YES
Janell D. Schmidt and
Lawrence W. Sherman
From Janell D. Schmidt and Lawrence W. Sherman, "Does Arrest Deter Domestic Violence?" in Eve S. Buzawa and Carl G. Buzawa, Do Arrests and Restraining Orders Work? (Sage Publications, 1996). Copyright @) 1996 by Sage Publications, Inc. Reprinted by permission.
DOES ARREST DETER DOMESTIC VIOLENCE?
During the mid-1980s, widespread concern about the incidence and prevalence of domestic violence led many big-city police departments to change radically the way they policed a crime that affects millions of women each year. The often-maligned "arrest as a last resort" tradition was replaced with written policies and state laws requiring arrest as the sole police recourse. Nationally, this enthusiastic shift generated, from 1984 to 1989, a 70% increase in arrests for minor assaults, including domestic. Yet, the movement to arrest batterers may be doing more harm than good. Research in six cities testing the "arrest works best" premise in deterring future assaults has produced complex and conflicting results. Police and policymakers are now faced with the dilemma that arrest may help some victims at the expense of others and that arrest may assist the victim in the short term but facilitate further violence in the long term.
The revolution in policing misdemeanor cases of domestic violence can be attributed, in part, to the 1984 publication of the Minneapolis Domestic Violence Experiment, the first controlled, randomized test of the effectiveness of arrest for any offense. Results from this endeavor were that arresting abusers cut in half the risk of future assaults against the same victim during a 6-month follow-up period. Alternative police responses tested were the traditional "send the suspect away for 8 hours" and "advise the couple to get help for their problems." The efficacy of each treatment was measured by interviews with victims and official records tracking the offense and arrest history of each suspect. Because arrest worked better than separating or advising couples, the authors recommended that states change laws prohibiting police from making warrantless arrests in misdemeanor domestic violence cases. They also advocated that replication studies be conducted to test the generalizability of the results in other cities with varying economic conditions and demographic complexions. But absent further research results, their recommendation to law enforcement was "to adopt arrest as the preferred policy for dealing with such cases, unless there were clearly stated reasons to do something else" (Sherman, Schmidt, & Rogan, 1992, p. 3).
Although the study authors opposed mandating arrest until further studies were completed, within 8 years legislatures in 15 states (including 1 in which a replication was being conducted) and the District of Columbia moved to enact laws requiring police to arrest in all probable cause incidents of domestic violence. This dramatic expansion of arrest practices has also been attributed to successful litigation against police departments who failed to arrest, to the recommendations of the 1984 Attorney General's Task Force on Domestic Violence, and to political pressure applied by women's advocacy groups.
It is not clear, however, how well these policies and laws have been followed or whether they have controlled repetitive acts of domestic assault. Observations of compliance of the Phoenix, Minneapolis, and Milwaukee police departments found that only Milwaukee officers consistently adhered to the policy. More important, the lack of labeling cases as domestic prior to policy changes renders attempts at before/after measures difficult. Further complicating evaluation or comparison efforts is the variable threshold for probable cause to arrest in incidents of domestic assault. In Wisconsin, only a complaint of pain is needed for police to effect an arrest; in Nebraska, visible injuries are required. Until 4 years ago, Florida law required the parties to be married or formerly married in order for the incident to be considered domestic.
What is known about the impact of police arrest policies relative to domestic assault is that the vast bulk of cases brought to police attention involve lower income and minority-group households. One reason may be a higher rate of domestic disputes among these groups; another reason may be a lack of alternatives short of police intervention that offer immediate relief. Although arresting thousands of unemployed minority males each year may assist the goals of victim advocates and provide a brief respite for the victims, the skepticism of many police and criminologists relative to the deterrent power of arrest still remains. The key question of whether other police alternatives could prove more powerful or whether the police could be effective at all led the National Institute of Justice (NIJ) to fund replication studies in six major urban cities.
Beginning in 1986 and early 1987, police in Omaha (Nebraska), Milwaukee(Wisconsin), Charlotte (North Carolina), Metro-Dade County (Miami, Florida), Colorado Springs (Colorado), and Atlanta (Georgia) began controlled experiments to replicate the Minneapolis findings. Each site was afforded leeway to improve the methodology of the Minneapolis study and to design alternative nonarrest treatments to build on its theoretical foundation. Researchers in all the cities sought to obtain a sample size larger than the 314 cases analyzed in Minneapolis in order to test for interaction effects among the various treatments. In MetroDade, for example, a sample of 907 cases was obtained so that researchers could compare arrest to no arrest, both with and without follow-up counseling by a specially trained police unit. In Colorado Springs, more than 1,600 cases were used to contrast arrest and nonarrest with immediate professional counseling at police headquarters or the issuance of an emergency protection order. In Milwuakee, the, police provided 1,200 cases for the researchers to test the length of time in custody--a short 2-hour arrest versus arrest with an overnight stay in jail, compared to no arrest. The experimental team in Charlotte included a citation response along with arrest, mediation, or separation treatments in its 686-case sample. Only Omaha followed the Minneapolis design with 330 cases but added an offender-absent window of cases to test the effect of having police pursue an arrest warrant.
The results from five of these six later studies (results from Atlanta are not forthcoming) have clouded the issue for police and policymakers, although some victim advocates remain strident in their view that arrest works best. Perhaps most striking is that none of the innovative treatments--namely, counseling or protective orders--produced any improvement over arrest versus no arrest. The citation used to notify offenders to appear at a future court date in Charlotte caused more violence than an arrest. Only Omaha broke ground and found an effective innovation in its offender-absent experiment. Offenders who left the scene before police arrived and whose cases were randomly assigned to the warrant group produced less repeat violence than did similarly absent offenders assigned to the nonwarrant group. The issuance of a warrant may have acted as a "sword of Damocles" hanging over an offender's head.
In short, the new experiments reported both deterrent and backfiring effects of arrest. Arrest cured some abusers but made others worse; arrest eased the pain for victims of employed abusers but increased it for those intimate with unemployed partners; arrest assisted white and Hispanic victims but fell short of deterring further violence among black victims. To understand these diverse findings and move toward a policy resolution, it is necessary first to focus on the effects of arrest compared to nonarrest, because that is the central issue for police and policymakers concerned with determining the most effective or appropriate police response (see Table 1).
One central finding is that arrest increased domestic violence recidivism among suspects in Omaha, Charlotte, and Milwaukee. Although these three cities produced some evidence of a deterrent effect of arrest within the first 30 days, victims found that this protective shield quickly evaporated and that they suffered an escalation of violence over a longer period of time. None of the followup measures produced the 6-month deterrent effect reported in Minneapolis. Some measures showed no difference in the recidivism of offenders arrested, compared with those whom police did not arrest.
Researchers in Colorado Springs and Metro-Dade found some support for the Minneapolis findings but only with limited measures. A narrow window of victim interview data (a 58% response rate in Colorado and 42% in MetroDade) confirms the deterrent power of arrest. But the less than ideal response rate might mean that victims who were interviewed were different from those who were not interviewed. Official records tracking recidivism in Colorado Springs did not uncover a deterrent effect of arrest, as some records did in MetroDade. Confounding the interpretation of the Colorado results was the fact that the vast majority of experimental cases (58%) were based on the offender's nonviolent harassing or menacing behavior toward the victim, perhaps distinct from the
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Table 1
Summary of Results of Six Arrest Experiments for Repeat Violence Against the Same Victim
Minneapolis Omaha Charlotte Milwaukee Colorado Springs Miami
6month deterrence
official measures Yes No No No No 1 of 2
6-month deterrence,
victim Interviews Yes Border No No Yes Yes
6. to 12-month
escalator, official
measures No Yes Yes Yes No No
6- 12-month
escalation, vicbm
interviews No No No No No
30- to 60-day
deterrence, official
measures (any or
samevicbm) Yes No Border Yes No 1 of 2
Escalaton effect for
unemployed _Yes _ Yes Yes
Deterrence for employed Yes Yes Yes
Note: = relationship not reported.
Source: From Sherman, Schmidt, and Rogan, 1992, p. 129.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
physical attack required to arrest for battery in the other cities.
The different results from different measures in these cities suggests, then, that arrest has a different effect on suspects from different kinds of households. This finding is best summarized by the following statement:
Evidence that the effects of arrest vary by suspect comes from Milwaukee, Colorado Springs, and Omaha. In each of those cities, nonexperimental analyses of the official records data suggest that unemployed suspects become more violent if arrested, but that employed suspects do not. This consistent pattern supports a hypothesis that the effects of criminal punishment depend upon the suspect's "stakes in conformity," or how much he has to lose from the social consequences of arrest. Similar effects were found in Milwaukee for unmarried versus married suspects; unemployed, unmarried suspects experienced the greatest escalation of violence after arrest. The unemployment result is the single most consistent finding from the domestic violence experiments, and has not been contradicted in any of the analyses reported to date. (Sherman et al., 1992, p. 17)
Could other factors explain this varying effect of arrest on different suspects in different cities? A comparison of the data on prosecution rates, level of victim injury, number of married couples, unemployment rate, and ages of the suspects across all studies showed no consistent variation between the two groups of cities finding a deterrent or escalating effect of arrest. The only major difference was that a larger proportion of black suspects was found in the "arrest backfires" cities (Omaha, Charlottej and Milwaukee), compared to the "arrest deters" cities (Colorado Springs, Minneapolis, and Metro-Dade). But this pattern is not consistent: One deterrent city (MetroDade) shared a similar rate of black suspects with a backfiring city (Omaha~ 42% and 43%, respectively,
How carefully should policymakers and advocates tread through this maze of diverse findings? Applying these results to crime control strategies is complicated by the dilemmas and choices they present. Urban legislators and police chiefs in at least 35 states can choose between continuing the status quo and not mandating arrest, a choice that will continue to harm some victims. They can also legislate arrest, a choice that may harm victims currently served by a lack of policy. Choosing between the lesser of two evils is best guided by the following summary of the facts and dilemmas gleaned from the domestic violence research published to date (see Sherman et al., 1992, pp. 19-20):
1. Arrest reduces domestic violence in some cities but increases it in others. It is not clear from current research how officials in any city can know which effect arrest is likely to have in their city. Cities that do not adopt an arrest policy may pass up an opportunity to help victims of domestic violence. But cities that do adopt arrest policies-- or have them imposed by state law-- may catalyze more domestic violence than would otherwise occur. Either choice entails a possible moral wrong.
2. Arrest reduces domestic violence among employed people but increases it among unemployed people. Mandatory arrest policies may thus protect workingclass women but cause greater harm to those who are poor. Conversely, not making arrests may hurt working women but reduce violence against economically poor women. Similar trade-offs may exist on the basis of race, marriage, education, and neigh boyhood. Thus, even in cities where arrest reduces domestic violence overall, as an unintended side effect it may increase violence against the poorest victims.
3. Arrest reduces domestic violence in the short run but may increase it in the long run. Three-hour arrests in Milwaukee reduced the 7% chance that a victim would be battered as soon as the police left to a 2% chance of being battered when the spouse returned from jail. But over the course of 1 year, those arrests doubled the rate of violence by the same suspects. No arrest means more danger to the victim now, whereas making an arrest may mean more danger of violence later for the same victim or for someone else.
4. Police can predict which couples are most likely to suffer future violence, but our society values privacy too highly to encourage preventive action. Largely because of the value our society attaches to privacy, especially marital and sexual privacy, no one has developed a recognized method, or even advice, for police to use in preventing domestic violence. A small group of chronically violent couples and incidents reported in apartment buildings produce most of the cases of domestic violence that police learn about, but the only policies now available react to the incidents, rather than to the patterns. Ignoring those patterns allows violence to continue; addressing them requires methods that many Americans would call invasions of family privacy.
Concomitant with these dilemmas is an even tougher question for officials charged with implementing effective policing strategies: Just how much research is enough to inform policy? The authors of the Minneapolis results were the target of much second-guessing and criticism from their colleagues over the reported findings and influence that the study enjoyed. Criminologists sought a more rigorous testing of the initial conclusions, perhaps foreseeing the risk of policy changes later proving to be unwise. Advocates, whose beliefs were validated by the results, and police policymakers, at least in Milwaukee, used the study to adopt arrest as the mandatory police response. In 1988, the Wisconsin legislature, perhaps less cautious than criminologists and motivated by ideological or politically pragmatic grounds, passed a law mandating arrest as the statewide response. This action occurred despite their awareness of the ongoing replication in Milwaukee testing the specific deterrent power of arrest. If a little medicine was good, a lot was even better.
The dilemma between limited research results and the need to do something about today's problems is also clearly illustrated by the Omaha offender-absent experiment. These findings may be far more compelling and relevant than the Minneapolis results because the offender is gone by the time police arrive in about half the cases brought to police attention. Yet, the study has had no observable influence on policy since its publication in an obscure journal. Modestly presented as a pilot study, no replications are being planned. Thus, there is little risk that the findings will inform policy and later be contradicted. In the meantime, assaults on thousands of victims could conceivably be thwarted if prosecutors heeded the policy implications.
Sherman et al. (1992) posited that:
the replication dilemma thus also poses a choice between two wrongs. Both using and burying research results entail risks of harm. But as Americans become more sophisticated about the scientific process, they may come to expect revisions of policy based on new scientific evidence in this realm of knowledge as in others. Americans are accustomed to constant revisions of findings about diet and disease. Cholesterol, sugar, caffeine, alcohol, jogging.. . the "latest" evidence about their relations to health and longevity has changed significantly and repeatedly over the last twenty years, and many people and businesses have changed their behavior in response. (p. 21)
To some the choice between two wrongs invokes despair and inaction. Yet, policing domestic violence may not be hopeless. Careful review of the policy implications, combined with the freedom to test alternative policies, can lead to more effective solutions. Use of the best information that Sherman et al. (1992, pp. 2~24) have to date guides the following five policy recommendations:
1. Repeal mandatory arrest laws. The most compelling implication of these findings is to challenge the wisdom of mandatory arrest. States and cities that have enacted such laws should repeal them, especially if they have substantial ghetto poverty populations with high unemployment rates. These are the settings in which mandatory arrest policies are most likely to backfire. It remains possible but unlikely that mandatory arrest creates a general deterrent effect among the wider public not arrested. Even if it does, however, increased violence among unemployed persons who are arrested is a serious moral stain on the benefits of general deterrence. The argument that arrest expresses the moral outrage of the state also appears weak if the price of that outrage is increased violence against some victims.
2. Substitute structured police discretion. Instead of mandating arrest in cases of misdemeanor domestic violence, state legislatures should mandate that each police agency develop its own list of approved options to be exercised at the discretion of the officer. Legislatures might also mandate 1 day of training each year to ensure that discretion is fully informed by the latest research available. The options could include allowing victims to decide whether their assailants should be arrested, transporting victims to shelters, or taking the suspects to an alcohol detoxification center.
3. Allow warrantless arrests. Whereas mandatory arrest has become the major issue in some states, warrantless arrest remains an issue in others. Sixteen jurisdictions have adopted mandatory arrest laws, but at last report 9 others have still not given officers full arrest powers in misdemeanor domestic violence cases that they did not witness: Alabama, California, Michigan, Mississippi, Montana, Nebraska, New York, Vermont, and West Virginia. The success of arrest in some cities suggests that every state should add this option to the police tool kit. Deciding when to use it can then become a matter of police policy based on continuing research and clinical experience, rather than on the massive effort required to change state law.
4. Encourage issuance of arrest warrants for absent offenders. The landmark Omaha experiment suggests that more domestic violence could be prevented by this policy than by any offender present policy. The kinds of people who flee the scene might be more deterrable than those who stay. A prosecutor willing to issue warrants and a police agency willing to serve them can capitalize on that greater deterrability. If the Omaha warrant experiment can be replicated in other cities --a very big if--then the warrant policy might actually deter more violence than do arrests of suspects who are still present. Because it will likely be years before more research on the question is done, such policies should be adopted now. They can easily be discarded later if they are found to be harmful or ineffective.
5. Special units and policies should focus on chronically violent couples. Because a limited number of couples produce most of the domestic violence incidents in any city, it makes little sense for police to treat all violent couples alike. It makes even less sense to frame the whole policy debate around responses to incidents when most of the problem is those chronic couples. The challenge is to develop procedures for violent couples that do not invade family privacy. Trial and error through research and development is required for any major breakthroughs. But an effective policy for dealing with chronic couples would have more impact than any other breakthrough. It deserves the highest priority in policing domestic violence.
The opposition to mandatory arrest laws presented here may frustrate or even anger many tireless advocates who have relentlessly grasped arrest as the preferred police response to incidents of domestic violence. To them, the suggestion that other institutions, such as shelters for battered women, treatment programs for victims and offenders, schools, and welfare agencies, may better serve victims is perhaps blasphemy. But they need not become too alarmed. However sensible that approach may be, the climate in many communities today is for law enforcement officials to get tough on crime. Regardless of the results of any scientific studies, the police will remain the primary institution coping with domestic violence among the poor and unemployed. This country's current fiscal crisis dooms any substantial investment in developing new programs in both the law enforcement and social services fields. The troublesome fact remains, however, that the punishment sought by advocates and community policymakers may encourage more crime.
REFERENCES
Sherman, L. W., Sclunidt, J. D., & Rogan, D. R (1992). Policing domestic violence: Experiments and dilemmas. New York: Free Press.