Public Awareness

Bulletins - Winter 2010 OPDV Bulletin

No Crystal Ball: Predicting Reabuse and Murder of Intimate Partners

Judges, police, victim advocates and others who deal with cases of intimate partner violence live with the fear that they will miss signs of extreme danger and a case they have a role in will end in tragedy. They hope for a risk assessment instrument that will allow them to try to prevent such dire outcomes. This article will consider what we know, what we can’t know, and whether risk assessment instruments can help.

What we know

Past behavior is the best predictor of future behavior. Roughly 30% of men arrested for a domestic violence offense will be rearrested for the same offense within a few years; 60% will reoffend within nine years.1 An abuser’s criminal history is highly predictive of repeat abuse. Younger men are more likely to reoffend. Those who have more to lose – employment, marriage, home ownership, social status – are more likely to desist to avoid consequences such as rearrest, job loss or divorce.2 (But there have to be consequences to achieve this effect.)

Some characteristics distinguish cases in which men have murdered their intimate partners from those in which they have inflicted non-lethal abuse: job loss increased the probability that an abuser murdered his partner by a factor of four; access to a gun increased it by a factor of five; and having used a gun against the victim increased the probability of lethal assault 41 times.3 Note that some men with none of these characteristics have murdered their partners; others with these characteristics have not.

What we can't know
There are inherent limitations on our ability to predict – and therefore prevent – bad outcomes. Actions taken by the victim and the criminal justice system may disrupt the potential for re-assault, making predictions less accurate. Similarly, a change in an offender’s circumstances (a new job; learning that his ex-partner is engaged; experiencing combat; drug rehabilitation) may increase or decrease his motivation and capacity to inflict abuse, especially potentially lethal assault. Our predictions are situation-specific and short-lived.

Intimate partner homicides are particularly difficult to predict because murders, though too common, are statistically rare. Rare events are unpredictable. Richard Gelles points out that there are 1,200 intimate partner femicides in the US annually out of about 90,000,000 heterosexual couples; murder is always unlikely.4 We can project that if a man is unemployed, has previously abused his partner and threatened her with a gun, he is more likely to kill his partner than anyone else, but it does not mean that he will. Gelles notes that analysis of past cases provides explanation, not prediction. Only hindsight is 20:20.

Can risk assessment instruments help us?

Some instruments are more accurate than victims’ predictions. The Danger Assessment (DA), designed for use with victims to predict lethality, and the Ontario Domestic Abuse Risk Assessment (ODARA), designed for use by law enforcement to predict reoffending, have achieved 70% accuracy in predicting severe reassaults and new offenses, respectively. 5, 6 They err on the side of caution; in at least 40% of cases categorized as high risk, the offender did not reassault his partner. Victims, conversely, are more likely to underestimate risk.

Whether and how instruments should be used depends on the context and purpose of the assessment. The DA might be used by victim advocates to train staff, help clients assess their safety, and communicate with police and prosecutors. The ODARA might be more useful in criminal justice settings because it relies on information on the offender’s criminal history. It is questionable whether courts should rely on instruments that cast too wide a net.

Reducing risk

There is no infallible method of prediction. Still, there are many reliable indicators that suggest preventive measures. When possible, domestic violence offenders should be required to surrender firearms. Decisions about prosecution, pretrial release and sentencing should take into account an offender’s criminal history, including drug and property crimes. Courts should impose the maximum penalty available and enforce court orders. Such an approach does not require risk assessment and has been found to decrease revictimization.7

Arrest for a domestic violence offense has even been associated with a decreased likelihood of homicide. Risk assessment instruments may be helpful but not necessary to gathering relevant information. The key lies in the response.

Klein, A. R. (2009). Practical Implications of Domestic Violence Research: For Law Enforcement, Prosecutors and Judges. NCJ22572. US DOJ, NIJ.
Feder, L., & Forde, D. R. (2000). A test of the efficacy of court-mandated counseling for domestic violence offenders: The Broward Experiment. NIJ-96-WT-NX-0008. Washington, DC: NIJ.
Campbell, J., D. Webster, J. Koziol-McLain, C. Block, D. Campbell, M. Curry, F. Gary, J. McFarlane, C. Sachs, P. Sharps, Y. Ulrich, and S. Wilt. “Assessing Risk Factors for Intimate Partner Homicide.” NIJ Journal 250 (November 2003): 14-19, NCJ 196547. http://www.ncjrs.gov/App/Publications/abstract.aspx?ID=196547
Men who Murder their Families: NIJ-Sponsored Symposium, US DOJ, June 2, 2009, Richard Gelles comment.
Campbell, J., O’Sullivan, C., Roehl, J. & Webster, D. Intimate Partner Violence Risk Assessment Validation Study. (2005). NCJ 209731. http://www.ncjrs.gov/App/Publications/abstract.aspx?ID=209731
Hilton, N., G. Harris, M. Rice, C. Lang, C. Cormier, and K. Lines. “A Brief Actuarial Assessment for the Prediction of Wife Assault Recidivism: The Ontario Domestic Assault Risk Assessment.” Psychological Assessment 16(3) (September 2004): 267-275.
Harrell, A., Castro, J., Newmark, L. & Visher, C. (2007) Final Report on the Evaluation of the Judicial Oversight Demonstration: Executive Summary. Washington, DC: US DOJ, NIJ, and The Urban Institute. NCJ 219386, http://www.urban.org/publications/411498.html