Minneapolis Domestic Violence Experiment - Replication

Replication

Beginning in 1986, the National Institute of Justice sponsored six replications of the Minneapolis Domestic Violence Experiment. While each site was an independent study, NIJ required that each study had to 1) use an experimental design (i.e., random assignment), 2) address domestic violence incidents that come to the attention of the police, 3) use arrest as one of the treatments, and 4) measure repeat offending using official police records and interviews with victims. The study sites included police departments in Omaha, Nebraska, Charlotte, North Carolina, Milwaukee, Wisconsin, Miami-Dade County, Florida, and Colorado Springs, Colorado. In Metro-Dade, 907 cases were used, compared to 1,200 cases in Milwaukee and over 1,600 cases in Colorado Springs. A study initiated in Atlanta was never completed. Although these five studies have been described as replications of the Minneapolis experiment, they each varied considerably from that study and from each other in methods and measures.

The initial findings from the five completed replications were reported independently beginning in 1990. The original authors' findings about the crime control effects of arrest varied depending upon the site studied, the measures of repeat offending used, and which alternative treatments were compared to arrest. Each replication reported multiple findings with some results favoring arrest, some showing no differences and some showing that arrest was associated with more repeat offending. None of the replications reported effects as strong as those reported for the Minneapolis Experiment.

Two articles synthesizing the findings from these studies report a crime control effect for the use of arrest for domestic violence. First, a meta-analysis of the published findings based on official police records from the Minneapolis and the SARP experiments reported a deterrent effect for arrest. Second, a re-analysis that applied consistent measures and methods to the archived data from the five replications reported that arrest was associated with as much as a 25% reduction in repeat offending and that those results were consistent across all five sites.