Findings

Transparency about the redevelopment work

2.1 ONS has regularly updated users of its progress on developing the new domestic abuse questions. For instance, the Developing a new measure of domestic abuse: April 2023 update gives a clear, transparent and accessible overview of the entire question redevelopment process. It sets out the criteria that ONS used to evaluate the success of the new questions.

2.2 ONS published an evaluation of the new questions in May 2025. The evaluation report presents detailed evidence on whether the new questions have met the evaluation criteria and is open and transparent about the areas where the statistics and accompanying tables can be improved. To make it easier for users to follow the progress of its evaluation, ONS should consider publishing a simple table summarising the evaluation criteria and the extent to which they have been met.

Measuring experiences of domestic abuse victims

2.3 ONS involved topic experts in drafting the new questions and in undertaking cognitive testing with individuals from diverse backgrounds to refine the questions. Involving experts from the outset helped ensure that the redevelopment process was robust and that the new questions reflect the lived experiences of victims.

2.4 ONS awarded the research contract to redevelop the domestic abuse questions to a consortium led by the Centre for Gender and Violence Research at the University of Bristol. It also involved the College of Policing and various domestic abuse charities and victim support organisations, including Women’s Aid Federation of England, Men’s Advice Line, IMKAAN and Welsh Women’s Aid. The Home Office, the Domestic Abuse Commissioner’s Office and academics were also consulted throughout the redevelopment process.

2.5 The consortium had clear aims from the start to develop and test new questions on domestic abuse for the CSEW to improve the headline prevalence measure, including frequency of abuse, to develop a way of measuring controlling or coercive behaviour within the overall prevalence measure, and to develop a measure of the impact of abuse. When drafting the questions, to ensure that they were accurately capturing the correct data, the consortium considered:

a. 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

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

c. how to mitigate against the possibility that experiences which are not abusive are reported by survey respondents as such

2.6 The suitability of the questions was tested in multiple ways. The consortium used online focus groups and interviews with domestic abuse victims to explore question framing and determine if they covered victims’ experiences. The survey questions were then amended based on the findings from these focus groups and interviews. The questions underwent cognitive testing with victims as well as with members of the public who had not experienced domestic abuse. These tests indicated that the questions worked well for both intimate partner abuse and family abuse and that non-physical abuse questions were clear and well received.

2.7 Prior to adding the new questions to the survey, ONS commissioned Verian (the contractor for the CSEW) to undertake usability testing of the questions to ensure that they work within the context of the self-completion module when handed to a respondent on a laptop or tablet. The testing involved face-to-face interviews with individuals with lived experience of domestic abuse and individuals with no experience of domestic abuse. ONS made minor amendments to the questions based on the results of this testing, including adding instructions about how to respond to questions and the reference period for questions.

2.8 Feedback from most users suggests that these new questions accurately measure what they are seeking to – the lived experiences of victims of domestic abuse. ONS told us that based on its one-to-one meetings with stakeholders and engagement with expert stakeholder groups, most users were happy that the new questions captured the experiences of victims.

2.9 ONS’s winter 2024/25 user survey asked users if they thought that the new questions accurately captured the lived experiences of victims of domestic abuse. The survey showed that 53.8% of people either “agreed” or “strongly agreed” that they do. However, the same survey also highlighted that 38.5% of users did not think that the new questions accurately captured the lived experiences of victims of domestic abuse. Users specified that faith-based abuse and technology-facilitated abuse were missing. While none of the questions specifically ask about these abuse types, cognitive testing with victims from a range of demographics found that they felt the new questions aligned with their lived experiences. Users also noted that the crime survey did not cover child victims witnessing domestic abuse and some other populations.

2.10 Of all the respondents, 76.9% agreed that the new survey questions should be permanently added to the CSEW. Based on the findings from the user survey and broader user engagement, ONS decided to roll out the questions to the full sample of CSEW respondents permanently from April 2025.

Ensuring the new questions align with legislation

2.11 One of the main aims of ONS’s redevelopment work was to align the new questions with the government definitions of domestic abuse in the Domestic Abuse Act 2021. ONS’s evaluation report explains how the estimates produced using the new questions closely align with the definitions in the Act.

2.12 ONS found that the new questions align or partially align with 48 out of the 82 (58.5%) example behaviours in the Home Office’s Domestic Abuse Statutory Guidance. It found that technology-facilitated abuse and abuse relating to faith are the behaviours that are least covered by the new questions. Reasons given for questions not aligning with the definitions in the Act include the information being outside the scope of the redevelopment work, not being able to capture the behaviour in a victimisation survey and having limited space in the survey.

2.13 ONS also considered how closely the new questions align with the definitions of controlling and coercive behaviour set out in the Serious Crime Act 2015. Again, the evaluation report outlines where the questions match closely with the legal

2.14 ONS provides some explanation on key areas where the survey does not align with the behaviours listed in the statutory guidance, for example, explaining that faith-based abuse and verbal abuse are not well covered by the new questions. ONS also invites users to contact the ONS Centre for Crime and Justice if they would like more-detailed information about which example behaviours listed in the statutory guidance are covered by the questions and which are not.

The quality of the new estimates

OSR review of the CSEW estimates

2.15 In October 2024, we reinstated the accredited official statistics status of all CSEW estimates, including the old domestic abuse estimates, after having temporarily suspended the accreditation in July 2022 due to quality concerns. A key concern was the substantial drop in the overall response rates from 70–76% before the COVID-19 pandemic to 42% in 2022–23 and how this may have impacted the quality of the CSEW estimates. These quality issues are also relevant to the new domestic abuse estimates, as ONS used data collected on a split-sample basis between April 2023 and September 2024 to produce the estimates in the July 2025 quarterly bulletin.

2.16 In our letter confirming the reaccreditation of the CSEW estimates, we said that we were satisfied that ONS had demonstrated that the quality of the current estimates was broadly similar to that of the pre-pandemic estimates, and considered that ONS provided enough reassurance to users that the lower response rate had limited impact on the quality of the estimates. We encouraged ONS to continue to monitor the quality of the estimates.

Impacts of the new questions on non-response rates

2.17 Two of ONS’s criteria for evaluating the success of the new questions are as follows:

a. They do not lead to a significant increase in non-response rates.

b. They do not lead to a significant increase in non-response bias.

2.18 Collecting data on a split-sample basis allowed ONS to examine non-response in the sample. ONS reported that the overall non-response rate for the new questions was 1.4 percentage points higher than for the existing questions (3.2% compared with 1.8% based on a sample size of just under 20,000 for each set of questions). Deeper analysis showed that the non-response rates largely stemmed from the family abuse questions, where non-response was 2.1 percentage points higher (4.3% for the new questions compared with 2.2% for the old questions).

2.19 ONS said that it expected to see an increase in the non-response rates and thought that this may be due to the extra length and detail of the new questions. ONS speculated that the larger increase in non-response for family abuse questions occurs because most respondents will have already answered many questions on partner abuse before they are asked about family abuse.

2.20 ONS also analysed the characteristics of non-respondents (such as ethnicity and age) to identify any potential impact and bias on the estimates derived from the new questions. It found that the non-response rates remained low, and all specific characteristics followed this trend. ONS did not explain the non-response rates for these characteristics in the evaluation report, but it did publish the data in the supplementary tables.

2.21 After considering ONS’s investigations, we do not have concerns about non-response. The overall non-response rate for the new questions has increased, but this increase has been slight, and the rate remains low across all respondent characteristics. We understand why this has not affected ONS’s decision to include the new questions and welcome ONS’s transparency about its analysis of non-response. It is good that ONS is continuing to monitor this aspect of the quality of the statistics and exploring the factors that may be causing the increase in non-response. We encourage ONS to engage with experts here and to publish any findings from this work so that users are informed about how the non-response rates impact the quality of the new domestic abuse estimates.

Precision of the new estimates

2.22 As the estimates from the domestic abuse questions are based on a sample of the population, the estimates are accompanied by a measure of uncertainty relating to how the estimate might differ from the “true value”. Confidence intervals indicate the degree of uncertainty of an estimate and help in determining how precise or close an estimate is to the “true value”. They specify a range of values likely to contain the unknown population value. These values are defined by lower and upper limits. For the domestic abuse estimates, ONS uses 95% confidence intervals. This means that if the questions were asked again to 100 different samples of similar size and nature, 95 times out of 100, the resulting estimate would fall somewhere between the upper and lower limits.

2.23 To assess the level of uncertainty of the new estimates, ONS compared the widths of the 95% confidence intervals around the estimates from the old and new questions. Although the estimates and confidence intervals are not directly comparable, if the confidence interval was considerably wider for estimates generated from the new questions compared with estimates generated from the old questions, it could suggest that there were issues with the accuracy and reliability of the estimates from the new questions.

2.24 The precision of the new estimates appears to be robust. ONS found that the confidence intervals are slightly wider for the estimates derived from the new questions compared to estimates from the old questions. However, the difference is fairly minimal, which should give users confidence in the quality of the new estimates. For instance, for the “domestic abuse since the age of 16 years” estimate, the confidence interval for the estimate derived from the new questions was 2.1 percentage points, compared with 1.9 percentage points for the estimate derived from the old questions.

Our judgement on the quality of the new estimates

2.25 We are satisfied that the quality of the new domestic abuse estimates is sufficient to meet the needs of users. Although we have not conducted an in-depth review of quality of the new domestic abuse estimates, based on the findings from our reaccreditation of the CSEW estimates and ONS’s evaluation of non-response and precision, the new domestic abuse estimates appear to be similarly robust to other CSEW estimates.

Ensuring the new questions and estimates meet user needs

2.26 ONS has engaged effectively with users and stakeholders through a variety of channels and throughout the redevelopment work. ONS has had one-to-one meetings with stakeholders in government, academia and the third sector. It ran multiple user surveys at different points throughout the development process to understand user needs and seek feedback on its work. ONS also hosted annual stakeholder events, where the stakeholders were given the opportunity to ask questions and provide feedback. Annex 4 summarises ONS’s main user engagement activities since summer 2020.

2.27 We welcome ONS’s proactive and collaborative approach to user engagement. Throughout the redevelopment process, ONS has demonstrated a willingness to maintain open dialogues, meeting or consulting with a wide variety of users in a range of ways. At each stage of question development, ONS sought feedback from stakeholders and addressed their views and implemented their suggestions where possible.

2.28 For example, measuring the impact of abuse on victims, such as their mental health, confidence or their physical health, was highlighted as essential by many users of the domestic abuse statistics. Users said that having data on impact is an important aspect of understanding how victims experience domestic abuse. As a result of this feedback, ONS developed questions that would measure the impact of domestic abuse on victims. After further testing and positive feedback from users, ONS decided to add the impact questions to the new domestic abuse questions.

2.29 Another good example of ONS seeking views from users and acting on their feedback is the October 2022 stakeholder event. Stakeholders were given the opportunity to provide feedback about whether the new questions would meet their needs and share their views on the redevelopment process. ONS said that the 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 did raise some areas of concern, such as the importance of ensuring that there was no sex bias in the questions. ONS amended the questions to address their concerns.

2.30 Not all user engagement activities were as effective as they could have been. For example, ONS ran a user survey between December 2024 and January 2025 to gather feedback on the content of the new questions, the new estimates, the methodology used to produce the new estimates and the supporting information around the estimates and the trial outputs. ONS only received 14 responses (compared to 39 in its 2020 survey) – a small number for a topic of high public interest with an engaged user base.

2.31 The low response rate was likely due to the survey running for one month over Christmas and New Year, when many stakeholders are busy. ONS told us the survey was run at this time to allow enough time to analyse responses and make a decision about rolling out the new questions permanently by the end of March 2025. The limited response rate to the winter 2024/25 survey underscores the need for a more strategic approach to timing and promotion, especially given the high public interest in domestic abuse statistics.

2.32 To gather further evidence on the effectiveness of ONS’s approach to user engagement, we spoke to several key users of the statistics that had been involved in the redevelopment process. We also monitored public comments on the evaluation report and domestic abuse estimates published in July 2025.

2.33 As noted above, many users were positive about the engagement and the extent to which the new estimates meet their needs. There were, however, some specific gaps highlighted. For example, one charity pointed to areas where the statistics do not meet their needs, such as children’s experience of domestic abuse not being captured. ONS explains in its evaluation report that only those aged 16 or over can participate in the CSEW. However, the survey does ask respondents about threats to someone or something they care about, and as such, some child victims may be captured by the survey.

2.34 ONS is continuing to explore the feasibility of a survey to measure the prevalence of child abuse, and in June 2024, it published a report on its work so far. ONS also explained that it will be reviewing the questions previously present in the “nature of partner abuse” module, which covers the respondent’s perception of children witnessing domestic abuse.

2.35 Some users we spoke to told us that it was unclear which types of questions refer to which types of abuse, and it was unclear what questions they are based on. These users also told us that they would have liked more data to evaluate the impact of the new questions.

2.36 Another user expressed frustration that the redevelopment has led to a break in the long-term time series, making it so that the new estimates are not comparable with the old estimates. However, ONS explained in its evaluation report that most users believed that improving the headline estimates was more important than maintaining the existing time series.

2.37 In the evaluation report, ONS sets out its view that it has listened to users and implemented changes where possible, but it has not been able to meet everyone’s requirements. The report states that 70% of the unique user requirements gathered throughout the redevelopment work have either been fully or partially met through the new questions.

2.38 We recognise that it is not possible to meet the needs of all users, and we consider that ONS has set out to balance the requirements and priorities of a wide range of users. ONS has been clear with users about which needs it has and has not met and details the reasons for this in its evaluation article. This engagement with users should continue, and any unmet user need should be identified and considered transparently alongside other resource and constraints and organisational priorities.

Clarity and insight of the new estimates

2.39 ONS published the first estimates from the new questions in the Crime in England and Wales: year ending March 2025 The presentation of the statistics is insightful, and ONS has provided clear guidance for interpreting the new estimates. For instance, the bulletin explains that the new estimates are not directly comparable with older estimates and that it is not possible to adjust the new estimates to make these comparisons.

2.40 ONS received specific feedback and suggestions for improvement in the winter 2024/25 survey. For instance, users found it helpful to have a table showing the prevalence of physical abuse, since the age of 16 years, by type of physical abuse and sex. This table was generated from the new questions. ONS plans to include the new table in its annual domestic abuse overview. Users also requested that ONS publish a time series table similar to those in the existing tables for domestic abuse prevalence and victim characteristics. ONS published a time series comparing estimates from the year ending March 2025 with those from the year ending March 2024 in the Crime in England and Wales: year ending March 2025

2.41 The evaluation report stated that ONS will include many of the suggestions for more data or detailed breakdowns in its plans for future outputs, including the annual domestic abuse overview. We will monitor these future outputs and follow up with ONS to understand the extent to which it has implemented these user suggestions.

2.42 Whilst ONS’s user guide to crime statistics for England and Wales talks about the split sample, the quarterly bulletin does not mention that the estimates are only based on half of the domestic abuse sample. Given that these estimates are widely used and feed into the combined prevalence measure used to monitor the UK Government’s mission to halve violence against women and girls, we consider that ONS should be transparent about the reduced sample size, as it may impact the conclusions drawn from the statistics.