Redevelopment of domestic abuse statistics: research update November 2022

The way domestic abuse manifests is constantly changing. To ensure our statistics continue to provide the most accurate information and meet the needs of users, we have been undertaking a user engagement, research and testing program to improve the collection of data on domestic abuse.

The questions on domestic abuse are asked as part of the self-completion section of the Crime Survey for England and Wales (CSEW). These have largely remained the same since they were introduced in April 2004, allowing for a long comparable time series. However, there are issues with the data currently collected, mainly: 

• they do not align with the definition of domestic abuse introduced in the Domestic Abuse Act 2021

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

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

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

Given these, significant changes to the questions were needed. In line with wider CSEW transformation, we also looked at survey mode to consider alternative options.

This update outlines our most recent research, findings, and plans for redeveloping domestic abuse statistics over the coming months. As the work progresses, we will continue to engage with stakeholders and provide updates.

In Autumn 2021, we put out a research tender to take forward the redevelopment of domestic abuse statistics following on from our previous work. There were two main parts to the research:

• to develop new survey questions on domestic abuse in the Crime Survey for England and Wales (CSEW) based on topic knowledge, user requirements and qualitative testing with victims

• to investigate the impact of survey mode when asking respondents about their experience of domestic abuse

In November 2021, we awarded a research contract to a consortium led by the Centre for Gender and Violence Research at the University of Bristol, that also involved the College of Policing, Women's Aid Federation of England, Men's Advice Line, IMKAAN and Welsh Women's Aid. The research took place between November 2021 and April 2022 and was carried out in two stages.

Stage one

The aim of stage one of the research was to develop and test an initial set of survey questions as well as investigate respondents preferred survey mode.

Developing questions

Based on the findings and recommendations from the previous research, alongside an analysis of questions used in other international domestic abuse surveys, an initial set of 25 questions were developed. These covered relationship status and context, controlling or coercive behaviour, physical assault and use of weapons, sexual coercion, and the impact of non-physical and physical abuse. When drafting the questions, consideration was given to:

• the difficulty of reflecting the subtle nature of controlling or coercive behaviours, and the possibility that victims may not recognise such behaviours as abusive while still in the relationship

• how to design questions that accurately reflect the experiences of both female and male victims

• how to mitigate against the possibility that experiences which are not abusive are reported

Findings from cognitive testing with victims 

To refine the questions, they were tested with 27 victims with differing characteristics and experiences through a mixture of online focus groups and interviews. The testing explored how the questions should be framed, whether they captured lived experiences, the ease of responding and preferred survey mode.

The research showed:

• the questions should be framed in the context of domestic abuse rather than crime

• the questions captured most victims' experiences, but suggestions for additions were made

• measuring frequency of abuse with numbers is not feasible, as most victims could not remember or count how many times abusive incidents happened to them; using options such as never, once, frequently, or weekly, monthly was preferred

• victims found the impacts included in the questions very relevant, but highlighted they are often not recognised while in an abusive relationship

• further testing was needed to explore the relationship between asking about current partner and experiences in the last 12 months

• victims preferred to have a choice of mode for completing the survey

Stage two

The aim of stage two of the research was to further develop the questions from stage one through cognitive testing with victims, members of the public who had not experienced domestic abuse, and consultation with survey stakeholders. In addition, survey mode was explored further.

Developing questions

Based on the findings from stage one, the survey questions were reduced and amended. As it was still unclear if asking questions about a current partner would produce better results than asking about experiences in the last 12 months, further testing was needed. It was also important to test whether the same set of questions were applicable in both intimate partner and family domestic abuse contexts.

Findings from cognitive testing with general public and victims 

The revised questions were cognitively tested with 30 individuals, a mixture of victims of domestic abuse and a sample of the public with no experience of domestic abuse. The cognitive interviews explored problems of cognition, recall, judgement, response, and question sensitivity.

Cognition

The testing showed most questions raised few issues in terms of cognition with most respondents suggesting they were easy to answer, straightforward and clear. The feedback also indicated possible wording revisions for some questions, and suggestions for amendments to the examples given to make some questions less ambiguous.

Sensitivity of questions

Although some behaviours asked about in the questions could be deemed innocuous to some, particularly those who have not experienced domestic abuse, they may also be experienced as abusive and controlling by others. The testing showed that the questions were not overly intrusive or uncomfortable, with victims suggesting they would be happy to answer them as they felt they described their lived experience.

Measuring frequency

The testing showed victims felt it was not possible to count individual acts of abuse, but preferred, for questions exploring non-physical coercion and control, a yes or no response 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.

"Last 12 months" versus "current partner"

The research showed that asking about domestic abuse experienced in the last 12 months would present a challenge to some victims, as those who are still living in an abusive relationship will often not recognise what they are experiencing as abuse. The testing explored whether specifically asking a respondent to think about abuse experienced by a current partner could instead be used as a more accurate measure for recent abuse. Participants were also asked what time period they were thinking of when they were recalling the information to answer the questions. The findings suggested that people who are still in an abusive relationship might not necessarily be able to answer the questions when phrased "in the last 12 months".

Survey mode

Participants were asked about their likelihood of responding to the questions via different survey modes and potential issues with completing an online survey on domestic abuse. The testing showed that people preferred to be given a choice as to how to complete the survey. Although the majority would prefer to fill it in online, an option for an interviewer visiting the respondent's home was still felt to be required in some instances.

Findings from stakeholder feedback

Consultation on the questions was carried out with key Crime Survey for England and Wales (CSEW) and domestic abuse stakeholders including representatives from government departments, survey experts, third sector organisations and academia. Stakeholders were asked for their views on both the question wording and options regarding the structure of the questions.

Stakeholders suggested amendments to the question wording, and identified additional examples of impacts of abuse which they felt were missing. Regarding the structure of the questions, stakeholders emphasised the importance of including questions covering current partner, ex-partner and family abuse each year so an overall domestic abuse prevalence measure could be produced. They also highlighted the need to include questions asking specifically about experiences in the last 12 months to provide a clear measure and reference point.

Research outcomes

Based on the findings from stage 2, the survey questions were amended. The final questions consisted of two sets: one regarding current or previous partner behaviour, and one regarding family behaviour. The questions covered non-physical coercive control, sexual abuse, physical violence, and impacts of such behaviours.

Questions about current and ex-intimate partners will be asked simultaneously followed by questions relating to family members. Questions will initially focus on whether specific behaviours happened since the age of 16, followed by summary questions relating to whether specific types of abuse (for example, non-physical, physical, sexual) happened in the last 12 months. This approach will provide detailed data on abuse occurring in both current and previous intimate relationships and familial settings while also retaining the ability to estimate a 12-month prevalence rate.

As the testing concluded it was not possible to accurately count separate acts of abuse within the survey, the questions instead focus on capturing frequency of physical violence using a banded approach.

The research also suggested that victim abuse profiles could be used as a possible method for deriving measures of prevalence of domestic abuse and coercive or controlling behaviour. Abuse profiles which take account of both the level of abusive behaviour and the impact of such behaviour on the victim could be used to differentiate types of victims. For example, it may be important to identify victims of "one-off" or infrequent abuse where the impact felt is reported as limited, separately from those suffering ongoing, coercive abuse.

Initial work to establish abuse profiles and assess severity using responses to the behaviour and impact questions suggested that profiles of high/high (high levels of behaviour and high levels of impact) and low/low (low levels of behaviour and low levels of impact) could be measured using statistical methods involving general linear models. This approach however would require further testing with a larger sample of data.

Recommendations

The research concluded with the following recommendations:

• the number of participants with experience of family abuse who took part in the cognitive testing was relatively small, therefore the questions require further testing with victims with experiences of a range of family abuse

• the questions require large-scale testing, for example as part of a split-sample experiment

• the questions should ask about current partner abuse, previous partner abuse and family abuse experienced since the age of 16, with follow up questions asking if any of the behaviours happened in the last 12 months

• subject to further testing, the headline prevalence measure should take account of both behaviours and impacts experienced by the victim and be calculated using abuse profiles

• abuse profiles should be published to provide comparison of those likely in need of services and support, and those who are less likely to need such support

• an online approach where respondents complete the questionnaire online at a scheduled time with the presence of an interviewer via an online link should be considered in combination with other options including face-to-face self-completion

The survey questions developed during the latest research, along with the implementation of these recommendations, will address the issues identified with the data currently collected. The new questions will capture a wider range of abusive behaviours, aligning with the definition of domestic abuse introduced in the Domestic Abuse Act 2021. The questions will also capture the offence of controlling or coercive behaviour and will measure the frequency of physical violence.

Following on from the outcomes and recommendations of the research led by the University of Bristol, we have undertaken further work over the last few months to update the questions to ensure they can be implemented within the context of the Crime Survey for England and Wales (CSEW) self-completion module.

Over the next few months, we plan to conduct further qualitative testing with victims with experiences of a range of family abuse to ensure those questions accurately capture lived experiences. We will also assess the total length of the new question set and, in conjunction with other crime topics included on the CSEW, prioritise what should be included and make any necessary amendments to the questions. Prior to inclusion on the survey, usability testing of the questions will be conducted to ensure the questions work within the context of the self-completion module when handed to a respondent on a laptop or tablet.

We aim to complete this work by early 2023 and introduce the new questions onto the survey from April 2023 on a split-sample basis.

Prior to the questions being added to the survey, we will continue to develop the method for calculating prevalence measures including defining abuse profiles. In Spring 2023, we will publish the final set of questions and our plans for calculating prevalence, alongside our evaluation criteria for assessing the success of new questions.

For further information on the current collection of data on domestic abuse see the domestic abuse quality and methodology information report.