Media Release

A new study by the NSW Bureau of Crime Statistics and Research (BOSCAR) that looked at the prevalence of coercive control behaviours in police domestic violence (DV) reports has found that in 57% of domestic violence (DV) events at least one coercive control behaviour was recorded by police. The most common coercive control behaviours were property damage and theft (present in 26% of DV events), intimidation and threats (24%) and verbal abuse (23%).

10% of recorded DV offences included a reference to threats of harm and 6% included a threat to kill.

Executive Director of the NSW Bureau of Crime Statistics and Research, Jackie Fitzgerald, says the study used text mining to analyse the narrative description of police DV reports looking for mention of coercive control behaviours.

“Coercive control relates to abusive behaviours which can exert domination and control over another person. These behaviours, which can include threats, financial control, social-isolation and surveillance, represent a growing awareness of the breadth of domestic violence behaviours,” Ms Fitzgerald said.

Emotional abuse and stalking by a current or previous partner affects a large proportion of Australian women. It is estimated that in 2021-22, 23% of Australian women have experienced emotional abuse since the age of 15 (including controlling or threatening behaviours, incessant insults and intimidation by a current or previous partner), while 20% have experienced stalking (including following/watching the person, maintaining unwanted contact and using social media or electronic devices to follow or track the person).

The latest DV assault offence trends in NSW to June 2023 show that police recorded incidents of DV assault continue to increase significantly.

“Over the five years to June 2023 the number of recorded DV Assault incidents in NSW increased by 13.5%. Domestic assault and sexual assault are the only major offences to show sustained increases over this time,” Ms Fitzgerald said.

Summary: Text mining police narratives of domestic violence events to identify coercive control behaviours

Key Analysis

A new BOCSAR study developed a text mining method to identify coercive control related behaviours from the free-text descriptions in police narratives of 526,787 domestic violence related events occurring between 1 January 2009 and 21 March 2020. This provided a measure of the prevalence of 48 separate behaviours.

The study found that in 57% of domestic violence events at least one coercive control behaviour was reported, with property damage and theft (26%), intimidation and threats (24%) and verbal abuse (23%) being the three most common behaviours. In 8% of events, three or more distinct subcategories of coercive control behaviours were detected. Other forms of coercive control, such as technological abuse, were detected but much less frequently.

For the four behaviours that were captured by both the police incident categories and the text mining system (stalking and intimidation, use carriage service to menace/harass/offend, property damage and trespass), there was substantial overlap with police offence classifications from fixed fields. However, the results suggest that, for all four of the categories examined, many more additional behaviours would be detected if the text mining measure was used in conjunction with the fixed field incident category recorded by police (between 30% and 60% more incidents of coercive control).

Notably, the study found that the coercive control measures provided no improvement in prediction over and above the demographic, offence characteristic and prior offending variables that are typically used to predict domestic violence reoffending.

This study highlights the inherent difficulty in defining, operationalising, and capturing coercive control. These challenges will also apply to any evaluation of the impact of the new coercive control legislation in NSW. If more than half of DV events that come to the attention of the police have at least one coercive control related behaviour reported (as estimated in this study), it is possible that a large majority of persons of interest identified in relation to DV events could potentially be charged with this new offence. Given the difficulties observing this offence, the way in which police record and enforce the offence (and the way in which victims report the behaviour) will largely determine how often the offence is used and the subsequent impact.