Developing a measure of controlling or coercive behaviour

On 29 December 2015, the government introduced the offence of controlling or coercive behaviour in an intimate or family relationship as part of the Serious Crime Act 2015. This prompted us to conduct research¹ to develop new questions in the Crime Survey for England and Wales (CSEW) to measure the prevalence of this offence. The CSEW includes questions on domestic abuse and, in 2004, these were extended to cover experiences of non-physical abuse.

We introduced new questions in the CSEW on a split sample basis² in April 2017, with half of the survey sample receiving the trial controlling or coercive behaviour questions and half receiving the current non-physical abuse questions. This has enabled us to compare responses between the two samples to identify the impact that the new questions had on the overall domestic abuse estimates and assess whether they are fit for purpose. Specifically, we have evaluated them using the following research questions:

• how do the prevalence rates of victims identified by the two measures compare?

• is there a consistent relationship between different behaviours experienced and their impacts?

• were any differences between how males and females responded to the controlling or coercive behaviour impacts as expected?

• how did responses to the controlling or coercive behaviour questions compare with similar questions found elsewhere in the survey?

This research presents a number of challenges, given the complexity of this topic. It has highlighted some expected results, such as lower estimates of domestic abuse generated by the more tightly-defined controlling or coercive behaviour questions. However, the scale and nature of some of the changes warrant further investigation before putting confidence in the new questions and their resultant estimates. There are three main areas worthy of further investigation:

• the considerable extent of the difference in domestic abuse prevalence estimates generated by the two questions requires more in-depth research and exploration

• the difference between men and women in the prevalence of domestic abuse changed with the introduction of the controlling or coercive behaviour questions – the extent of this difference and the reasons why such a change occurred needs further investigation

• the wording and other aspects of the questions may be drawing in people who are not victims, or be missing those who are

Therefore, we have agreed that further development is needed and have removed the trial questions from the CSEW from April 2019 whilst further research is conducted.

This article presents the results of the research to date and outlines the next steps of the work.

Notes for: Summary of the research

1. This research was supported by the Office for National Statistics (ONS) Domestic Abuse Statistics Steering Group. This group is chaired by ONS and includes representatives from across government, academia, the voluntary sector and Kantar Public who run the CSEW. See Annex A for further information.

2. Respondents were randomly assigned to either the original questions on non-physical domestic abuse or the new questions on controlling or coercive behaviour.

We currently produce estimates of domestic abuse using data collected in the Crime Survey for England and Wales (CSEW). The survey includes a self-completion module¹ that asks respondents aged 16 to 74² years about their experiences of domestic abuse. This module captures data on family abuse and partner abuse, and includes questions on non-physical abuse, threats or force, sexual assault and stalking.

Until the introduction of controlling or coercive behaviour as a separately identifiable offence, there was no specific criminal offence for any type of domestic abuse. Instead, all domestic abuse-related offences were recorded under other offences, for example, assault with injury³.

The new offence of controlling or coercive behaviour captures many of the non-physical acts of domestic abuse. In the current survey questions, non-physical abuse is estimated using a single measure.

Non-physical domestic abuse question

A question is included on the CSEW to measure prevalence of non-physical abuse. Respondents are asked whether they had been:

• prevented from having their fair share of the household money

• stopped from seeing friends and relatives

• repeatedly belittled to the extent that they felt worthless

Respondents are asked in turn if they experienced any of these behaviours by a partner or ex-partner, or by a family member. They are asked if they have experienced these behaviours since the age of 16 years and, if so, whether they had experienced them in the last 12 months.

Notes for: The current survey measure

1. For more information about the CSEW self-completion module, please see Section 2.3 of the User guide to crime statistics for England and Wales.

2. The module was previously restricted to respondents aged 16 to 59 years. The age limit was increased to 74 years in April 2017. All analysis in this article was conducted on data from 16- to 59-year-olds.

3. This is still the case for domestic abuse-related offences that do not constitute controlling or coercive behaviour.

Controlling or coercive behaviour does not relate to a single incident, but is a pattern of repeated or continuous behaviour that occurs over a period of time. It captures many of the non-physical acts of domestic abuse that are used to exert control in a relationship, as well as physical and sexual abuse within an intimate or family setting.

The cross-government definition of domestic violence and abuse outlines controlling or coercive behaviour as follows:

• coercive behaviour is an act or a pattern of acts of assault, threats, humiliation and intimidation or other abuse that is used to harm, punish, or frighten a victim

• controlling behaviour is a range of acts designed to make a person subordinate and/or dependent by isolating them from sources of support, exploiting their resources and capacities for personal gain, depriving them of the means needed for independence, resistance and escape, and regulating their everyday behaviour

The statutory guidance (PDF, 296KB) for the controlling or coercive behaviour offence further explains that, for an offence to apply, the:

• controlling or coercive behaviour must take place “repeatedly or continuously”

• pattern of behaviour has to have a “serious effect” on the victim

• behaviour must be such that the perpetrator knows or “ought to know” that it will have a serious effect on the victim and

• perpetrator and victim have to be personally connected when the incidents took place

With this statutory guidance in mind, the current measure for non-physical abuse only partially captures elements of the new definition of controlling or coercive behaviour and may include elements that do not constitute the offence. This is because it does not:

• cover the range of behaviours suggested in the statutory guidance

• specify “repeated and continuous” nature of the behaviours experienced

• measure the “serious impact” of the behaviours on the victim

Given this, we initiated research into how we could expand the Crime Survey for England and Wales to include controlling or coercive behaviour in our estimates more accurately. It is important to note that controlling or coercive behaviour is not limited to non-physical abuse.

To reduce complexity in this early research, we substituted the non-physical abuse question with the controlling or coercive behaviour question in the split sample trial. Further research will look into the interaction between controlling or coercive behaviour and other types of abuse.

Questions to measure controlling or coercive behaviour were developed through the Office for National Statistics (ONS) Domestic Abuse Statistics Steering Group (DASSG), by drawing on both the legislative guidance and advice from members of the group who had worked closely with victims of this type of domestic abuse.

The statutory guidance considers that both the behaviour experienced, and the impact, are important in considering whether somebody has experienced controlling or coercive behaviour. In line with best practice for question design, two questions were developed to reduce complexity and the resultant cognitive burden from asking two questions within one. The first question identifies the behaviour(s) experienced and the second question asks about the impact(s) on the respondent.

The behaviours question uses a multiple response format where the respondent can select one or more behaviours that they have experienced. Unlike the current non-physical abuse question, the new controlling or coercive behaviour question only asks about experiences relating to the last 12 months. This is to reflect the advice given in the UN guidelines (PDF, 2.19MB) against asking questions relating to psychological abuse with a long period of recall.

The behaviours included largely align with those outlined in the statutory guidance. To limit respondent burden, the majority of the DASSG agreed that the number of behaviours should be reduced to nine for this initial research. The wording of the question specifies that behaviours should have occurred “repeatedly or continuously”, in line with the government definition of the offence. To reduce respondent burden, the question does not require respondents to record how frequently each behaviour was experienced.

The impact question is asked of respondents who reported having experienced one or more of the behaviours. The scale of the impacts included was designed to provide a proxy measure of the severity that these controlling or coercive behaviours had on the respondent’s everyday life. This closely aligns with the statutory guidance that “serious effects” must be experienced by the victim for the offence of controlling or coercive behaviour to apply.

During the question development, the Crime Survey for England and Wales (CSEW) contractors, Kantar Public, carried out two phases of cognitive testing of the questions with members of the public¹. This supported how we developed and refined the question wording and allowed us to check the question validity. Following this cognitive testing, there were also some outstanding suggestions for alternative wording. These may be incorporated into future development of the questions.

The questions used for this split sample trial follow. As with the current non-physical abuse question, respondents are first asked if they experienced these behaviours by a partner or ex-partner. They are then asked the same questions in relation to behaviours experienced by a family member.

Trial questions

In the last 12 months, has a partner or ex-partner ever repeatedly or continuously done any of the things listed below?

By partner we mean a boyfriend, girlfriend, husband, wife or civil partner.

Please select all that apply.

1. Unfairly controlled how much money you could have or how you spent it

2. Isolated you from your friends and family

3. Monitored your letters, phone calls, emails, texts or social media

4. Enforced rules or activities which humiliated you

5. Controlled how household work or childcare is done

6. Kept track of where you went or how you spent your time

7. Bullied or intimidated you, for example by punching walls or destroying property

8. Forced you to engage in sex or certain sexual acts against your will

9. Threatened to harm children in the household

10. None of these

11. Don’t know/can’t remember

12. Don’t wish to answer

Thinking about these actions you experienced in the last 12 months, to what extent did you suffer any of the following as a result?

A. Fear that violence would be used against you

B. Feeling unable to leave the relationship/household due to fear of coming to harm

C. Constantly living in fear which affected your day-to-day activities

D. Significant changes in routine, behaviour, or appearance to try to avoid the abuse

E. Forced to give up work, education, or volunteering due to fear of coming to harm

F. Fear that you would lose contact with your children

When answering the second part of the question, respondents are asked to consider each impact, and indicate whether they suffered it:

• very much

• quite a lot

• a little

• not at all

or they could respond with:

• don’t know/can’t remember

• don’t wish to answer

• does not apply

If the respondent indicated that they experienced at least one behaviour, the impact question would be asked. The respondent is not asked this question for each of the behaviour(s) they indicated they experienced, but collectively for all.

In April 2017, we introduced a split-sample experiment on the CSEW to assess whether the questions were fit for purpose and support how we would define a victim. For half of respondents, the new controlling or coercive behaviour questions replaced the original non-physical abuse questions at the same point in the survey. The other half of respondents received the original non-physical abuse questions. Respondents were randomly assigned to the two samples and were asked all other questions in the domestic abuse, sexual assault and stalking module. Appendix Table 1 shows the sex and age profile of the samples.

All analysis was conducted on data from 16- to 59-year-olds² and all data cover the first year of the trial, April 2017 to March 2018. Throughout this analysis, we acknowledge that, although these sensitive questions were asked in the self-completion module, there are limitations to a survey approach in whether victims of such experiences would complete this section at all.

Notes for: Developing the controlling or coercive behaviour questions

1. A summary of this research can be found in Annex B.

2. Respondents eligible for the self-completion module were expanded in April 2017 from adults aged 16 to 59 years to adults aged 16 to 74 years. Due to having only one year’s worth of data on this new age group, all analysis has been conducted on 16- to 59-year-olds to allow for comparisons over time.

For the existing measure of non-physical abuse, and all types of domestic abuse, an individual is identified as a victim if they experienced one of the behaviours.

For this initial research, we developed a first approach to defining a victim. As part of this, we assessed the responses to the controlling or coercive behaviour questions against the following research questions:

• are all the behaviours in the question appropriate for defining a victim of controlling or coercive behaviour?

• should a respondent have experienced multiple behaviours to be defined as a victim of controlling or coercive behaviour?

• are all the impacts in the question appropriate for defining a victim of controlling or coercive behaviour?

• is it appropriate to consider all impacts equally when defining a victim of controlling or coercive behaviour?

The analysis was conducted on unweighted data¹ and is summarised in this section.

Behaviours

Are all the behaviours in the question appropriate for defining a victim of controlling or coercive behaviour?

To provide an indication of whether all behaviours should be included, we identified whether all behaviours were relevant and which were the most and least likely to be experienced (Appendix Tables 2a and 2b). The analysis showed that:

• each of the nine behaviours were experienced to some degree

• there was some variation in the proportion of respondents reporting each behaviour for those experienced by a partner or ex-partner (ranging from 0.2% to 2.3%)

• the responses to behaviours by a family member were lower, with less variation in the proportion of respondents reporting each behaviour (ranging from 0.1% to 0.8%)

Should a respondent have experienced multiple behaviours to be defined as a victim of controlling or coercive behaviour?

As all behaviours feature within the statutory guidance, and they were all reported to some degree, we thought that all nine should be included in the victim definition. This was supported by analysis into whether respondents were more likely to experience only one type of behaviour or multiple behaviours. We also examined whether this differed between males and females (Appendix Table 3).

For respondents who experienced at least one type of behaviour, a large proportion did not experience multiple behaviours (with 48.9% experiencing a single behaviour when the behaviour was by a partner or ex-partner and 64.1% when by a family member).

As a large proportion of respondents experienced only one type of behaviour, this supported that the victim definition should cover “at least one” behaviour experienced. This is consistent with the statutory guidance, which does not specify that a number of different, or specific, behaviours, should be experienced for the offence to apply.

It is possible that future research could explore whether additional behaviours (such as those indicated in the statutory guidance and cognitive testing) should also be included.

Impacts

Are all the impacts in the question appropriate for defining a victim of controlling or coercive behaviour?

To answer this question, we analysed which impacts had the highest number of responses that the impact “does not apply” (Appendix Tables 4a and 4b).

Two impacts elicited a higher proportion of “does not apply” responses than others, for behaviours experienced both by a partner or ex-partner and by a family member:

“Forced to give up work, education or volunteering due to fear of coming to harm”

“Fear that you would lose contact with your children”

From this, we concluded that these two impacts should be removed from the definition of a victim for this initial stage of the research.

Responses to “Fear that you would lose contact with your children” impacting the respondent “to some extent”² were marginally higher in relation to behaviours experienced among men by a partner or ex-partner. There was concern among some members of the Office for National Statistics (ONS) Domestic Abuse Statistics Steering Group that this impact was likely to illicit a relatively high response among men, which may not truly reflect controlling or coercive behaviour but rather a relatively common outcome in dissolved relationships. As such, at this early stage of the research, we took the decision not to include this impact in the definition of a victim. This is something we will investigate in later stages of the research.

From this, the following impacts were included in the definition of a victim:

1. Fear that violence would be used against you

2. Feeling unable to leave the relationship/household due to fear of coming to harm

3. Constantly living in fear which affected your day-to-day activities

4. Significant changes in routine, behaviour, or appearance to try to avoid the abuse

Is it appropriate to consider all impacts equally when defining a victim of controlling or coercive behaviour?

When considering whether all impacts should be treated equally when defining a victim, we primarily referred to the statutory guidance. As the gravity of each impact listed varies, we decided the severity with which a respondent should have experienced these impacts should reflect this.

Of the four impacts included, where “fear” is referenced, a respondent is classified as a victim if they answered that they were affected “very much”, “quite a lot” or “a little”. This reflects the statutory guidance, which states that any level of fear³ experienced would be considered serious under the offence.

For the remaining impact, a respondent is defined as a victim only if they suffer the impact “very much”. This high threshold aligns with the statutory guidance, which requires that any impact on the day-to-day activities should be substantial for the offence to apply. Individuals may often alter their behaviour or routine due to unhappy (but not abusive) relationships with partners and/or family members, however, these experiences would not constitute controlling or coercive behaviour as outlined by law. To reflect this, and to reduce the risk of overestimating the number of victims, we have restricted this impact to the most severe option on the scale.

In conclusion, we defined a victim of controlling or coercive behaviour as:

1. Somebody who has experienced at least one of the behaviours

and

2a. Said they were affected either very much, quite a lot, or a little by any of the first three impacts:

1. Fear that violence would be used against you

2. Feeling unable to leave the relationship/household due to fear of coming to harm

3. Constantly living in fear which affected your day-to-day activities

or

2b. were affected very much by the fourth impact:

4. Significant changes in routine, behaviour, or appearance to try to avoid the abuse

This victim definition reflects the statutory guidance by placing a greater emphasis on the impacts suffered than the behaviours themselves. Although the remaining two impacts in the question did not form part of the victim definition, they have been included in some of the further analysis in following sections. As this work progresses, changes may be made to this victim definition and new questions may be developed. Before finalising an approach, further research will be conducted.

Notes for: Defining a victim of controlling or coercive behaviour

1. As data were not weighted to the population, significance testing was not carried out.

2. “To some extent” covers respondents who suffered the impact “very much”, “quite a lot” or “a little”. Data for this can be seen in Appendix Tables 4a and 4b.

3. The word “fear” is used differently in “fear that you would lose contact with your children” in that it is not connected to a fear of violence or a fear of coming to harm, therefore this consideration does not apply. Future research will consider careful use of the word “fear”.

In order to assess whether the questions were fit for purpose as a measure of the controlling or coercive behaviour element of domestic abuse, we assessed them against the following questions:

• how do the prevalence rates of victims identified by the two measures compare?

• is there a consistent relationship between different behaviours experienced and their impacts?

• were any differences between how males and females responded to the controlling or coercive behaviour impacts¹ as expected?

• how did responses to the controlling or coercive behaviour questions compare with similar questions found elsewhere in the survey?

Analysis was conducted on one year of the split-sample data from the Crime Survey for England and Wales (CSEW), using data for 16- to 59-year-olds². We used standard statistical tests to determine whether differences in prevalence between the two samples were statistically significant.

How do the prevalence rates of victims identified by the two measures compare?

We produced prevalence rates for controlling or coercive behaviour and compared these with the prevalence rates for non-physical abuse (Appendix Table 5).

The prevalence of controlling or coercive behaviour estimated using the new questions is significantly lower than the prevalence of non-physical abuse estimated using the existing questions. This is true where the perpetrator is either a partner or ex-partner (1.7% compared with 3.0%) or a family member (0.6% compared with 1.3%).

As such, the prevalence of overall domestic abuse was lower for the controlling or coercive behaviour sample compared with the non-physical abuse sample (4.5% compared with 6.1%) as shown in Figure 1.

It is important to note that, for this prevalence rate analysis, respondents who reported that they “don’t wish to answer” (Appendix Table 6) are excluded from the numerator (the number of victims). As such, we cannot conclude that any differences in the prevalence rates are due to how many respondents “don’t wish to answer” the two measures.

Despite this, only a small percentage of respondents reported that they “don’t wish to answer” the two questions. This is true in relation to both behaviours by a partner or ex-partner and by a family member. There is little difference between the two measures, suggesting that the more direct and specific wording in the controlling or coercive behaviour question does not contribute to a lower willingness to reply to this question compared with the non-physical abuse question.

It is possible that those who are reluctant to respond to these questions may be more likely to say they had not experienced a behaviour at all, rather than saying “don’t wish to answer”. Future analysis could investigate the number of respondents who provided an answer to the question prior to the controlling or coercive behaviour or non-physical abuse questions, who then did not go on to respond to the two measures at all.

The lower prevalence of victims of controlling or coercive behaviour compared with non-physical abuse is consistent with our expectations. Despite this, due to the complexity of the controlling or coercive behaviour offence and the definitional differences between the two measures, further research is required to understand whether the extent of this difference is accurate.

The difference between men and women in the prevalence of controlling or coercive behaviour by a family member is similar to that of non-physical abuse by a family member (Appendix Table 5). For non-physical abuse by a family member, the prevalence of female victims was 0.2³ percentage points higher than the prevalence of male victims (1.5% compared with 1.2%). For controlling or coercive behaviour, the prevalence of female victims was 0.5 percentage points higher than the prevalence of male victims (0.8% compared with 0.4%).

Contrary to this, when considering these measures in relation to a partner or ex-partner, the differences between men and women are greater (Figure 2). Non-physical abuse by a partner or ex-partner is 3.0 percentage points higher for women than for men (4.5% compared with 1.5%). For controlling or coercive behaviour, although we see the prevalence decrease for both sexes, the difference between men and women is smaller. The prevalence for women was 1.1 percentage points³ higher than for men (2.2% compared with 1.2%).

The same patterns were also seen when considering the prevalence of domestic abuse estimates, when calculated using the two measures (Appendix Table 7).

We expected that this difference between the sexes would vary between the two questions based on expert guidance from the Office for National Statistics (ONS) Domestic Abuse Statistics Steering Group (DASSG). However, some of the group suggested that the sex difference would be larger for controlling or coercive behaviour than for non-physical abuse. Our findings are not consistent with this expectation. We cannot be sure whether the extent and direction of the difference shown by the analysis is a true finding or if they are an artefact of the questions.

Is there a consistent relationship between different behaviours experienced and their impacts?

The question asks the respondent to indicate which impacts they suffered in relation to all behaviours reported (not each behaviour in turn, if more than one was indicated). Due to this, it is difficult to ascertain specific relationships between individual behaviours and impacts.

We have conducted some preliminary analysis identifying the impacts suffered where only one behaviour was reported by a respondent. This isolates the relationship between that behaviour and the impact(s) suffered. The numbers of respondents in this analysis are very small, however, an early review of the findings suggests that different impacts appear to be more applicable to some behaviours than others. Further analysis on additional data collected from the second year of this split-sample trial (April 2018 to March 2019) will enable this to be explored in more detail.

It would be interesting to investigate behaviours where the impact had a high proportion of “not at all” responses. Findings from this could answer whether:

• respondents consider all behaviours included in the question as controlling or coercive (as opposed to some being considered normal within domestic relationships)

• the behaviour would be considered controlling or coercive by a respondent, but the language used does not have sufficiently negative connotations such as “bullied” or “forced” to be considered to have an impact

Were any differences between how males and females responded to the controlling or coercive behaviour impacts as expected?

Given the sex differences in the controlling or coercive behaviour prevalence rates, we produced a sex breakdown for each impact and the extent to which it was suffered (Appendix Table 8). Taking into account that the controlling or coercive behaviour sample was made up of 46.3% men and 53.7% women (Appendix Table 1), the findings show that:

• for most impacts, a higher proportion of women than men responded that they suffered the impact “very much”

• the only impact that was suffered “very much” by a higher proportion of men than women was “fear that you would lose contact with your children” (52.0% compared with 48.0%)

• for the impact “fear that you would lose contact with children” over two-thirds of respondents who said “not at all” were women (68.6%) and 62.8% said “does not apply”

Some elements of this analysis are consistent with expectations from the DASSG, namely that more males than females experienced the fear that they would lose contact with their children.

With the exception of this, from the current analysis (and higher prevalence estimates for women seen in Appendix Table 5) there is a concern that the current impacts, and their wording, could disproportionately apply to female experiences of controlling or coercive behaviour.

It is possible that the impacts experienced by men have not been well-reflected in the trial question, leading to an under-representation of male victims in the trial statistics. However, research suggests that when controlling or coercive behaviour is taken into account, the differences between the experiences of male and female victims become more apparent, with women experiencing greater negative impacts than men. Future research is required to investigate this further and to consider whether we should reword the current impacts, or include additional impacts, to address this.

How did responses to the controlling or coercive behaviour questions compare with similar questions found elsewhere in the survey?

The partner abuse module covers four main topics:

• sexual assault

• threat or force

• stalking

• non-physical abuse

In the split-sample trial, the non-physical abuse questions were replaced with controlling or coercive behaviour questions.

The following analysis focuses on how the two groups responded to some of these questions in the partner abuse module. Both between-group and within-group analysis has been conducted to identify where there was agreement between similar measures and to explore where there were differences.

Sexual assault

The findings show that there is only a small overlap between the two questions for sexual assault (Appendix Table 9). Only a third of respondents who said that a partner had repeatedly and continuously “forced them to engage in sex or sexual acts against their will” in the controlling or coercive behaviour question also experienced “any sexual assault by a partner”.

This low proportion is likely due to the controlling or coercive behaviour question being more open to interpretation than the sexual assault question, which is more clear and concise. The controlling or coercive behaviour question also specifies that the behaviour should be repeated and continuous, whereas the sexual assault question asks about experience that happened once or more.

Non-physical abuse compared with controlling or coercive behaviour questions

To make a direct comparison, we have conducted analysis on responses to the non-physical abuse question alongside responses to only the behaviours element of the controlling or coercive behaviour question. The analysis looks only at behaviours by a partner or ex-partner (Appendix Table 10).

The analysis shows that a similar percentage of women reported behaviours of non-physical abuse and controlling or coercive behaviour (5.2% and 4.7%, respectively). This relationship is different for men, where a lower proportion of men reported behaviours of non-physical abuse compared with controlling or coercive behaviour (1.8% and 4.2%, respectively). One possible explanation for this could be that men identify more closely with the specific listed behaviours in the controlling or coercive behaviour question than the more general behaviours in the non-physical abuse question.

This analysis excludes the impact questions. It therefore shows a different pattern when compared with the prevalence of victims, as shown in Figure 2. The difference seen in this analysis demonstrates the effect of including impacts in the definition of a victim and the importance of accurately doing so. Further research is required to ensure the victim definition is appropriate.

How do victims of controlling or coercive behaviour engage with follow-up partner abuse questions?

Analysis has been conducted to identify how respondents who were defined as a victim of controlling or coercive behaviour engaged with the follow-up partner abuse questions, compared with respondents who were victims of non-physical abuse. In the CSEW, the follow-up questions are asked of all respondents identified as a victim of partner abuse in any of the four areas: sexual assault, threat or force, stalking, and non-physical abuse/controlling or coercive behaviour.

The first of the partner abuse follow-up questions asks whether respondents consider their experiences abusive:

“Thinking about all the abuse you have suffered in the last 12 months, would you call what happened to you domestic abuse?”

Although we acknowledge that controlling or coercive behaviour is not limited to non-physical abuse, to directly compare how victims of controlling or coercive behaviour and victims of non-physical abuse answered this follow-up question, we analysed respondents who were solely a victim of controlling or coercive behaviour, or non-physical abuse (Appendix Tables 11a and 11b). This is to ensure that responses to the follow-up question in the analysis were not in respect of the other three offences (sexual assault, threat or force, or stalking).

The analysis showed that victims of controlling or coercive behaviour were more likely to identify that their experiences were abusive than victims of non-physical abuse (29.6% compared with 22.8%, respectively). In addition, the percentage of respondents who selected they do not wish to answer this question was much lower for victims of controlling or coercive behaviour (15.5%) than for victims of non-physical abuse (21.7%). This suggests that victims of controlling or coercive behaviour are more engaged with the follow-up question.

These findings could be due to the more detailed nature of the behaviours included in the controlling or coercive behaviour question and the additional question focusing on the impacts. The increased specificity may have increased the salience of the abuse to the respondent, leading to increased engagement with the follow-up questions and identification of their experiences as abusive.

At this stage, we are unable to draw any conclusions from any similarities or inconsistencies between the controlling or coercive behaviour questions and similar questions found elsewhere in the survey.

Notes for: Assessing whether the trial questions are fit for purpose

1. Behaviours and impacts analysis is based on all six impacts included in the question and is not limited to those impacts included in the definition of a victim.

2. Respondents eligible for the self-completion module were expanded in April 2017 from adults aged 16 to 59 years to adults aged 16 to 74 years. Due to having only one year’s worth of data on this new age group, all analysis has been conducted on 16- to 59-year-olds to allow for comparisons over time.

3. These differences are calculated from unrounded data. Numbers in Table 5 are rounded, so these differences quoted in the text will not sum to those in Appendix Table 5.

Our analysis of the controlling or coercive behaviour questions has highlighted that there is uncertainty in whether the measure adequately captures victims of the offence as outlined in the statutory guidance. As such, the questions require further development before we can be confident that they are fit for purpose. To allow for this, the questions tested in the split-sample experiment were removed from the survey in April 2019. The non-physical abuse questions and all other questions in the module will continue to be asked of the full sample whilst the development takes place.

Removing the questions at this stage does not prevent us from being able to conduct further analysis on the test questions. Additional data to those analysed in this article will be available up to March 2019 (after which the questions were removed). This will provide a larger sample for the analysis conducted previously. This research has concluded that the questions tested were not fit for purpose in their current form. As such, there is little value in including them for an extended duration, compared with the increased confidence we can have in our results by increasing the sample size for the existing questions.

Further discussions with stakeholders will support the next steps in developing the controlling or coercive behaviour questions. We also plan to revisit the suggestions generated during the cognitive testing of the trial questions and will investigate the complex relationship between other forms of abusive behaviour, which can be part of enacting controlling or coercive behaviour. Analysis on the additional year of data could also enable us to investigate question responses in more detail.

Furthermore, changes to the offence indicated by the draft Domestic Abuse Bill (PDF, 2.24MB) published in January 2019 could inform the question development. This states that the statutory guidance will be updated to include how economic abuse may manifest itself as controlling or coercive behaviour.

The Office for National Statistics (ONS) Domestic Abuse Statistics Steering Group is chaired by ONS and includes representatives from across government, academia, the voluntary sector and Kantar Public who run the Crime Survey for England and Wales.

The Group was set up in November 2015 to:

• review the questions on domestic abuse asked in the Crime Survey for England and Wales to ensure they accurately capture the data required

• review domestic abuse statistics produced and published by the ONS to ensure they meet user requirements for information on domestic abuse

Please note: where references to decisions and discussions involving the group are made throughout the article, these were made on an advisory basis and may not have been unanimous or democratically decided.

We’d like to thank members of the group for their advice and guidance during the development and analysis of the questions on controlling or coercive behaviour.

To support development of the controlling or coercive behaviour trial questions, Kantar Public (the Crime Survey for England and Wales (CSEW) survey contractors) conducted two phases of cognitive testing.

The first phase involved cognitive interviews with the general public to develop questions on controlling or coercive behaviour. Twenty interviews were conducted across two rounds of testing as part of the standard CSEW cognitive interviewing. These interviews focused on how questions would be understood by people who had not experienced any domestic abuse; as such, they were designed to understand how these questions would be approached by the majority of people taking part in the survey. Changes were made to the questionnaire between rounds.

The second phase involved cognitive interviews with victims of domestic abuse. Overall, 13 interviews were conducted, nine of which were with women and four with men. Updates to the questionnaire were made during this second phase of testing and the revised questions were tested in the later interviews.

Summary of findings

In the first round of testing, all participants clearly understood that these questions related to serious issues of abuse, intimidation, bullying and control.

A few participants had suggestions for additional behaviours and impacts that could potentially be included, such as: constant monitoring of where they went and how they spent their time, and threats to isolate them from their children. As such, these were added to the question as a behaviour and impact, respectively.

As well as these two additions, other suggestions were made. One participant said that blackmailing was something that was not included – another referenced verbal aggression. Others talked about the online facets of a potentially abusive relationship (for example, sharing inappropriate photos online without consent; spreading rumours and so on). The behaviours included in the question were designed to match the legal definition of controlling or coercive behaviour and therefore these additional suggestions were not added for round two of the testing.

During the testing, there was some discussion around the scale used to measure the severity of impacts. Although some participants (who were not victims of controlling or coercive behaviour) preferred a simpler scale, only respondents who experienced controlling or coercive behaviours would proceed to answer the impacts question. When interviewing these individuals, the more detailed answer scale used in this trial was found to be more appropriate.

Participants also raised that the wording around threats could be an area of difficulty for respondents; often in a coercive relationship, threats are not explicit but are implied, or a victim may feel threatened if previous abuse has taken place. This is something that will require focus in future research.

Based on their specific experiences, one participant suggested that it may be helpful to specify the intention behind the actions. This may be something to consider in future research, as the criminal offence specifies that the perpetrator “ought to know” the behaviours would have a “serious impact” on the victim.

Overall, in both the interviews with the general public and those with victims of domestic abuse, the questions around controlling or coercive behaviour were understood well and were generally felt to be a fairly comprehensive list of the types of behaviours that might be encountered. Participants felt that the questions, though very explicit, were reasonable and that they accurately reflected their experiences, and they were able to understand why the level of detail included in the questions was required.