were least captured were those relating to technology-facilitated abuse and abuse relating to faith. While the new questions do not include any specific questions that ask about these abuse types, cognitive testing with victims from a range of demographics (as explained in Section 4 of our Developing a new measure of domestic abuse: April 2023 article), found they felt the new questions aligned with their lived experiences.

The DAA contains further information on domestic abuse that we have not included in this assessment, because it is not relevant to measuring abuse or possible for us to capture in a victimisation survey. These sections include:

• the Domestic Abuse Commissioner

• powers for dealing with domestic abuse

• local authority support

• protection for victims and witnesses in legal proceedings

• miscellaneous and general

7.2.9 Alignment to the new questions

The DAA was one of the main reasons for the redevelopment of the domestic abuse questions on the CSEW and this section has outlined where the new questions do and do not meet the definition it sets out.

Our analysis found that the new questions align closely to the definition. Where they did not, this was largely because collecting this information would fall outside of the remit of the redevelopment of the domestic abuse questions and a victimisation survey. The new questions are limited by survey space, meaning we cannot measure all aspects of domestic abuse. However, we expect that victims of these abuse types will be able to identify with some of the behaviours meaning that no victim is completely excluded.

7.3 Estimates produced using the new questions closely align with government definitions of domestic abuse in the Serious Crime Act 2015

One of the issues with the existing domestic abuse questions is that they do not capture the offence of controlling or coercive behaviour in an intimate or family relationship, which was first introduced as part of the Serious Crime Act 2015. We set a research criterion assessing the alignment of the estimates from the new questions with the definition of the offence, and whether any differences were clear to users.

This section outlines the definition set out in the Act, explaining where the new questions do and do not align.

7.3.1 The Serious Crime Act 2015

The Serious Crime Act 2015 (SCA) was introduced to ensure that the National Crime Agency, the police and other law enforcement agencies had the powers they needed to effectively pursue, disrupt and bring to justice serious and organised criminals. Section 76 of the Act introduced the criminal offence of controlling or coercive behaviour in an intimate or family relationship, to close the gap in the law around patterns of controlling or coercive behaviour and to recognise its severe impact.

7.3.2 At the time of the behaviour, the victim and perpetrator are personally connected

The SCA states that a victim and perpetrator are "personally connected" if they are in an intimate personal relationship, or they live together and are either members of the same family or have previously been in an intimate personal relationship with each other. Like the DAA, we found the new questions closely aligned with this definition. However, the new questions do not ask respondents if they currently live with the family member or person they have previously been in an intimate relationship with.

The SCA states that, regarding the controlling or coercive behaviour category, "the perpetrator does not commit an offence if at the time of the behaviour, the perpetrator has responsibility for the victim, and the victim is under the age of 16 years". The new questions only ask about experiences where the victim and perpetrator were aged 16 years and over.

7.3.3 The behaviour has a serious effect on the victim

The SCA states that a behaviour has a "serious effect" if it causes the victim to fear, on at least two occasions, that violence will be used against them, or it causes serious alarm or distress which has a substantial adverse effect on their usual day-to-day activities.

To understand the effects of abuse, the new questions ask about the impacts of the abusive behaviours, for example respondents are asked if they believed the perpetrator would "hurt you or your children (physically and emotionally)". Where applicable, the word "fear" has been included within the impact question wording.

We do not ask about the frequency of impacts that victims experienced. Cognitive testing, as outlined in the University of Bristol's report, showed victims felt it was not possible to count individual acts of abuse. It is likely that victims would have similar difficulties recalling the frequency of impacts they experienced as a result of the abuse.

Some of the questions ask about impacts that could be defined as serious alarm or distress. For example, respondents are asked if they have "suffered with mental ill health, for example, anxiety or depression", "self-harmed, or had thoughts of self-harm" and "attempted suicide, or had suicidal thoughts". Most, if not all impacts asked about in the questions, could have substantial adverse effect on the victim's day-to-day activities, however, this is not explicitly asked.

Our cognitive testing and previous research have shown that not all victims who experience domestic abuse realise what they are experiencing is abuse and may not report a substantial adverse effect on their day-to-day activities. For this reason, and because of limited survey space, the new questions do not ask if each behaviour experienced had an adverse effect on day-to-day activities.

7.3.4 The perpetrator knows, or ought to know, that the behaviour will have a serious effect on the victim

The SCA states that the perpetrator must know, or ought to know, that their behaviour would have a serious effect on the victim. However, it is a defence if they believed they were acting in the victim's interests, and the behaviour was in "all the circumstances reasonable".

As the CSEW is a victim-based survey, we cannot assess what the perpetrator's thoughts or intentions were, meaning the new questions cannot align to this under the design of a victimisation survey.

7.3.5 The perpetrator repeatedly or continuously engages in behaviour towards the victim that is controlling or coercive

Most of the new questions do not ask about the frequency of abuse. Cognitive testing, as outlined in the University of Bristol's report, showed victims felt it was not possible to count individual acts of abuse, but preferred for questions exploring non-physical coercion and control to have "yes" or "no" response options, as opposed to frequency scales.

We have chosen to use words like "constantly" and "repeatedly" within the abusive behaviour questions where possible to align to this part of the definition. For example, "repeatedly prevented, or tried to prevent, you from spending time with family, friends, or colleagues".

7.3.6 Controlling or coercive behaviour

We also compared the new questions with the Home Office's Controlling or coercive behaviour: statutory guidance framework, which accompanies the definition provided in the Act.

The guidance provides 38 examples of behaviours within the range of controlling or coercive behaviours, of which 23 (60.5%) aligned with the new questions. There were no clear themes of behaviour types that did not align with the new questions, however abuse relating to faith and verbal abuse were least captured. The guidance also outlines 14 examples of impacts that constitute a "serious effect", and the new impact questions align with 11 of these.

7.3.7 Alignment to the new questions

The definition of controlling or coercive behaviour, as defined in the SCA, is complex. Developing a set of questions on a survey limited by space, but is easily understandable and can be used to derive measures of controlling or coercive behaviour that meet the SCA definitions, has been challenging. To achieve this, the questions would need to cover:

• the full range of abusive behaviours

• whether each behaviour has happened continuously and had a serious effect on the victim

• whether the perpetrator knew that the behaviour would have a serious effect on victim

We have communicated this challenge throughout the redevelopment process, for example at user events, and when we first published the new questions in Section 7 of our Developing a new measure of domestic abuse: April 2023 article.

Of the prevalence measures we have already published in our data from the split-sample trial, this section has highlighted that the estimates from each of the abuse types include behaviours that would be considered controlling or coercive.

We anticipate that the abuse profiles, as outlined in Subsection 4.4 in Section 4: Relevance, will provide measures that are more closely aligned with the offence of controlling or coercive behaviour. This is because they will include the impacts of abusive behaviours, which other prevalence measures do not. This will be a lifetime (since the age of 16 years) measure, so will not indicate whether the abuse was recent or many years ago. It will also not specify if the perpetrator knew their behaviour would have a serious effect on the victim.

Overall, given the difficulties of aligning the questions with the offence of controlling or coercive behaviour, we believe the new estimates will align as closely to the SCA's definition as is possible within the parameters of a victimisation survey. Producing a set of estimates achieving this alignment was an important aim of the redevelopment, and we believe the new questions and estimates provide a substantial improvement compared with the existing questions.

Accessibility is the ease with which users can access statistics and data, the format in which data are available and the availability of supporting information. Clarity refers to the quality and sufficiency of the commentary, illustrations, accompanying advice and technical details. For more information, see the 'Accessibility and Clarity' subsection of the Government Analysis Function's Quality statistics in government article. This section evaluates the two criteria we set in the context of the accessibility and clarity quality dimension.

8.1 Estimates produced using the new questions are presented in a format that is easily available and clear to users

As we outlined in Subsection 4.3 in Section 4: Relevance, we published a set of trial outputs using estimates from the new questions alongside Section 3 of our Redevelopment of domestic abuse statistics: research update December 2024 article and asked for user feedback on their accessibility in our winter 2024 to 2025 user survey.

The contents of the data tables were provisional. They were published to show the type of outputs that could be produced from the new questions and to seek feedback on their content and format. When designing the outputs, we aimed to keep the table style and ordering similar to those currently published in our Domestic abuse prevalence and victim characteristics dataset, which meet accessibility requirements.

In our user survey, we asked users how easy they found it to access the trial outputs. Around 4 in 5 users (81.8%) found it "very easy" or "quite easy" to access them, while the remaining 18.2% found it "neither easy nor difficult".

Most users (72.7%) found it "very easy" or "quite easy" to interpret the data presented in the trial outputs, with 18.2% finding it "neither easy nor difficult" and 9.1% finding it "difficult" (Figure 12).

It is important to note that the percentages are based on a relatively small number of people.

Additional feedback from users suggested that those who found it difficult to interpret the trial outputs thought there should be more information on which new questions were used to produce each of the estimates and breakdowns presented in the tables. In Subsection 4.2 in Section 4: Relevance, we detailed how the new estimates are produced, and Section 11: Concepts and definitions explains which abusive behaviours fall under each abuse type. We also plan to include this information in the table notes in the future, so that it is clear to users how the new questions feed into our outputs.

Overall, user feedback suggests that the trial outputs are accessible and easy to interpret, and therefore we plan to publish the estimates from the new questions using this format in the future.

8.2 Supporting information about the changes to the questions, and estimates produced using them, is easily available and clear to users

In Subsection 7.1 in Section 7: Comparability and coherence, we outlined the changes we have made to the domestic abuse questions to address the issues with the existing questions. It is important to ensure that information on these changes is easily available and clear to users, to enable understanding of the data and potential data uses.

We have published regular updates providing supporting information about the changes to the questions and estimates in our Redevelopment of domestic abuse statistics research updates since 2021. We have published these updates alongside our annual domestic abuse compendium, so that the information is easily accessible to users at a time when we know they will be accessing our website and reading our articles. The exception to this is our December 2024 research update article, which we published after the compendium because of the inclusion of trial estimates from the new questions. We have provided further details on the changes to the questions in Section 7: Comparability and coherence to ensure users are aware that the new domestic abuse estimates are not comparable with existing estimates. We will continue to include information on this alongside the publication of estimates in the future.

Alongside our Redevelopment of domestic abuse statistics: research update December 2024 article, we published a user survey where we asked users to what extent they agreed with the statement:

 "Supporting information about the changes to the questions and estimates is easily available and clear".

The majority of respondents (61.5%) either "strongly agreed" or "agreed" with this statement, 15.4% "neither agreed nor disagreed" and 23.1% "disagreed" or "strongly disagreed".

The feedback suggests that most users felt the information on the changes to the questions and the estimates was easily available and clear, however not all users agreed. We hope that the information provided in this report on the changes to the questions and impacts of these, as well as how the estimates are produced, will ensure that users fully understand the changes. While the new questions have been included permanently on the Crime Survey for England and Wales (CSEW), we will continue to work on improving the accessibility and clarity of our domestic abuse publications and supporting information.

Following extensive evaluation, as documented in this article, we have decided to include the new domestic abuse questions on the Crime Survey for England and Wales (CSEW) permanently from April 2025 onwards, fully replacing the existing questions. The new questions provide huge opportunities to explore and present the lived experiences of victims of domestic abuse. This will help government, law enforcement, victim services, academia and the general public to better support victims, and ultimately help to prevent domestic abuse.

This article has provided the findings for each evaluation criterion for the new questions. There remain areas where we will seek to make improvements over the coming months and years. For example, we will continue to work to meet the criterion: "The majority of users are satisfied with the outputs produced from the new questions".

We acknowledge that we have not shared methods for producing abuse profiles at this stage, and therefore we have not yet met this criterion. However, we will endeavour to publish this information alongside our domestic abuse compendium later in 2025.

Although not all the evaluation criteria have been fully met, as outlined, we have plans in place to address gaps, and therefore this did not prohibit our ability to make a decision on the status of the new questions.

The design of the new questions should mean that future large-scale changes to the domestic abuse module are not necessary, and the beginning of a new long time series can commence. We aim to publish headline estimates from the new questions alongside our Crime in England and Wales: year ending March 2025 bulletin in July, as well as more detailed estimates in our annual domestic abuse compendium later in 2025.

We are extremely grateful to everyone involved in the redevelopment of the domestic abuse survey questions. We look forward to future engagement to help shape our new domestic abuse outputs, as we continue to develop them, following the permanent inclusion of the new questions.

Current partner

A current partner is someone with whom the respondent was in an intimate relationship at the time of completing the Crime Survey for England and Wales (CSEW).

Domestic abuse

Domestic abuse is not limited to physical violence and can include a range of abusive behaviours. It can also be experienced as repeated patterns of abusive behaviour to maintain power and control in a relationship. The Domestic Abuse Act 2021 defines domestic abuse as any incident or pattern of incidents between those aged 16 years and over who:

• are a partner

• are an ex-partner

• are a relative

• have, or there has been a time when they each have had, a parental relationship in relation to the same child

The Domestic Abuse Act 2021 outlines the following behaviours as abuse:

• physical or sexual abuse

• violent or threatening behaviour

• controlling or coercive behaviour

• economic abuse

• psychological, emotional, or other abuse

The Domestic Abuse Act 2021 recognises children under the age of 18 years who see, or hear, or experience the effects of the abuse, as a victim of domestic abuse, if they are related or have a parental relationship to the adult victim or perpetrator of the abuse.

Existing questions

CSEW domestic abuse questions that were permanently included on the survey since April 2004. They were removed from the survey in April 2025 and replaced with the new questions. The last estimates produced from the existing questions were for the year ending March 2024.

Ex-partner

An ex-partner is someone with whom the respondent was in an intimate relationship, but at the time of completing the CSEW is no longer in this relationship. Data collected from the CSEW on ex-partner abuse will include abuse by someone with whom the respondent was in an intimate relationship at the time of the abuse, as well as abuse from someone with whom they were no longer in an intimate relationship at the time of the abuse.

In the last year

In the last year refers to the time period for which domestic abuse questions on the CSEW are asked. Specifically, respondents are asked whether they have experienced something in the last 12 months.

Family member

In the context of domestic abuse on the CSEW, family member includes the respondent's father or mother, son or daughter, brother or sister, or any other family member who was aged 16 years and over (for example grandparent, grandchild, cousin, uncle, or aunt). It also includes family members who are step, adopted, foster, or in-laws.

New questions

CSEW domestic abuse questions that have been developed and trialled on the survey between April 2023 and March 2025. They were added to the full survey permanently from April 2025.

Since the age of 16 years

Since the age of 16 years refers to the time period for which domestic abuse questions on the CSEW are asked. Specifically, respondents are asked whether they have experienced something since they were 16 years old. This allows for an understanding of whether something has happened in the respondent's adult life.

Split-sample trial

A split-sample trial is used to test new methods. Respondents are randomly allocated to groups and a different method is applied to each group. In the split-sample trial used in the redevelopment of the domestic abuse questions, half of respondents received the new questions and were not presented with the sexual assault and stalking questions. The other half of respondents were presented with the existing domestic abuse questions, along with the sexual assault and stalking questions. The split-sample trial took place between April 2023 and March 2025.

Victim

Using the new domestic abuse questions on the CSEW, a victim of domestic abuse is defined as someone who has been a victim of at least one of the specified abuse types, and therefore experienced one or more of the abusive behaviours.

Victims of domestic abuse types

Emotional abuse 

A victim in this category is defined as someone who experienced one or more of the following behaviours:

• constantly monitored who you talked to or where you went, for example, called you or monitored your text messages, social media, emails

• repeatedly prevented, or tried to prevent, you from spending time with family, friends, or colleagues

• tried to turn your family, friends, or others against you, or convinced them you were "crazy" or had a mental health problem

• acted in an overly jealous way, or repeatedly accused you of being unfaithful

• constantly blamed you for their behaviour or their alcohol use, drug taking or self-harm

Economic abuse

A victim of economic abuse is defined as someone who experienced one or more of the following behaviours:

• controlled, or tried to control, your spending or access to the household income (including your benefits) or savings

• deliberately got you into debt, for example, by taking out credit in your name without your consent

• made you give them your income or personal belongings, or took them without your consent

• prevented you from getting a job or studying, or made it difficult for you to do so

• repeatedly refused to pay their share of the rent, mortgage, or household bills

Health abuse

The health abuse category was added to measure more abusive behaviours that are prevalent in controlling or coercive relationships as stated in the Home Office's Controlling or coercive behaviour: statutory guidance framework. The term "health abuse" is not one used in the DAA, but is a term coined to summarise specific behaviours. A victim of health abuse is defined as someone who experienced one or more of the following behaviours. Specifically, a person: 

• controlled, or deliberately deprived you of, your daily essentials such as food, sleep, medication or living aids, for example, your wheelchair

• deliberately tried to control whether or not you used contraception, became pregnant, or forced you to terminate a pregnancy (for women)

• deliberately tried to control whether or not you used contraception (for men)

• deliberately prevented you from accessing support services or medical assistance, for example, for injuries or illness

Marital status-related abuse

The marital status-related abuse category was added as this type of abuse is included in the DAA and the Home Office's Domestic Abuse: statutory guidance. The term "marital status-related" is not one used in the DAA, but is a term coined because the behaviours are marriage related. However, the behaviours are not limited to forced marriage, as they also include the prevention of marriage. A victim of marital status-related abuse is defined as someone who said they have experienced one or more of the following behaviours:

• forced, or tried to force you, to marry someone you did not want to marry

• prevented, or tried to prevent you, from marrying someone of your choice

Domestic threats

A victim in this category is defined as someone who experienced one or more of the following behaviours:

• deliberately destroyed or damaged property or your belongings, for example, punched walls, threw or smashed things

• threatened to use sensitive personal information about you to discredit you, for example, your disability or mental health condition, sexuality or immigration status

• threatened to hurt or kill themselves if you did not do what they wanted

• used the "authorities" as a threat, for example, social services, mental health services, police, immigration agencies or your employer

• threatened to hurt or kill you

• threatened to hurt or kill someone or something you care about, for example, your children, family, friends, or pets

Domestic stalking

A victim of domestic stalking is defined as someone who experienced one or more of the following behaviours. Specifically, a person:

• sent you threatening, offensive or unwanted mail, emails, texts, or messages on social media more than once

• made threatening, offensive or nuisance phone calls to you more than once

• followed you or hung around your home, workplace or somewhere else you regularly go to more than once 

Domestic sexual assault

Domestic sexual assault has previously been derived using the CSEW questions from the sexual assault module. The new domestic abuse questions include questions on sexual assault. Domestic sexual assault estimates will now be derived from these questions. The sexual assault module is currently being redeveloped and will provide more detail than the domestic sexual assault questions on the new domestic abuse module. Any differences between estimates derived from both modules will be clearly communicated.

Using the new domestic abuse questions, a victim of domestic sexual assault is defined as someone who experienced one or more of the following behaviours. Specifically, a person:

• has shared, or tried to share, intimate photos or videos of you without your consent

• agreed to sexual activity because it was less hassle to give in to their demands

• agreed to sexual activity because you were afraid of what they might do if you didn't agree

• has forced you to take part in any form of sexual activity when you did not want to, or you were unable to refuse 

Physical abuse

A victim of physical abuse is defined as someone who experienced the following behaviour. Specifically, a person:

• used force on you or attacked you, for example, pushed, shoved, hit or kicked you