He Never Hit Her. Then He Killed Her

That same research found that what most reliably distinguishes women who are killed from those who survive abuse is not the number of police calls. Firearm access was the single strongest predictor, increasing femicide risk roughly five times. Estrangement from a controlling partner was a major risk factor. Prior strangulation. Threats to kill. Forced sex. These things happen in private. They leave no paper trail.

A separate study of men convicted of killing their intimate partners found that in fully half the cases, no physical or sexual assaults against their partners had been reported in the year before the homicide. Coercive control, monitoring, isolation, and psychological domination were present across the sample, including in cases with no prior physical violence.

Key points
Prior domestic violence is a serious and documented risk factor for intimate partner homicide.
But the absence of a police record is not the absence of danger.
The factors that predict whether a woman will be killed rarely generate police reports or criminal records.
Official systems are best at detecting the type of domestic violence least likely to end in death.
A new study out of Australia made a finding that at first sounds reassuring: in a database of more than 62,000 violent incidents, cases flagged by police as domestic violence were less likely to end in homicide than other violent incidents. If you’re in a relationship where things have gotten physical, you might read that and exhale.

But the finding reflects how inconsistently police record domestic violence, not how dangerous it is. And something else; The cases most likely to end in murder are sometimes cases where official systems saw very little, or nothing at all. Two cases are frightening examples.

used with permission from iclipart
Source: used with permission from iclipart
When the Danger Is Seeable But the System Is Blind

On February 19, 2020, in the Brisbane suburb of Camp Hill, Rowan Baxter jumped into the car of his estranged wife, Hannah Clarke, as she was leaving her parents’ home on the school run. He doused the car in gasoline and set it on fire. Hannah, 31, managed to pull herself from the burning vehicle and told bystanders and first responders what had happened before dying in the hospital that evening with burns to 97 percent of her body. Their three children, Aaliyah, 6, Laianah, 4, and Trey, 3, died in the car. Baxter stabbed himself at the scene.

A Domestic Violence Protection Order had been in place. Hannah had been in contact with the police and a domestic violence caseworker. She had told the caseworker that, on one occasion, Baxter strangled her, one of the strongest predictors of lethal violence. She had told a colleague she feared Baxter was going to kill her.

And yet, a 2022 Queensland coroner’s inquest after Hannah’s death found that every agency she talked to had failed to recognize the extreme risk she faced. Deputy State Coroner Jane Bentley described Baxter as “a master of manipulation” whose murderous plans could not, in her assessment, have been stopped by any further action from police or service providers. Bentley concluded that failure “probably came about because Baxter had not been violent and had no relevant criminal history.”

In other words, he didn’t beat her. Most of his abuse was years of coercive control. Monitoring her movements. Isolating her from family. Controlling what she wore. Tracking her phone. Using the children as leverage. The system was not built to read that pattern as a lethal threat.

When the Danger Was Invisible

Three years later, Christine Banfield was killed in her home in Herndon, Virginia. Her husband, Brendan Banfield, a former IRS criminal investigation agent, had spent months arranging her murder to avoid what he calculated would be a financially devastating divorce. He created a fake profile on a fetish website, lured a stranger to the house under false pretenses, and staged the scene to look like a home invasion. He was convicted of aggravated murder on February 2, 2026.

article continues after advertisement

Christine Banfield was not isolated. She was not financially dependent. She was a pediatric ICU nurse with her own career. There is no evidence that she was afraid of her husband. She died in her sleep.

This is a different kind of danger and a different kind of invisible. Baxter’s danger was felt by everyone who loved Hannah, even if the formal system could not capture it in time. Banfield’s danger may not have been felt by anyone, including Christine, because it had nothing to do with the relationship’s dynamics.

It had to do with money, and with a man who looked at his wife and saw a problem to be solved rather than a person to be reckoned with. No pattern of escalating control. No threats to document. No history of violence. Just a long, careful plan executed by someone practiced in deception: an affair carried on for months, a fake identity constructed online, a cover story maintained through a police investigation.

What the Research Says about Who Gets Killed

Jacquelyn Campbell’s landmark 11-city study of intimate partner femicide, the foundation for the Danger Assessment tool used by law enforcement and advocates across the country, found that roughly four in five women killed by intimate partners had been physically abused or stalked by the same partner beforehand. Prior intimate partner violence is the leading risk factor for intimate partner homicide.²

But that same research found that what most reliably distinguishes women who are killed from those who survive abuse is not the number of police calls. Firearm access was the single strongest predictor, increasing femicide risk roughly five times. Estrangement from a controlling partner was a major risk factor. Prior strangulation. Threats to kill. Forced sex. These things happen in private. They leave no paper trail.

article continues after advertisement

The Danger Assessment was built for exactly this reason: because counting police calls was the wrong way to find the women most at risk. It asks about gun access, controlling behavior, strangulation, threats, and separation, the factors that predict death rather than the ones that generate incident reports.

Link to full article here.