WEEK 2 (B)

Assigned Reading: Thistlethwaite & Wooldredge, Part 1 Chapter 2- The Minneapolis Domestic Violence Experiment: Does arrest deter crime? pp. 40-47

After reading the article answer the following questions in a discussion posting:

1. Do you think the methodology used to complete this study was valid?

2. What limitations do you see in the way the study was conducted?

3. Based on your knowledge and experience do you think proactive domestic violence arrest policies deter future acts of domestic violence? Why or Why not?

the minneaPolis Domestic violence exPeRiment: Does aRRest DeteR cRime? Sherman, L., and R. Berk (1984).

“The Specific Deterrent Effects of Arrest for Domestic Assault.” American Sociological Review 49:261–72. Background Domestic violence is the most common form of violence encountered by the police. More women are injured by spouses or partners than by any other person. Despite this fact, the traditional police response used to be nonintervention. The public perceived domestic violence as a “private” matter and the police were reluctant to get involved formally by making an arrest. Police offi-cers would utilize their discretion by either separating the parties (having the suspect leave for a “ cooling-off” period) or would attempt some type of counseling to help mediate the situation. Interest in protecting the victims of domestic violence increased in the 1970s as battered wom-en’s groups and victims’ rights groups organized to provide shelters and legal assistance (Pleck 1989). These same groups were also responsible for calling attention to the lack of formal police response. From 1970 to 1983, 36 states and the District of Columbia passed laws to modify offi-cial responses to domestic violence (Buzawa and Buzawa 1985). Several states passed laws per-mitting an officer to make a warrantless arrest if the officer had probable cause to believe that a protection order had been violated. Other states imposed additional responsibilities upon officers such as remaining with the victim until she was out of danger or taking the victim to a shelter. In spite of these legislative changes, the police remained largely apathetic toward domestic violence and it was considered a low priority (Sherman et al. 1992, Zorza 1992). Most police departments had explicit nonarrest policies and officers were instructed to discourage victims from pressing charges. The most common police responses were to order the perpetrator to leave for a cooling-off period or attempt mediation. These actions often led to an escalation of vio-lence. According to a study on spousal homicides conducted by the police foundation in 1977, 85 percent of the sample had called for police assistance at least once in the preceding 2 years and 54 percent had called five or more times (Sherman et al. 1992). There was also an increase in liti-gation during the 1970s. Several lawsuits were filed against police departments for their failure to arrest when the police had probable cause that an assault had occurred. The threat of liability was enough for many police departments to implement voluntary changes in their domestic violence policies; however, most departments still did not allow their officers to make a warrantless arrest in cases where there was no apparent physical injury. In 1982, only five states required the police to arrest if the officer had probable cause that the assault had taken place (Zorza 1992). the experiment The most significant change in police response to domestic violence can be attributed to an experiment conducted by Lawrence Sherman and Richard Berk (1984) with the assistance of the Police Foundation and the Minneapolis Police Department. Beginning in 1978, police officers in Minneapolis were allowed to arrest in cases of misdemeanor domestic violence but it was left up to the officer’s discretion to arrest, separate the parties, or attempt mediation. The experiment began on March 17, 1981, and lasted until August 1, 1982. The purpose of the experiment was to determine which of three police responses (arrest, separation involving an 8 hour cooling-off period, or mediation) produced the lowest recidivism rates for offenders involved in misde-meanor domestic violence. The study involved an experimental design in which the responses were randomly selected prior to each shift. Police officers were given color-coded report forms and were instructed to respond to each case of misdemeanor domestic violence according to the predetermined response. Encounters involving felony domestic assault (with physical injuries) were excluded from the study because officers were required to arrest in those cases. The random selection of responses enabled the researchers to control for any preexisting differences between offenders so that the exact influence of each response could be determined. Research assistants were assigned to ride along with the officers during a sample of shifts to make sure that the officers complied with the random selection process. Participation by the officers was voluntary. During the study period, 314 cases of misdemeanor domestic violence were reported. The experiment

 was conducted in two of the city’s precincts with the highest concentrations of domestic violence to yield a large number of cases in a short period of time. Two measures of recidivism were exam-ined: official police reports indicating whether or not a suspect was rearrested, and follow-up interviews with the victims indicating any repeat occurrence. The victim interview data were collected to address one of the major problems with official measures of crime: the fact that many crimes go unreported to the police. Victims are often reluctant to report abuse for reasons that include embarrassment, fear of retaliation, economic concerns, or the perception that the police will not help. Victimization studies often reveal a higher incidence of domestic violence compared to police reports and the researchers wanted to accurately determine the extent of repeat violence. The experiment raised a potential ethical concern. Police officers were responding to cases of domestic violence differently. The responses were predetermined which meant that the police would not be permitted to use their discretion in choosing perhaps the most appropriate response. For example, if a police officer arrived on the scene with the predetermined response of coun-seling but felt that separation would be a better response after observing interactions between the offender and the victim, the officer was supposed to only counsel. The separation may have prevented violence in the future whereas counseling might not. This issue frequently arises in the field of medicine with experimental trials designed to determine the effectiveness of new medi-cations or treatments. Doctors must randomly assign patients to either the experimental group (patients who receive the treatment) or a control group (patients receiving a placebo). The ethical problem is that some patients may not receive treatment that makes them better. With the domes-tic violence experiment, some offenders may not have received the response that might have pre-vented future violence. This issue was addressed by Sherman who argued that randomly assigning suspects to one of three dispositions was appropriate because at the time there was no evidence to suggest one response was better than the other two (Sherman et al. 1992). In this case, random assignment of dispositions was no different than the police officers’ professional judgment. Results Both the official measures and victim interviews revealed that the arrested suspects were less likely to recidivate compared to offenders that were separated or counseled. It appeared as if by arresting a suspect, the arrest served as a deterrent for future violence. The results are presented in the graph on the next page. Based on their conclusions, the researchers made three recommendations. First, it was sug-gested that all states adopt policies permitting a police officer to make warrantless arrests for domestic violence. Second, although the researchers cautioned against a policy of mandatory arrest, they did contend “ . . . an arrest should be made unless there are good, clear reasons why an arrest would be counterproductive.” (Sherman and Berk 1984, p. 270). Finally, the researchers recommended that the study be replicated in other locations across the United States. criticisms Experimental designs are considered the ideal evaluation research design, yet experimental research involving human subjects outside of a controlled setting is rare in criminal justice. The Minneapolis Domestic Violence Experiment remains one of the best examples of this type of research. The researchers, however, ran into a problem inherent in field experiments: the dif-ficulty of ensuring random selection to the treatment groups. With any experiment it is impor-tant to establish internal validity. Internal validity means that the researcher was able to control for all extraneous influences on the outcome aside from the treatment effects. Unfortunately, the

Percentage of Repeat Domestic Violence 40 37 33 30 24 20 10 10 19 19 Police Data Victim Data 43 0 Arrest Counsel Separate Police Response figURe 1.1 Minneapolis Domestic Violence Recidivism Minneapolis police chief determined that certain encounters between the police and a domestic violence suspect should end in an arrest regardless of what the color-coded card indicated. For example, the police were instructed to arrest based on the preference of the victim or if the sus-pect physically attacked the officer. There was also the possibility that some officers may have just refused to follow the cards. The discrepancy in random assignment could have led to certain types of suspects being arrested more often, making it difficult to determine the true effect of each police response (Berk et al. 1988). Even so, eighty-two percent of the suspects did receive the proscribed treatment. Reasons for not following the predetermined responses included many of the exceptions mentioned above. In cases where an officer was assaulted or the victim requested arrest, an arrest was made (Berk et al. 1988). Collecting the official data on recidivism was easy compared to the victim interview data. Many of the victims could not be located or refused to cooperate with researchers. Only 62 per-cent of the victims were interviewed, which raises the risk of selection bias if victim characteristics influenced who participated in the interviews and who did not. The researchers were able to examine victim characteristics and a thorough examination of the data revealed no significant dif-ferences between the victim participants and nonparticipants (Sherman and Berk 1984). Critics also took issue with the implication that arrest is the best response simply because it appeared to “work” better than separating or counseling (Binder and Meeker 1988). The Minneapolis study only examined three police interventions, yet the influence on police policy was significant. Domestic violence is a social issue that many believe requires a much broader response than simply arresting offenders. The biggest challenge to the experiment’s results came from the replication studies. Replication is crucial to the generalizability of an experiment’s results. Establishing that the find-ings would be consistent across different police agencies in different cities was important for lending credibility to Sherman and Berk’s findings. The experiment was replicated 6 times in five different cities: Omaha, Charlotte, Milwaukee, Colorado Springs, and Miami. All of these replica-tion studies utilized an experimental design and were close approximations (although not exact) to the original Minneapolis study. Only two of the replication cities produced findings consistent with Minneapolis. In Colorado Springs and Miami, the arrested offenders had the lowest inci-dence of repeat domestic violence, but data from the other three cities revealed that the arrested suspects had the highest occurrence of domestic violence (Sherman et al. 1992). exPlaining the inconsistent ResUlts The relationship between arrest and recidivism for domestic violence was obviously more complex than originally thought. It was possible that the inconsistent findings were a result of treatment and/or methodological differences between stud-ies. The arrested suspects across cities spent different amounts of time in jail following their arrest (Garner et al. 1995, Sherman et al. 1992). In some of the cities, the police (as in Minneapolis) followed color-coded cards to ensure random assignment, in other cities a police dispatcher or a civilian made the assignment (Sherman et al. 1992). One alternative explanation provided by the researchers was that arrest only reduced domestic violence among suspects who had something to lose if arrested—suspects with a stake in conformity. A subgroup analysis of the data from each of the cities revealed that arrested suspects who were employed were less likely to commit domes-tic violence in the future, but the unemployed arrested suspects were more likely to recidivate. A similar finding was also found for suspects who were married at the time of the arrest (Pate et al. 1991). Employment and marital status appeared to have an influence over the effectiveness of an arrest. A second explanation for the different findings across studies involved what actually hap-pened to the suspects after arrest. Not all arrested suspects were prosecuted and convicted, and convicted offenders were subjected to different forms of punishment (jail sentences versus pro-bation versus counseling). The Minneapolis Domestic Violence Experiment and the replication studies did not examine the outcomes following arrest. significance and subsequent Research Sherman subsequently published a book titled Policing Domestic Violence in 1992 in which he included his original Minneapolis findings and results from the replication studies. A study of the most cited works in criminal justice put his book at the top of the list for 1995 (Cohn and Farrington 2012), and policymakers took notice. Dramatic changes in domestic violence policy occurred following the release of the experiment’s results. By 1989, 84 percent of police agencies in large urban areas adopted policies favoring arrest, and 76 percent imposed a mandatory arrest policy (Sherman 1992). The 1994 Crime Bill advocated the adoption of mandatory arrest policies throughout the United States as well. The Minneapolis Domestic Violence Experiment was the first controlled experiment to examine the effectiveness of arrest. The results of the study along with the replication research increased our appreciation of the connection between research and policymaking. Domestic violence was no longer perceived to be a private matter but a social issue deserving official action. Arrest remained the predominate police response in domestic violence situations, but the failure of arrest to work for all offenders has led to the investigation of other police and court practices that have been shown to be effective in reducing domestic violence. According to Lawrence Sherman, “ . . . . the lesson of the Minneapolis Domestic Violence Experiment and its replications is that one size is not likely to fit all cases, regardless of the public appetite for simple solutions” (p.352). David Ford and Mary Regoli (1993) conducted an experiment in the early 1990s to determine the effectiveness of prosecution on domestic violence recidivism in Indianapolis. Prosecution did not reduce the likelihood of repeat violence in general, but there was an impor-tant exception worth mentioning. Cases that were initiated by the victim and involved “severe violence” did benefit from prosecution. The study also found that most (90 percent) of the offenders arrested for misdemeanor domestic violence were never charged or had their charges dropped. Prosecutors dismissed cases for a variety of reasons, including insufficient or improp-erly obtained evidence, or at the request of the victim. Dropping charges at the request of the victim was especially problematic in misdemeanor domestic violence cases because often there was no physical evidence. Prosecutors were left to build their cases on the testimony provided by victims. Several jurisdictions subsequently adopted “no-drop” policies with regard to prosecuting domestic violence. Regardless of whether or not the victim cooperated with prosecutors, pros-ecutors were encouraged to pursue charges. The implementation of these policies varied signifi-cantly across jurisdictions. Ford and Regoli’s experiment compared two types of no-drop policies. With the “soft” no-drop policy, victims had input into whether or not charges could be dropped against their perpetrator, but only after the abuser had appeared before a judge. With the “hard” drop policy, victims were informed that once charges were filed, they could not be dropped at the request of the victim. The study found that allowing the victim to drop charges actually resulted in less recurrent domestic violence compared to the stricter no-drop policy. Research has also been conducted examining differences in whether or not domestic vio-lence defendants are convicted, participate in some type of diversionary treatment, or receive other sanctions. Ford and Regoli (1993) found no difference in the Indianapolis data between arrestees randomly assigned to pretrial diversion with counseling, probation with counseling, or a combi-nation of fines, probation, and/or jail. Research on the effectiveness of court-mandated batterer intervention programs has produced mixed results. One study conducted by Lynette Feder and Laura Dugan (2002) involved the random assignment of 404 convicted domestic violence offend-ers to either probation or probation with weekly counseling sessions. There was no significant difference in recidivism rates between the two groups. Research by Robert Davis et al. (2000), how-ever, found that participants who had completed a 6-month treatment program recidivated less compared to offenders who only completed an 8-week program or received community service. More recently, researchers conducted a randomized experiment examining the effectiveness of a “second responders” program in California. Second responders are teams of police officers who visit domestic violence victims after a complaint of abuse has been filed. Their goal is to provide victims with information on available legal and social services. In the experiment, complaints were randomly assigned to three groups: those who received a second responder’s visit within 24 hours, those who received the visit within one week, and those where no visit was made (Davis et al. 2010). The experiment found the second response program to be ineffective in preventing future violence and even suggested that the program may have escalated violence (Davis et al. 2010). Several jurisdictions have implemented “specialized” courts for processing domestic assault offenders and to provide greater assistance to victims. There are over 208 such courts in operation in the US.2 Victims today have greater access to court enforced protection orders, shelters, and other resources designed to help them separate from their abusive partners. Offenders are pro-vided greater opportunities to participate in treatment programs. Policymakers now have access to an enormous body of research that explores the causes and correlates of domestic violence and the effectiveness of criminal justice responses to the problem. Furthermore, the Minneapolis Domestic Violence Experiment provided important insight into how police officers use their dis-cretion and the outcomes of policies designed to reduce police discretion.

Berk, R., G. Smyth, and L. Sherman (1988). “When Random Assignment Fails: Some Lessons from the Minneapolis Spouse Abuse Experiment.” Journal of Quantitative Criminology 4:209–23. Binder, A., and J. Meeker (1988). “Experiments as Reforms.” Journal of Criminal Justice 16:347–58. Buzawa, E., and C. Buzawa (1985). “Legislative Trends in the Criminal Justice Response to Domestic Violence.” In Crime and the Family, edited by A. Lincoln and M. Straus, 134–47. Springfield: Charles C. Thomas. Cohn, E., and D. Farrington (2012). “Scholarly Influence in Criminology and Criminal Justice Journals in 1990–2005.” Criminal Justice Review 37:360–83. Davis, R., B. Taylor, and C. Maxell (2000). Does Batterer Treatment Reduce Violence? A Randomized Experiment in Brooklyn. Washington, DC: National Institute of Justice. Davis, R., D. Weisburd, and E. Hamilton (2010). “Preventing Repeat Incident so Family Violence: A Randomized Field Test of a Second Responder Program.” Journal of Experimental Criminology 6:397–418. Feder, L., and L. Dugan (2002). “A Test of the Efficacy of Court-Mandated Counseling for Domestic Violence Offenders: The Broward Experiment.” Justice Quarterly 19:343–75. Ford, D., and M. Regoli (1993). The Indianapolis Domestic Violence Prosecution Experiment. Washington, DC: National Institute of Justice. Garner, J., J. Fagan, and C. Maxwell (1995). “Published Findings From the Spouse Assault Replication Program: A Critical Review.” Journal of Quantitative Criminology 11:3–29. Pate, A., E. Hamilton, and S. Annan (1991). Metro-Dade Spouse Abuse Replication Project Draft Final Report. Washington, DC: Police Foundation. Pleck, E. (1989). “Criminal Approaches to Family Violence, 1640–1980.” In Family Violence, vol. 11, Crime and Justice: An Annual Review ofResearch, edited by L. Ohlin and M. Tonry, 19–57. Chicago: University of Chicago. Sherman, L. (1992). “The Influence of Criminology on Criminal Law: Evaluating Arrests for Misdemeanor Domestic Violence.” Journal of Criminal Law and Criminology 83:1–45. Sherman, L. (2010). “Forward.” Journal of Experimental Criminology 6:351–54. Sherman, L., with J. Schmidt and D. Rogan (1992). Policing Domestic Violence: Experiment and Dilemmas. New York: Free Press. Zorza, J. (1992). “The Criminal Law of Misdemeanor Domestic Violence, 1970–1990.” Journal of Criminal Law and Criminology 83:240–79.